Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
Polestar. Discover true electric performance with the test drive at
Polestar Dublin, Postar two, Postar three and Pollstar four three
extraordinary expressions of design and innovation. Ready to change the
way you experience driving. Orders now for two five to
two delivery from forty four twenty five year ro Polestar
(00:23):
the way electric cars should be. Book your test drive
today at Polestar dot com.
Speaker 2 (00:30):
Tesco Club Card gives you the power de more hel
Can you make sumwherre meal time special? Bring the family
together with three for two across twenty three finest barbecue
meats like juicy and flavorful Irish Angus quarter pounders and
are made for sharing. Sweet and smoky Irish lam kababs,
or add something sweet and savory with our Korean style
Irish corn fed chicken tie kin kebabs. Get more from
(00:52):
your shop with your Tesco club card or app shop
in store or online. Tesco Every Little helps available at
most doores, prices very express.
Speaker 3 (01:06):
You're listening to Luke's English podcast. For more information, visit
teacher Luke dot co dot uk. Hello listeners, Welcome back
to Luke's English podcast. Here is another episode about artificial intelligence. AI.
In fact, using AI to help you learn English. We're
(01:29):
going to look at some specific ways that AI can
help you to do that in combination with this podcast. Now,
as I always say, listening to plenty of podcast episodes
for example, is a fantastic thing to do for your English.
But there are other ways that you can push your
English further, other ways that you can apply yourself to
(01:51):
your English learning, and I think that AI can definitely
make this easier. And in this episode, the idea is
to show you some specific ways that you can do that. Now,
AI is incredible, right, We live in an amazing time
thanks to the potential for learning that AI gives us.
(02:14):
But obviously artificial intelligence has its negative side as well.
There are controversial points relating to this, as I discussed
not long ago in episode nine hundred and nine. We'll
talk about some of those things as well. There'll be
a bit of a discussion about that, but the main
point is learning English with AI. How can you actually
do it? I'm joined today by Alista Budge from Leonardo
(02:38):
English dot com and Alistair has been working away looking
at specific ways that you can use AI as a
learning tool, and he's come up with a number of
very useful prompts and exercises, and he's going to be
kind enough to share them with us today. So there
should be lots of very practical tips and things to
(02:58):
learn from this episode, so get yours ready. As usual,
there's an episode pdf which contains a lot of notes
and prompts and a transcript of this episode which should
be very useful to help you to get the most
out of the episode. You'll find a link in the description. So,
Hello Alistair, how are you doing? Hello link?
Speaker 4 (03:17):
I am doing wonderfully today.
Speaker 3 (03:19):
How are you? Yeah, not bad? Thank you. I'm looking
forward to this. I'm looking forward to seeing what you've
come up with. So what have you been doing? You've
been just sort of playing with AI then, yeah.
Speaker 5 (03:30):
I guess like hundreds of millions of people around the
world and nothing particularly original, But have been trying to
see how I personally and how I as a language
learner can get the maximum benefit from AI tools, and
(03:50):
also as a podcaster and someone who makes podcasts to
try and help people improve their English. Have been, you know,
experimenting as much as I can with different ways in
which you can use AI tools like chatchipt in conjunction
(04:10):
with podcasts to kind of go one level deeper really
and help you learn more efficiently with podcasts. So yeah,
I've been I've been doing this for quite a long
time and have kind of created lots of different prompts
that I've found to be quite useful which people can
(04:31):
use in conjunction with any podcast episode to help them
improve their listening, comprehension, vocabulary, writing, grammar and so on,
and even a little bit of speaking. So it's it's
really been super interesting to play around with and also
see the kind of evolution of tools like chatchipt specifically
(04:55):
for the language learning, because they've come on a long way.
Speaker 3 (04:59):
Right, Definitely, Yeah, I mean, I guess it's mostly chat GPT,
although there are lots of other platforms, but I mean
I use chat GPT, I use others as well, like
Microsoft co Pilot and Google Gemini and Google Notebook LM.
There are all these different things. I've been playing around
(05:19):
with them a lot too. But yes, if we just
talk about chat GPT first of all, Yeah, I've definitely
noticed that it has improved. It keeps getting better, it
keeps learning, doesn't.
Speaker 5 (05:28):
It It certainly does, and it's it has become far
more useful from the perspective of a language learner as well.
Speaker 4 (05:38):
But I still think that lots of people.
Speaker 5 (05:41):
Are using it for, you know, just kind of text
generation type what tasks as in, write some text for
me that I can therefore use and often pass off
as my own. And what I've been really going quite
deep into is different ways in which you can in
(06:04):
which you can use let's just say, chatchapt rather than chatchat.
Speaker 4 (06:08):
And other AI tools.
Speaker 5 (06:09):
I just I'll refer to chatchabt to mean all AI
tools for the time being, rather than just adding a
disclaimer every single time. But yeah, ways in which you
can use chat shipt so you kind of act as
a kind of tutor when listening to a podcast episode
(06:33):
and just get a lot more out of it than
you would do otherwise. So it's been it's been really
really useful for me. I've been trying out with lots
of listeners and they the responses have been pretty pretty
positive so far. So I'm very excited to share what
I found with people listening to your show.
Speaker 3 (06:54):
Okay, great, So can you give us an overview of
some of the key ideas or key things you're going
to show us? Sure?
Speaker 5 (07:03):
Well, the maybe before I go right into that, I
just wanted to take one step back and talk about
podcasts in general, because I absolutely love podcasts. From the
point of view of learning a language. You always see
whenever there are interviews with successful language learners, polyglots, whatever,
(07:24):
they almost always mentioned podcasts. Clearly, we both believe that
podcasts can be very helpful from the point of view
of learning English. But when you listen to a podcast,
you're kind of on your own really because you are
you know, you've got headphones in, you are listening to
someone talking, and perhaps they are talking about vocabulary grammar
(07:47):
as well, or they're just giving you an example of
good English, good native level English, and then it's kind
of up to you to kind of copy that try
to try to kind of learn from that. And if
you are just kind of listening and wandering around or
doing the dishes, walking the dog or whatever, then you
(08:13):
you know, you're not really doing anything else with that
podcast episode, and I think you are kind of holding
yourself back a little bit from the from its maximum potential.
So how I like to think about Chatchip Team slash
Ai Tools in terms of learning with a podcast is
(08:36):
that they can be a little bit like a sort
of tutor who's listening to the podcast with you and
kind of guiding you, helping you pinpoint different interesting words
and phrases, helping you test your comprehension, helping you and
encouraging you to do little mini writing exercises, correcting your grammar,
(08:59):
that kind of stuff of all based around this this
podcast episode, and I think it's the kind of thing
that can be so helpful because it turns that slightly
passive activity of listening into some kind of short active
exercises that can help develop your comprehension, your your vocapary,
(09:21):
writing grammar, and so on. So that's kind of where
I see chatchy BT, slash AI tools kind of fitting
into the picture. And I really think that it doesn't
have to be as time intensive as lots of people
kind of listening to this might think, because you think,
(09:42):
oh God, now he is saying you've got to like
go to your computer and fire out CHATCHYBT and do
all these things. It's going to be super complicated. But really,
the things that I would like to talk about today,
it might take you kind of five or ten minutes.
So if you're thinking about listening to a podcast episodes
that might be half an hour, an hour, two hours long,
(10:04):
and then adding on a relatively short activity after that,
it doesn't take up so much time out of your date,
and I think it can be very helpful in kind
of joining the dots between what you've listened to and
what you're actually able to produce. So yeah, I'm excited
(10:25):
to talk about that with with with l EP in
particular because one thing, I was just reading some of
the comments on your on your YouTube videos before we
started this, and one of the things that kind of
the very common thing that people seem to say is like, Wow,
I love how Luke talks, and like I wish I
(10:46):
could talk like Luke.
Speaker 4 (10:48):
And you know, you've provided the world with.
Speaker 5 (10:52):
I don't know how many thousands of hours of examples
of how Luke talks, but clean not everyone who is
listening to your podcast is able to talk just like you.
So I think chat GIBT can be helpful in terms
of really kind of teaching people about different kind of
(11:15):
expressions that are used in the podcast, helping them understand
the kind of language that's used in a way that
would just be very difficult to do if you're on your.
Speaker 3 (11:25):
Own, so it's like an assistant sort of thing, like
an assistant, you can think of it like that. And
you're talking about in terms of the process. We'll get
into the details in a bit, but you're talking about
essentially like, for example, listening to an episode and then
after listening, just quickly using chat GPT to just activate
(11:51):
your learning a bit more. So you'd listen to the episode,
do something with chat GPT, and then kind of, you know,
do these exercises that you can talk about. So it's
kind of in that order because some people listening who
perhaps aren't so familiar with, you know, how it works,
might listen to that and think, oh, what, so you
mean somehow you can incorporate chat GPT while I'm listening
(12:15):
to the episode. No, you listen to an episode first,
and then you essentially give chat GPT the episode you've
listened to and ask it to do certain things, and
Alista's going to tell us what those you know, what
you ask it, the specific questions or prompts that you
give it, and then you kind of do the little
exercises that it presents to you. And that can be
(12:38):
either in written form texting like a text chat, or
it can be in spoken for because chat GPT these
days has quite a kind of functional voice chat voice
conversation mode, although the free version, I think that's kind
of limited. On your phone, for example, it will only
(12:58):
do that with you for a certain amount of time.
And obviously if you if you use the paid version,
then you've got you get a sort of unlimited amount
of talking time with it. Yeah, now, I before we
get into that stuff, and I want to focus on
that as much as possible, and we're going to talk
about things like using it to improve your listening comprehension,
(13:18):
to develop your vocabulary, to work on your writing, to
work on your grammatical accuracy, to practice speaking, and I
think other things like just helping you to be a
let's say, better listener to the podcast, like being able
to search my episode archive more effectively, or even having
(13:40):
a kind of conversation with chat GPT after having heard
an episode. So some of those things, but I feel
like I need to address certain things as well. There's
there's still a stigma attached to using AI. Have you
noticed this? Do you know what? People listening might not
(14:01):
know what a stigma is, but essentially a sort of
negative association, there is still like that slight sense of negativity.
Maybe that's never going to go away, you know, because
AI is such a complicated thing, and as we know,
there are all sorts of you know, controversial things relating
to it. But so I'll give you an example. Whenever
I'm teaching in a classroom with a group of students
(14:23):
and I mentioned chat GPT or show them something. For example,
I say, oh, you know, I asked chat GPT and
this is what it said. If we're discussing a language
point or discussing something or whatever, and if I go
to chat GPT and show it on the screen, there's
a sense in the room that people have a slightly
(14:44):
funny reaction to it. That's the best thing I can say,
meaning that they are maybe maybe people laugh, you know
at the idea that let's see what chat GPT says.
It's kind of funny to use it live in the classroom.
Sometimes there's a slight sense of like an intake of breath.
Oh like that, as if there's something kind of controversial
(15:07):
or negative about it, Like I think also some teachers
maybe don't want to admit that they use chat GPT
or some of us don't want to admit it. Why
would we not want to admit it because there's a
slight sense of stigma attached to it, and that would
be things like maybe it's associated with with with laziness.
(15:28):
It's a sign of laziness somehow, or it's considered in
some way to be a sign of cheating, like using
chat GPT to do anything is equivalent to cheating or
being lazy. There's also that association with academic cheating. We've
all heard about the stories of universities attempting to assess
their students and the students are all using chat GPT
(15:51):
to write their own essays. So that's, you know, considered
to be a very bad thing, you know, the effects
on education and assessment. There's copy issues relating to it,
and also just the quality concerns with chat GPT and
other software. Know what it produces, is it actually good?
Is there actually any value or quality to it? These
(16:14):
are all things that I'd like to discuss later on,
but I just wanted to mention those things at the beginning. Also,
there's the issue of using chat GPT as a language
learning tool. I think that it works best when you
do it in the right way. It can be a
very powerful tool. It can definitely help you learn your
(16:35):
learner language effectively and efficiently, but you have to do
it right, and I think doing it wrong, or using
chat gpt in the wrong way, can actually potentially be
harmful for your English, right. I think that learners of
English should think about this and what does this mean
using it wrong? I think examples of using chat gipt
(16:59):
in the in a way that could be harmful include
just like asking it to translate something directly for you,
using it to do all the work of producing some
written English which, as you said earlier alistair, which you
then pass off as your own. Just asking chat gpt
to write a reply to a certain comments in a
(17:19):
comment section, asking it to reply to a certain email,
and then just pasting what it says into what you're
writing your reports or essays or whatever. You know, that's
a kind of like short term way to use it.
It can be a short term solution, but that might
not be what's best for your English long term. And
(17:43):
what's wrong with this? If I can just say a
few more things before we get into the stuff you've
prepared alistair, you know you might become so used to
using AI to correct you or to generate correct text
for you, which you then just copy paste into whatever
you're writing. You may become so used to doing that
that you end up kind of doubting your own ability
(18:06):
or losing your own ability to do it yourself. You know.
So I think that AI should you should use it
to train you to be a better producer of English,
rather than just using it as a thing that just
does all the work for you, because you know, ultimately
this can lead to things like a lack of confidence,
a lack of ability, which which really hits you when
(18:29):
you're in a real world situation and AI isn't available.
It could lead you to sort of lose those productive
skills which are so vital for producing good English on
your own. And it can lead to passive learning where
you accept corrections from AI. Because this is another thing.
People say, Oh, you just you know, ask chat GPT
(18:50):
to correct your English, which is great, But again, if
you just accept all those corrections without really asking why,
or without really learning why the errors happen in the
first place, then you don't really maybe internalize English correctly,
don't really internalize the rules or patterns. Another thing is
(19:14):
that it could potentially lead to a loss of your
own authentic voice. It's very important to find your own
voice when you're learning a language. I think, and AI
English can be a little bit generic and characterless and
a bit overly formal, and it just might lack a
certain special personal touch, which I think is important when
making human connections. And also, finally, there is you know,
(19:38):
data concerns. You know, you know the fact that you know,
you end up sharing a lot of your personal data
and just simply spending all your time with AI means
that you spend less time with humans, which might turn
you into some sort of weirdo or something like that. So,
(19:58):
you know, So I want to have a balanced look,
you know, I want to acknowledge that there are negative
sides as well, and also acknowledge that I think there's
a right way to use AI and if you're using
it in the right way, it can really superpower your English,
but you have to do it, you know, correctly, And
I think I think what you're going to tell us
about is probably a set of really useful things that
(20:23):
can benefit your English.
Speaker 5 (20:24):
I hope so. And I certainly agree with everything you said.
And just to add one thing, is that I'm not
here to say that a I should replace any kind
of human teacher or anything like that. And if you
are in a situation where there's a human tutor who's
listening to episodes of louts English podcasts with you and
(20:46):
providing you with exercises, that's way better, right, that's a
much better solution. But for most people that's that's not
that's not a practicality, and it's it's a expensive thing
to be a very expensive thing to hire a private
user just to set your exercises for Luke's English podcast.
Speaker 4 (21:10):
Maybe it's a very worthwhile investment, who knows.
Speaker 3 (21:12):
I should also add that a lot of the time
my podcast episodes do come with additional material that there's
normally some sort of PDF these days, which includes often
a vocab list and often different exercises, maybe a vocab
quiz that you can use to help you practice, you know.
(21:32):
A lot of this time it's all included in a PDF,
and there's a transcript and stuff like that, and of
course there's neep premium as well, which is you know,
stuff that you can use. You know, it's a bit
like me teaching you more directly in those episodes. But yeah,
let's have a look at some specific things then that
people can do alistair. What have you got for us first?
Speaker 5 (21:54):
So first I just wanted to give a little bit
of background to how to how to set this thing,
how to set it up really, because what you will
need to do before starting working with any episode is
you know, to give chatchipt the context of that episode,
(22:16):
and the easiest way of doing that is just by
sharing the entire transcript with chatchipet. So maybe I'll just
kind of start sharing my screen now and I can
show I can show people watching this on YouTube how
it's working. And for people listening to the audio version
of this, I'll just talk you through what I'm doing.
(22:38):
So don't worry if you're if you're just listening on audio. Okay,
let me just.
Speaker 4 (22:43):
Do this now.
Speaker 5 (22:45):
So I've just got chatchipt up here. I'm just going
to go this one GPT four Oh, and what I
am doing. I've just gone to YouTube and I'm just
going to copy the entire trans scripts from one of
your episodes, episodes number nine hundred and twenty eight, which
(23:06):
dedicated Luke's English podcast fans will know is the Abduction
Learn English with this short story so I am going
to go and I'm going to paste that in there.
Speaker 3 (23:17):
Wait, so how do you get the transcript because some
people don't know how to do that.
Speaker 4 (23:21):
Okay, so I will just I should say this.
Speaker 3 (23:25):
Is on YouTube, where there are automatic subtitles available for episodes,
and either there are like non automatic subtitles, like people
have added subtitles into the episodes manually in some way,
or YouTube's standard voice to text automatic subtitles, and you
(23:46):
can access them on pretty much any YouTube video unless
the subtitles have been switched off, you can access them,
and you can copy and paste them. And a lot
of the time these automatic subtitles are really very accurate,
especially these and especially when it's one person, especially when
it's one person talking on their own. Yeah.
Speaker 5 (24:09):
So I'm just showing my screen now and I'm on
the episodes on YouTube. I've just gone down here, and
there is a button that says show transcript. I click
on that and it takes me up to this bit
here on.
Speaker 3 (24:25):
The right top right and corner of the screen. The
little window opens up with the with the transcript written
there exactly.
Speaker 5 (24:32):
And I'm just going to take the whole thing and
I go down here and I'm going to copy it.
Speaker 3 (24:37):
So it's just copy pasting the entire thing into.
Speaker 5 (24:41):
Chat chipet and I'm just gonna I'll just switch my screen, sorry,
and I'm pasting it in there. But as I've just
pasted it into the chatchpt textbox. But an important thing
that I will need to do as well is tell
chatchapt why I've done this, because if I've just given
(25:02):
chat you put a transcript of a podcast episode, it
won't really know what I'm trying to do. It might
be a little bit confused and little sort of say, Okay,
what do you want me to do with that? So
what I am going to do is I'm going to
I'm going to pace something else and I'm going to
read it out to you. And I think, Louke, this
(25:22):
is going to be in the episode PDF, so people
can just copy it directly from there.
Speaker 3 (25:29):
It's this so above above that within the same sort
of chat box as the script above the transcript you're
going to type in your prompt.
Speaker 5 (25:40):
I've typed in this, which is I want to use
this podcast transcript to improve my English. I will past
the full transcript below. Once I've done that, wait for
me to tell you what I want to do next,
do not respond until I give you an instruction. And
then below that is just the entire transcript from the
(26:02):
podcast episode.
Speaker 4 (26:04):
So I will just share this.
Speaker 5 (26:08):
I just press send and hopefully, okay, good, chat gipt
has come back and said, got it. Let me know
what you'd like to do with the transcript. So this
is this is a very important step because now chat
ChiPT has the entire context of the of the podcast
(26:28):
episode and we can start to work with it to
do whatever we'd like.
Speaker 3 (26:34):
Okay, very good. Yeah. I should also add that again,
you're using a paid version of chat ChiPT here, and
if you're using the free version, it might not accept
all of that text. I think it sort of limits
it to a certain number of characters, So you know,
you just might you might have to play around with
it a little bit and see what the what the
limitations are in the free version versus the paid version.
Speaker 4 (26:57):
But you're right, I've got the what's it called, the
I think the.
Speaker 5 (27:01):
Plus version, which is I think it's twenty twenty dollars
a month, which for me is very very worth it.
I don't want to sound like a CHATCHYPT salesman, but
I get a lot of value out of it. I
think with the free plans you have certain daily limits.
(27:21):
I'm not quite sure exactly what they are, and they
change all the time, so perhaps by the time you're
watching this it'll be slightly different. But the idea is
that we want to paste the transcript of the podcast
into CHATCHBT and give it the context of why we're
doing that. The episode I chose is a relatively long one.
(27:47):
I think it's an hour and fifty one minutes, so
it's quite long. But of course you can choose shorter ones.
You can choose to only work with a part of
an episode as well. The important thing is to provide
the transcript of what you would like to work with.
Speaker 3 (28:05):
Yeah, mm hmm. Okay, so now it's ready. So you've
given it the transcript, you've said to it. Right, So
here's a script. Don't do anything yet. Let me just
wait until I tell you what I want to do.
And chat Chi's pt has said, right, got it, let
me know what you'd like to do with the script.
And so this is where you add in your next
(28:25):
prompt exactly.
Speaker 5 (28:27):
So if we use the analogy of a friendly English tutor,
the tuser has now listened to the episode and it's
sort of ready to ready to help us work with it.
So the first kind of category of activities that I'd
(28:47):
like to talk about is just to do with listening comprehension,
because it's sort of the most obvious thing that the
people might think about developing when they're talking about listening
to podcast. Because you listen to a podcast, you develop
your listening. But one of the reasons that I think
(29:08):
it's important is that it sort of forces you to
listen more actively then you might otherwise have listened. As
if you're just wandering along listening while you know, walking
by the river or or doing whatever you might be doing,
it's easy to to not focus on what you're actually
(29:28):
listening to. But if you sort of know that, oh, teacher,
cha actually PT is going to ask me some questions afterwards.
Speaker 3 (29:34):
Then.
Speaker 5 (29:37):
Part of you, especially in me, the kind of lazy
learner in me, thinks, okay, well I better I better
really listen carefully if there's gonna there might be some
questions coming up. So that's why I think it's quite
quite a kind of useful thing to do. Now I'm
going to go back and paste something else in here.
I'll read it out as I as I paste it.
But I'm just gonna go back to my notes and
(30:00):
find the first activity. So I'm going back into chat GBT,
and I have pasted in this and again this is
going to be in the PDF, but I will read
it out now so it says you are my English tutor.
I've just listened to a podcast episode and pasted the
transcript above. Number one, ask me five comprehension questions based
(30:23):
on the episode, to wait for my answer after each one. Three,
if I'm wrong or stuck, give me a small hint
before telling me the correct answer. And four after all five,
give me a quick summary of how well I understood
and suggest one area to improve. Now, before I send
(30:44):
this to chat GIBT, I want just to highlight one thing,
which is the third point. If I'm wrong or stuck,
give me a small hint before telling me the correct answer. Now,
I think kind of starting to think a bit like
this way you're interacting with CHATCHYPT can be a really
useful thing because if you think about what a good
(31:07):
teacher would do, it's not just immediately tell you the
answer or say you're wrong or you're right, They would
kind of nudge you, kind of push you gently towards
getting it for yourself, towards finding out the answer for yourself.
So I think this is a really useful way for
people to start thinking about their interactions with CHATCHYPT saying,
(31:30):
don't just tell me the answer, give me a few hints,
help me kind of on my journey to figuring it
out for myself, and then, you know, tell me the
answer if I get completely stuck. So with treating this
more as an interaction with a tutor who's helpful rather
(31:50):
than just a simple kind of question, question and an answer.
So I'm going to go and press send. And this
is obviously life, so let's hope it's all working.
Speaker 3 (32:05):
Okay.
Speaker 5 (32:06):
I have to confess that I haven't listened to the
whole episode, so I don't know the answers.
Speaker 3 (32:12):
See if I can remember, let's.
Speaker 5 (32:13):
Say, so I'll read this out. CHATCHPT said, great, let's
get started. Question one, what was the main reason Jose
Martinez moved to the UK?
Speaker 3 (32:23):
So I think the main reason? So what the learner
at this point would do is right in the answer. Okay,
So let's say the main reason Jose moved to the
UK was to improve hearing his English by immersing himself
in a place where English was spoken. Very good, I
will immerse himself in English.
Speaker 5 (32:45):
Let's say, okay, I'll just say to immerse and I'm
actually going to add a little grammar mistake there too,
to see if Chatchipt picks up on it. I can
say to immerse herself in English and see if it
corrects me on that.
Speaker 3 (33:01):
Afterway, can you add a preposition mistake as well? So
can we say to immerse herself to English? Why not to.
Speaker 4 (33:11):
English? Okay?
Speaker 5 (33:12):
So what was the main reason Jose Martinez moved to
the UK? And I am going to answer to immerse
herself to English, and I'm going to send that now.
And chat Chipet has responded saying, almost right, You've got
the idea, but there's a small mistake with the pronoun hint. Remember,
(33:33):
Jose is a man. Would you like to try again?
Speaker 3 (33:36):
Okay?
Speaker 5 (33:38):
And so it's picked up. It's given me a hint
that I've used the I've used the wrong pronoun that
it shouldn't be herself. So I might think, okay, great,
it must be himself. But I will keep the same
mistake that Luke asked me to add and let's see
what happens here? Do you immerse himself to English? So
(34:01):
I've just typed in to immerse himself to English, and
let's see what happens now, Okay, close, You've got the
right pronoun. Now, just one small adjustment needed with the preposition.
We don't say immersed to English. Would you like to
try again? And luky UK for me.
Speaker 4 (34:20):
Just to put the correct one?
Speaker 5 (34:21):
Yeah, go ahead, okay, we'll go to immerse himself in English? Okay,
and chat Gupta said perfect, Yes, Jose moved to the
UK to immerse himself in the English language and culture.
Question number two, what unusual event happened to Jose while
(34:41):
he was closing up his garage one evening. Now, I
imagine we don't have to go through the through the
entire quiz, but hopefully people listening to this and watching
it can get an idea of how this would work
in practice.
Speaker 4 (34:58):
We'll go through the entire quiz.
Speaker 5 (34:59):
And and then after question five, Chat Gpt will hopefully
give us a summary of how well we've done, perhaps
highlights any grammar mistakes, corrections with our English and so on.
So that's that's just one very simple example of how
(35:19):
if I'd listened to that whole episode and I thought, right,
I'll just try and test how well I've how well
I've understood it, how I can practice my listening comprehension
and also get a kind of free little grammar grammar
lesson there too, based on the mistakes that I've made.
Speaker 6 (35:41):
This May, there are nine must be one prizes of
two hundred and fifty thousand euro for Irish euro Millions players,
not maybe one must be one in May play euro
Millions for your chance to be one of nine guaranteed
winners of an extra quarter of a million euro the
National Lottery. It could be you play responsibly.
Speaker 7 (36:00):
Onla Sonri and Horteleisahershtak Horses Umproscula or Nina de Bosher
and Naseehina and Sha Mehev Dunig and So and Hortelishers
Rig boss Er and punk E tal Slash School transport
(36:21):
Nelen Aris in the ticket the Skullle and my younger
artist Jane, my ticket Don Proscola or Falla Hi the
Lina wilts naheren.
Speaker 3 (36:33):
M So. I would like to say at this point
that this shows that as always being a sort of
independently minded learner being a sort of intelligent learner who
applies themselves to the learning process. That pays off, and
taking responsibility for your own learning again, it pays off.
(36:54):
So you you might need to take the initiative two.
For example, in the case of episode nine to eight,
the abduction, you might need to think to yourself, right, well,
actually this is a long episode. But thinking about it,
what Luke has done is he's told us a story,
and then he's explained the story, and then he's gone
into a lot of vocabulary teaching. You know, the episode
(37:16):
is in those sections in fact offer and on YouTube
there are chapters which show you. So what you could
do is, if you want to focus on listening comprehension,
you would take a section of the episode and you
can find the time codes in the transcript. You can highlight,
copy paste just that section. For example, here it would
(37:37):
be the story of Jose Martinez. You'd put that into
chat GPT and then ask it to give you listening
comprehension questions based on that, and that's going to maybe
focus you more specifically on your understanding of the story
as it was told to you. And then the vocab
teaching part. That's not so much a listening comprehension task
(37:58):
for me is more like question of how much of
the vocab do you understand? How many of the things
I taught did you remember? And how can you actually
practice and use them. You might need to look at
other tasks for that section, but yeah, you need to
kind of think in a sort of analytical way a
little bit. Take the initiative and you know, make decisions
(38:20):
about how you ask chat GPT to give you listening
comprehension practice, namely choosing just a certain section of the
episode to focus on for listening comprehension.
Speaker 5 (38:31):
Yeah, precisely, and everything that I will kind of show
and talk about today I hope listeners will take as
sort of this is one example of how it can
be done. Now take this and think about it for yourself.
You can just copy and paste the prompts that will
(38:53):
be in the in the episode PDF, but you know,
think really critically about what you're trying to get out
of the episode. If it's if it's one of your
sort of short story episodes, then you probably don't need
the entire transcript. If you're just just listening for kind
of comprehension of the story part of the episode. Similarly,
(39:18):
if you're working, if you're sort of trying to develop
your vocabulary, as we'll talk about talk about a little
bit later on, then you probably want to prompt CHATBT
in a way that tells it that part of the
part of the transcript focuses on certain vocabulary words or
structures that Luke is talking about, and therefore that's what
(39:41):
you want to focus on.
Speaker 3 (39:43):
Yeah. And also another thing to consider is that you
could if we just take for example, an ielt's listening
exam as an example, one of the things they do
there is in the different papers, in the different parts
of the exam, they are testing different listening skills. So,
for example, different listen skills could be understanding the general
meaning of what was said right that's listening for gist
(40:05):
or being able to find specific information from what you
listened so facts, figures, names, dates, other specific details, and
also understanding the general tone of what you heard, for example,
how did the speakers feel about the subject they were
talking about. These are all things that you could ask
(40:25):
chat GPT to test you on. You know, if it's
a conversation with a guest, you could ask it to
focus its questions on your ability to understand let's say,
the humor that was at play, or maybe how the
guests felt about the things they were talking about. Right,
So you can ask it to kind of focus on
(40:47):
those specific things as well. Yeah, okay, certainly.
Speaker 5 (40:52):
And if you know you have a particular area of
the you know, the the ails listening is a problem
area for you, then create exercises based just on that
particular skill so that you're you're working on that and
not working so much on skills that you're already at
at a decent.
Speaker 3 (41:12):
Level at Yeah, there's that you can go to. By
the way, you can go to ils dot org and
find preparation resources and a guide for the test, which
will tell you, for example, what different skills are being
tested by different sections of the exam. It's quite a
good idea to look at that stuff in order to
be to become literate in the language of eyels, which
(41:38):
by extension means becoming literate in the language of you know,
the skill set of English.
Speaker 7 (41:45):
You know.
Speaker 3 (41:46):
Anyway, let's carry on, then, Alistair, what else have you
got to show us?
Speaker 5 (41:50):
I can show you one other kind of quite fun
listening comprehension style one. I can probably add some more prompts.
They will be in the PDF that we might not
talk about in the episode, just so people have got
more examples of how you could do this. But let's
just talk about one more listening comprehension one. So I'm
(42:11):
going to I'm just going to write into chat GPT,
let's stop this this particular exercise, and I'm going to
paste in something else that I will read out now,
and that is based on the transcript I provided. Choose
(42:32):
a short paragraph that contains useful vocabulary and is slightly
challenging for a I've put B two learner, but of
course people can adapt this as they see fit. Then
I have said number one, remove six to ten keywords
to create a fill in the blank exercise number two
present the blanketed version first, and number three after I
(42:54):
try to complete it, show me the correct answers and
explain any tricky vocabulary. So I will just press send
there and let's see. Let's see how we get on.
So you'll see that I'm not going to read out
the entire piece of text from CHATCHIBT, but what it
(43:16):
has done has created a fill in the blank exercise.
It's provided us with the timestamp saying approximately eleven minutes
and thirty around when Jose encounters the aliens for the
first time, and we have the fill in the blank
exercise there. So what this means is it's very easy
for me to go back to the podcast episode to
(43:40):
just go to eleven minutes and thirty and press play
and then try to complete this fill in the blank exercise.
And of course, like before, Chatchibt will act as a tutor,
giving us hints and it'll explain the harder words for
(44:01):
us and obviously help us with any mistakes that we made.
Speaker 3 (44:06):
So what you have to do is sort of type
out the paragraph, including the missing words. And I'm just
looking at the paragraph. It's like Jose. Jose tried to
sit up, but his body felt hmmm and his head
was hmm, like he'd just come back from a trip
to the pub with his English mate Barry, so if
you remember listeners, his body felt heavy and his head
(44:27):
was spinning. So this is quite good that it's picked
out those things because those are quite sort of meaningful,
useful bits of vocab to try to recall. That's definitely
a useful thing to do and are there numbers? So
the gaps are not numbered. But as you're writing here, Alistair,
(44:48):
I can see that you're putting number one heavy, number
two spinning, so you could number them. Chat GPT is
going to understand what that means as instead of just
typing up the whole paragraph again, you could just put
number one heavy, number two spinning, and so on. That's
a useful activity. Those are probably words that I explained
as well later on in the episode. I certainly remember
(45:10):
explaining his head was spinning. So that's a useful thing
to do. And it's it's just randomly picked that paragraph, right, Yeah,
just did that randomly.
Speaker 5 (45:22):
It just completely randomly picked that that paragraph. I'm sure
if listeners tried this at home, they might not get
the same paragraph. I'm sure if you were just trying
it now, you probably wouldn't get the same paragraph either.
So it's a kind of useful way to interact in
you know, just in a deeper way than people might
(45:44):
otherwise be otherwise be doing.
Speaker 3 (45:47):
Yeah, I would say sorry, you could all if you
wanted to, You could pick a certain paragraph and ask
it to do that, you know, input the paragraph and
give the same prompt and it'll do that. You know,
if there's a specific section that you want to focus on,
and you could probably ask it to be more specific
and say, remove any adjectives or remove all the prepositions,
(46:08):
or you could say, you know, gap out the verbs
and I'll have to put them all back in. You know,
you could probably get it to do those things too,
or you just say, just pick out some as you said,
vocab suitable for a challenging for a B two learner
or something.
Speaker 5 (46:23):
Yeah, exactly, should I just should we just see if
we can get it to do one of those more
specific exercises now to see Yeah, yeah, let's strue, shall
we see? Okay, so let's just do this, do this life.
Let's say, let's move on from this fill in the
blank exercise, create another fill in the blank exercise that
(46:48):
takes a part of the story, not Luke's X vocabulary explanations.
I struggle with let's say verbs, Should we say yeah,
verb tenses, verb verb forms, verb forms, and would like
(47:12):
to focus on this, Should I say, choose the most
complicated part of the story.
Speaker 3 (47:18):
How about this? How about this? Find a section which
includes lots of gerunds and infinitives. I don't know if
it'll be able to do that, but let's see.
Speaker 5 (47:33):
And then then I've said, use the previously defined format
for the exercise.
Speaker 3 (47:41):
Now, listeners, if you're thinking, well, I can't. You have
to be very articulate to be able to get chat
GPT to do what you wanted to do. But you know, again,
check the PDF because a lot of this stuff is
written out specifically and has been already written by Alistair,
and you can just essentially copy paste those things.
Speaker 4 (48:00):
Exactly.
Speaker 5 (48:01):
And hopefully what this demonstration is showing is some of
the different ways you can just take these excises and
modify them based on your own goals, your level, and
what you want to be doing. This is also live,
so let's see. Let's see what we come up with here. Okay,
(48:22):
I just read that out again for the benefit of
people listening. First I wrote, let's move on from this
fill in the blank exercise just to chalel jackshpt that
I wanted to do something else. Then I said, create
another fill in the blank exercise that takes a part
of the story brackets, not Luke's vocabulary explanations. I struggle
(48:45):
with verb forms and would like to focus on this.
Find a section, which includes lots of germs and infinitives.
Use the previously defined formats for the exercise. Perhaps Luke
I should also say and blank out these words.
Speaker 3 (49:05):
Yeah, yeah, to make.
Speaker 5 (49:06):
Sure that chatchipt knows that we're we want to focus
on the germs and infinitives, and that's what we need to.
Speaker 4 (49:13):
That's the words we need tofined.
Speaker 3 (49:15):
It's very specific, this one, so I'm curious to see
if it will do what I wanted to do.
Speaker 4 (49:20):
Let's see, Okay, let's just wait for it to finish.
So it has says it's done.
Speaker 5 (49:31):
It He's a fill in the blank exercise of focusing
on gerons and infinitives from a story section, not the
vocabulary explanations has given us the timestamp approximately twenty one
minutes forty five. When how is it interacts with the
aliens and shares jugs.
Speaker 3 (49:46):
Can I read this out?
Speaker 4 (49:47):
Yes?
Speaker 3 (49:48):
Yes, So the paragraph it's chosen is another section from
the story, and it goes like this. Despite their differences
and the fact that they had basically kidnapped him, Jose
found himself growing fond of his captors, appreciating their wisdom
and the peacefulness of their society. Despite their complete lack
of humor and their insistence on blank. And the word
(50:09):
is probe. So, despite their lack of humor and their
insistence on him as part of their research, he began blank.
The verb is share. He began share stories, music, and
literature with them. So, listeners, what form should these words take.
I'm quite impressed because it has done more or less
what I wanted it to do here. So it's despite
(50:30):
their insistence on probing him. So that's a gerin form
because normally after a preposition you would have a gerund
form of a verb. They insisted on probing him. Despite
their insistence on probing him, right, he began sharing stories.
That's because after the verb begin, we typically if it's
(50:52):
another verb following it, it would be in the I form.
So they began sharing stories then. Also we've got He
shared music literature with them along with his recipe for
how the perfect paella. The verb is make how to
make the perfect piela? So yeah, it's done the job.
(51:16):
So if you listeners, if you do the exercise, you would,
you know, just list the verbs in the correct form
in the order that they should appear. But you could
also ask shaut Gpt to explain its answers. You know,
you could say, can you give you the answers and
explain why they are these aren't? How would you put that?
(51:40):
Can you explain the answers?
Speaker 4 (51:42):
Yeah, you could say what I'm doing now, I'm going
to write the.
Speaker 5 (51:48):
Answers here, and then afterwards maybe I will also say,
please let me know how many I got right and
explain any errors that I made. But what I was
also going to do is you said that the second one,
the answer was he began to share story.
Speaker 3 (52:07):
I began sharing, But yeah, we began to share that, So.
Speaker 5 (52:10):
I was going to put the wrong one, but one
that was that also made grammatical sense, and see what
happens there. So yeah, so what I've written is despite sorry,
he began to share stories, even though in the transcript
and the original version it is it began sharing. And
(52:33):
I'm going to say, can you or so sorry, how
many did I get right? And please explain any I
got wrong? That's not wonderful English. But let's let's see
what it.
Speaker 3 (52:51):
Says, right. So, just to be it's written quite a
lot of stuff here, which is quite a lot for
us to read out, but essentially the ones that it
considered to be grammatically correct. It's given a green tick.
So for example number two to share, it says correct.
(53:11):
The verb began is followed by the infinitive to share
in this context, as it indicates the start of an action.
Now it doesn't do what I would do as a
teacher in this situation when giving feedback on the task,
which would be immediately to say yes, you could say
began sharing or began to share. They're both correct because
begin is one of those verbs that actually takes both
(53:33):
forms without a difference in meaning. But it hasn't actually
done that, you know, maybe it's no big deal, but still.
The one that you got wrong was number five, having
I don't remember the specific context of that it was.
If you go back up to the text.
Speaker 5 (53:54):
It also seems to be quite confused, says incorrect. The
correct answer is having. So your choice is actually correct,
but I want to explain why, just to make sure
you're clear.
Speaker 3 (54:02):
Right, So hold on a minute. So it's so you
you put for gap number five, you put the verb having.
Can you go back up so I can see what
that sentence actually was. Although the zilara showed no signs
of yeah, it should be having showed no signs of
having any sense of humor, so having is correct, and
you wrote having, but it's marked it incorrect. So this
(54:24):
is like a little bug. This is kind of made
a mistake here. It said it's incorrect, even though in
fact so it's instantly contradicted itself. It says incorrect. The
correct answer is having, which is exactly what you wrote.
So your choice is actually correct, but I want to
explain why, just to make sure you're clear. So going,
(54:46):
you got the right answer, but I just want to
make sure that that wasn't pure lucky. The phrase showed
no signs of is followed by a gerrand. We say
show signs of having something or show signs of being something.
In this context, it means the aliens didn't display any
indication of possessing a sense of humor. So that's odd.
(55:07):
It's like it's kind of got marked it wrong just
to kind of make itself feel useful.
Speaker 5 (55:14):
So we have a we have a live demonstration of
how CHACKGBT. It is not quite there yet in terms
of getting everything right, but it's it's better, it's better
than nothing, but but always always verifying that you you know,
because it can make mistakes as you've seen here.
Speaker 3 (55:35):
But what it has done, it hasn't given that the
mistake it's made is fairly negligent. Negligible. I mean, it's
not that serious because the response it's written is still correct.
It's not teaching you anything that's not correct. I find
that CHAT, GPT and other you know, generative language models,
(55:56):
they almost never make language errors. I've never ever seen
a grammatical mistake made by a language model like this
because it's almost impossible for them to do it, because
that is the basis of their programming, isn't it. It's
based on language, so.
Speaker 5 (56:16):
Large language model, so it's it's just it's not going
to make mistakes when it comes to language. This is
just some kind of generation eraror where it's it thinks
it's incorrect, but it knows it's actually correct.
Speaker 4 (56:28):
So it's got a little bit confused.
Speaker 3 (56:30):
But the grammar is right, the explanation it's given is correct,
but it's just yeah, it's logic, this is incorrect, and
then it says, yeah, the correct answer, actually your choice
is actually correct. But I wanted to just explain it
to you anyway, because I wanted to feel like I
was at least doing something useful. All right, there we go.
Speaker 5 (56:52):
Okay, so we have those as as sort of a
few examples of how you can create exercise is based
on listening comprehension. Do you want to go John to
do more and talk about kind of vocabulary development stuff
or do you want to move into more kind of
(57:12):
general discussion of.
Speaker 4 (57:16):
Chat GBT.
Speaker 3 (57:17):
That's true, Let's do a bit of stuff on vocab development,
because I think this is a really important useful thing.
What we want, right, what we want is for chatch
ept to help you two a help you check that
you've understood correctly, and we've kind of looked at that,
(57:38):
but also to help you essentially broaden your range of vocabulary,
and that includes finding exercises, finding ways of helping you
to notice, understand, maybe record or remember and recall meaning action,
(58:00):
not just remember it in your head, but actually produce
new vocabulary and then actually to personalize it to use
it to express yourself. That's the sort of process that
we want. So let's let's have a look at some
of that.
Speaker 5 (58:16):
Sure, what I think I might do is just choose
a different episode, given that in this episode you said
there's quite a heavy focus on vocabulary explanations, So maybe
I'll choose one where there are it's a kind of
natural conversation, and we can try to try to use
that as a as a basis for this.
Speaker 3 (58:37):
Let's do nine hundred and twenty one. From Comedy to
Cozy Crime with Ean More. Let's try that nine hundred
and twenty one. I was talking to a friend of
mine who has moved in his career from being a
stand up a comedian to being a writer of crime stories.
He writes books, he's published. You find his books in
(58:59):
all the main books shops and stuff now. So it
was a conversation about that.
Speaker 4 (59:04):
Okay, excellent.
Speaker 5 (59:05):
So what I'm going to do, I'm just going to
refresh the chat because Chatchibt doesn't need the context of
the previous episode. And what I'm going to do is
what I did before, which is I've taken the entire
transcript and I'm going to paste in the kind of
(59:25):
preparation prompt which I would just do now. So again
that is, I want to use this podcast transcript to
improve my English. I will paste the full transcript below.
Once I've done that, wait for me to tell you
what I want to do next. Do not respond until
I keep you an instruction and below that. I have
just pasted the entire transcript from episode number nine and
(59:46):
twenty one, so.
Speaker 3 (59:47):
I'm going to do that. What I'm thinking here is
that as well as the conversation with Ian in this episode,
there was quite a lot of other stuff from me,
So I spent about fifteen minut and it's at the
beginning of the episode setting a bunch of comprehension questions,
like loads of them, so there's quite a lot of
(01:00:07):
just question asking at the start, then a conversation, and
then at the end. I spent probably about half an
hour in fact, giving answers to all the questions I
set at the beginning. So that's all going to be
included as well. So maybe the smart learner would think,
right well, having listened to that episode, I will choose
the script for the conversation. But whatever, we're going to
(01:00:31):
load in the whole thing and we'll see what happens.
Speaker 4 (01:00:35):
Cool.
Speaker 5 (01:00:36):
I'm going to paste in this prompts here which I
will read out now, which reads, based on the transcript
I pasted, find ten natural sounding expressions that native speakers use,
but that be one level learners like me might not.
Of course, we can adapt this level as we see fit.
(01:00:59):
And I've said, for each expression, ask me what I
think it means, then explain the meaning and how it's
typically used, ask me to write a new sentence using
the expression, and give feedback after each one. One thing
I just wanted to kind of draw listeners attention to
here is the Is this the ten natural sounding expressions
(01:01:22):
that native Use speakers use, but the B one level
learners like me might not? Because I think this people
often tend to think of vocabulary in terms of like
this is a really advanced word, and this is a
basic word, so you know there is a basic word,
and I don't know, psoriasis or something is a very
(01:01:44):
advanced word that it's probably also difficult for most people
to spell. But I think there's this kind of almost
like second category of words and expressions that are are
things that are not complicated, necessarily not sort of super
advanced or rare, but are the kind of natural expressions
(01:02:07):
that native speakers or advanced English learners will find themselves
using a lot, but you don't often find in kind
of textbooks or or kind of like formal learning so much.
And podcasts can be a real sort of gold mine
when it comes to these, especially these more conversational style ones.
(01:02:31):
So in my experience, chatchipt is pretty good at finding
these kind of expressions that people listening to the podcast
might not necessarily have picked up on and thought, oh,
that's that's a really interesting way that Luke used to
introduce that subject, or you know, that's it's the way
(01:02:53):
that his guest kind of explained that in a different way.
That's that's interesting. So it's helping people pinpoint these kinds
of expressions, explain them, and getting them to use them
for themselves.
Speaker 3 (01:03:10):
Yeah great. Can I just add one maybe one point
in your list in your prompt so you've got to
ask me what I think it means. You know, for
each expression, ask me what I think it means. Then
explain the meaning and how it's typically used. And third
asked me to write a new sentence using the expression
I feel like I'd like to I'd like it to
include the expression in context. So include a sentence with
(01:03:34):
the expression in context. That's what I want chat gpt
to present to me as a learner. I don't just
want a definition, so it would be like I want
chat GPT ultimately to say, here's something that Ian said
near the beginning of the conversation, and then it includes
a sentence or two containing the phrase. Because I always
(01:03:56):
think that learners need to see vocab in a grammatical
context or cotext as it's actually known. They need to
see it included in a sentence right in order to
see because you know, it's vital to see language in
(01:04:18):
its natural environment, if you know what I mean.
Speaker 4 (01:04:21):
That's a very good point.
Speaker 5 (01:04:23):
I think this prompt would provide them the example of
the context of the episode.
Speaker 4 (01:04:28):
But what I think could also be useful is.
Speaker 5 (01:04:30):
We say, make sure you include the word a phrase
in the context of the episode and give another example.
So maybe listeners have the the example of how it
was used by you or your guests and another example
of how it could be used. So you've got both.
Speaker 3 (01:04:47):
There more examples that the better. I think. When you
know learners see a phrase being used again and again
and again in different sentences in different contexts, it really
helps to kind of provide a meaningful sort of yeah context,
which really helps to emphasize the meaning of something.
Speaker 4 (01:05:09):
Okay, I will just read out one more time.
Speaker 5 (01:05:12):
I've what I'm asking chat chipet to do, which is
based on the transcript I pasted. Find ten natural sounding
expressions that native speakers use, but the B one level
learners like me might not. For each expression, ask me
what I think it means, then explain the meaning and
how it's typically used. Ask me to write a new
(01:05:32):
sentence using the expression, and I've written at the bottom.
Make sure that you include the word or phrase in
the context of the episode, and at least two other
examples in different contexts. Maybe we should have put that
after number two, but I think it should be. Okay,
and let's see what chatchipt does does there? Okay, So
(01:05:59):
there's ten ten words and phrases. Okay, Okay, it's given
them one by one. So the first one is fly
by the seat of your pants. It's given us the
context from the episode, and it's asked us to guess
what this expression means.
Speaker 3 (01:06:21):
Yeah, can I read that out? Yes?
Speaker 4 (01:06:24):
Yeah.
Speaker 3 (01:06:24):
The first one, it says expression one fly by the
seat of your pants, context from the episode. And this
is something that Ian said, my guest. He said, there
tends to be two schools of thinking in terms of
crime writers. There's one school, which are the plotters and
the other school, which are the pants as it's spelt
(01:06:45):
that wrong, because that's pants is actually a kind of
made up word relating to the word pants p a
n t e p a n t s the things
that you wear, but it's written them p a n
c e r s. Anyway, the plotters and the pants
as they're called the panzers spelt correctly now, which means basically,
(01:07:08):
you're just flying by the seat of your pants. The
question is what do you think fly by the seat
of your pants means? So, just to be clear, two
types of two ways to approach writing a crime story.
One is to plot everything and the one and the
other approach is to fly by the seat of your pants.
(01:07:28):
You know, it hasn't given Oh, it's given us an explanation.
Speaker 5 (01:07:32):
I think it's what it's done, which might be might
be quite clever, or perhaps it was just the way
that I scrolled. But it's given us the expression, and
then we can if we like, I can, I can
respond to the question before I scroll down and see
the explanation. Yeah, So if I wanted to really challenge myself,
(01:07:53):
I would read that first and I'd say, I'm going
to really think about this. I'm going to write my answer,
and then I'm going to go and read the explanation,
which is yeah, I mean something to This expression means
to do something without planning, using your instinct and improvisation
rather than careful preparation. It comes from aviation, where pilots
(01:08:15):
would rely on their instincts rather than instruments to fly
the plane. And it's given us two other examples which
both seem to be pretty okay to me. When the
surprise guests arrived, I just flew by the seat of
my pants and made dinner with whatever I had. It's
not something that I would say, but perhaps it's something
(01:08:36):
that you would.
Speaker 3 (01:08:36):
Say, Luke, Yeah, I would yeah, yeah, yeah. If you know,
just a situation where you haven't prepared and you need
to kind of make it up as you go along,
you need to improvise, you know, I can imagine. I
think I've explained this. I think I talked about this
actually during the episode at some point, maybe in the
explanation part at the end. I said something like, you know,
(01:08:58):
I may imagine going into a classroom without a lesson plan,
and one of your colleagues says, have you if you've
not got a lesson plan for that class, and I'd say, well,
there wasn't anything available and I don't have time. Look,
don't worry, I'll just I'm just going to have to
fly by the seat of my pants on this one.
I mean, it's quite it's quite a funny expression. It's
quite a cumbersome expression. I might say, don't worry, I'll
(01:09:21):
just improvise. It's probably what would be easier, but still
it's something that I would say that though as well.
It depends in a very stressed out moment where every
second counts as I'm heading into a classroom trying not
to be late, I might not say, oh, don't worry
I'll just fly by the seat of my pants on
(01:09:41):
this one. Too many syllables, it's like there's not enough time. Instead,
I might just say, don't worry, I'll improvise, you know.
But anyway, Yeah, it's certainly one that I think that
people shouldn't know, you know.
Speaker 5 (01:09:56):
I think the typo on pantss, I imagine that's come
from them that AI generated transcripts, rather than chat GPT
creating a writing it wrong.
Speaker 4 (01:10:07):
I think that's probably where it's come from.
Speaker 3 (01:10:08):
Right, that's YouTube's transcript you see if you've got the
transcript from my website, that there wouldn't have been there
that would mistaken. Yeah, yeah, made a rookie mistake there.
Speaker 5 (01:10:18):
Actually an interesting point to add is that what you
can do, as opposed to going to YouTube and copying
down the entire transcript, you can upload the pdf. I
think you provide PDFs Luke. Yeah, so you can just
upload from upload from computer and you can just say,
(01:10:39):
you know, this is the PDF of the episode transcript
of episode number nine and twenty one.
Speaker 3 (01:10:45):
Or you might need to just be a bit more
selective because often my PDFs don't just have a transcript,
they have a bunch of other stuff too, including loads
of like vocab lists and vocab quizzes and things like
that as well. So you might want to open the transcript,
select the open up the pdf, find the section which
is the transcript. Often that's that's normally the sort of
final part, and just copy paste all of that and
(01:11:08):
then paste it into chatch GBT. You know, you do
need to think for yourself to a certain extent.
Speaker 5 (01:11:13):
Listeners, yes, they certainly do, so should I do you
want to see what the next word it comes up with?
Speaker 3 (01:11:24):
Do one more and then we'll move on to something else.
Speaker 5 (01:11:27):
Should I try and write a sentence, and I'm going
to try, and I'm going to make some grammar mistakes,
and I'm going to use it in the wrong context.
Speaker 4 (01:11:33):
I for fly by the seat of my pants. So
I went on the aeroplane to France, and I decided
to flew by the seat of his pants.
Speaker 3 (01:11:50):
What normally, normally, what what learners do when they're using idioms,
is they don't use all of the different parts, so
they'd say I flew by my pants, or or well,
I decided to fly by the seat my pants, So
just forget the of.
Speaker 5 (01:12:09):
Okay, let's do this one. So I went on the
airplane to France, and I decided to fly by the
seat my pants.
Speaker 3 (01:12:18):
No one knows what that means. But that's all right.
Let's see what chat GPT comes up with.
Speaker 5 (01:12:23):
What happens here almost there, Your sense is grammatically correct.
But the expression fly by the seat of your pants
doesn't literally mean flying on airplane. I mean's doing something
without planning or preparation. Here's a more natural use. I
didn't study much for the exam, so I had to
fly by the seat of my pants during the tests,
try and rewriting a sentence to show a situation where
(01:12:45):
you're improvising or not well prepared. Shall I just see
what happens if I copy this, but I just still
make the same grammatical mistake.
Speaker 3 (01:12:54):
Yes, so you're just copying the example it gave and
removing the preposition of the you removed before. But it
didn't notice. And we're going to just see if it
is sort of observant enough and meticulous enough as a
teacher to meet your very particular high standards.
Speaker 5 (01:13:14):
And I have found sometimes that chat GBT will kind
of go through one mistake at a time, and it'll
say you made this one mistake, even if you made
two mistakes, and then you'll fix that mistake, but you'll
keep the other mistake, and it'll say this is the
other mistake, which is a little bit frustrating.
Speaker 4 (01:13:36):
Sometimes.
Speaker 5 (01:13:38):
Let's see how it does in this example. So I've
written I didn't study much for the exam, so I
had to fly by the seat my pants during the test.
And it has actually done what we wanted it to do.
It's a great job. You've got the meaning and context
right this time. Just one small correction. Your sentience should
be I didn't study much for the exams, so I
(01:14:00):
had to fly by the seat of my pants during
the test. And it's not very clear, but of is
bolded here, So it's drawing our attention to what we
did wrong and said, you just missed the word of
after seat. Other than that is perfect well done. Yeah,
so it's and then it's obviously remembered the context of
(01:14:21):
the wider exercise, which is the finding native level expressions.
And the next one is take its toll context from
the episode, it has really taken its toll physically and mentally, and.
Speaker 4 (01:14:36):
Question what do you think take its toll means?
Speaker 3 (01:14:39):
Now? Hopefully right, hopefully As we said before, this is
stuff that my listeners will be doing after having heard
the episode. Right, So this is after listening stuff because
you know, I said, I want it to give the
sentence context for the phrase. It has done that, but
it's not quite enough. So all you've got is it
(01:15:01):
had really taken its toll physically and mentally. We don't
know what it is, but if you've listened to the episode,
you probably would remember the general meaning of what he
was talking about. He was talking about why he I
think why he quit doing comedy or why he quit
touring so much, traveling around and doing I can't remember exactly,
(01:15:24):
but that's a nice expression. It's a very useful expression.
It's a very common expression. It's quite a specific expression,
and it's I think, exactly the sort of thing that
people should learn. I don't know if it's be if
exactly BE one level. Normally, at B one level, I
think learners would probably be focusing on slightly more high
(01:15:47):
frequency language and getting it correct before they move on
to this sort of idiomatic language. But nevertheless, it is
definitely the sort of thing that I would find myself
teaching in my class and in an intermediate class. So
to take its toll, it had really taken its toll
physically and mentally. What does take its toll mean?
Speaker 5 (01:16:07):
We could have perhaps changed that prompt to be even
more specific. So rather than saying that language that a
B one learner might not know, we say the kind
of language that BE one learner might not know and
should be focusing on for to achieve their B one level.
Speaker 3 (01:16:25):
That be too B two.
Speaker 5 (01:16:29):
So perhaps chatchpt is thinking, Okay, this is just something
a B one level learner won't know. Therefore, let's just
let's just give it to them rather than focusing on
more level appropriate language.
Speaker 3 (01:16:42):
To be honest, if one of my learners in a
B one class successfully used that phrase properly, you know,
for example, if they were saying, oh, I'm going to
give up my job? Oh really? Why? Well, it's just
to be honest, it's very I'm very tired. Why because
traveling so much has really taken its toll. I think, wow,
(01:17:04):
what's this person doing in an intermediate class? So to
be fair, you know, that is the sort of thing
that would get you out of an intermediate class and
into a higher level. So fair enough, I suppose.
Speaker 4 (01:17:16):
Yeah, So yeah.
Speaker 5 (01:17:18):
This is one example of how you can sort of
use chatchypt to help you focus on different words or phrases.
And of course, in many of your episodes, you focus
on these things yourself, so you're you're automatically drawing listeners
attention to them and talking about usage.
Speaker 3 (01:17:39):
And so on.
Speaker 5 (01:17:42):
But in episodes where you are not doing that as much,
or if they're more kind of conversational style episodes, this
can be a useful a useful way in which you
can think about how to use chatchy BET to kind
of get some of those get some of that value
from the from the audio that you might not necessarily
(01:18:05):
have got otherwise.
Speaker 3 (01:18:07):
Yep, absolutely yeah. Now I'm sure that you've got loads
of other prompts and other things that we could look at,
but we can't spend all that much time going through
them all, So I don't know if there are others.
So okay, So we looked at listening comprehension. We looked
at sort of like focusing on grammar for example. We've
looked at expanding vocab. Just you know, looked at a
(01:18:29):
few different things. Is there anything else before we sort
of maybe have a discussion about all of this and
then end the episode.
Speaker 5 (01:18:37):
So I think there's there's lots and lots of other
things that we could talk about, but I think it
will probably be a little bit boring for people just
to watch me sharing my screen and sort of us
battling with Chack GBT saying stuff's incorrect when it's actually correct,
and kind of going through exercises. What I think is
(01:18:58):
probably the the best thing we can do is we
can I can leave a load of those prompts the
in the PDF, and then people can go through them
and practice using them practice adapting them for themselves. Hopefully
what we've done now gives people enough enough of an
(01:19:20):
idea about how they can how they should be thinking
about adapting them for themselves, and some of the possibilities
of using chatchupt with podcast episodes, and of course how
to just set it up from a technical point of view.
Speaker 3 (01:19:36):
Yeah, really good, that's great. There were actually a couple
of other things though, that I wanted to ask you about,
just a couple of others before we have our sort
of discussion, and that would one of those is you
mentioned to me before before we started recording that you
can use AI to help you search my episode archive,
because I've got a very long episode archive. There's like
(01:19:58):
over nine hundred episodes, and I'm sure that sometimes my
listeners don't quite know where to start if they're looking
for episodes where I talk about certain things or whatever.
Maybe AI can help them to find the relevant episode
that they're looking for. Yeah.
Speaker 4 (01:20:16):
I just found it to be.
Speaker 5 (01:20:21):
Be in many ways of superior way to search the
search and kind of archive of anything than Google. Because
if you're saying, if you could go into Google and say,
you know, episodes of Luke's English podcast about travel or something.
Then it might find some things, but it's probably not
going to give you a proper explanation and sort of
(01:20:44):
why why you might want to listen to it. But
if you might, if you said, you know, I am,
I would like to listen to an episode of Lukes
English podcast where Luke talks about his life in France.
Speaker 4 (01:20:55):
Let's say, and it's under an hour. Lone, for example,
doesn't exist. You've broken chetchypt the results. But I think, like.
Speaker 5 (01:21:13):
Like searchingly for anything, really often the results you'll get
with AI just it takes into account better context than Google.
Speaker 4 (01:21:23):
Search currently does. So I think for.
Speaker 5 (01:21:28):
Listeners who haven't listened to all nine hundred and twenty
eight as of the time of recording.
Speaker 3 (01:21:35):
Oh oh, there's more more. By the time this goes up,
it'll be nine hundred and thirty watching record, I don't
know quite a.
Speaker 5 (01:21:40):
Lot if you haven't listened to all of them, or
even if you have and you want to go back
and then listen to something about whatever it might be
where Luke talks about cheese or.
Speaker 3 (01:21:53):
No, that's good, I did this whole episode called the
cheese episode.
Speaker 5 (01:21:57):
Well that's truely, that's probably an example one's easy to
find on Google. Don't search for that, but you know
ones where Luke talks about Luke uses particular types of
language that you like to you like to focus on.
Speaker 4 (01:22:12):
I think it's just a.
Speaker 5 (01:22:15):
Better a better way than using Google because it has
all the context straight.
Speaker 3 (01:22:22):
Yeah. It just searches more intelligently than what Google does,
which is often it prioritizes certain things, and you know,
there's all those Google just works in a different way,
but AI might have a more direct approach, might understand
exactly what you're looking for and be able to find
it better. So how do you actually do that though?
Do you just go into chat GPT and say, can
(01:22:43):
you recommend episodes of Luke's English podcast where he talks
about food exactly?
Speaker 4 (01:22:48):
That's all I do.
Speaker 3 (01:22:50):
So it can sorry, it can search the web these days,
because a couple of years ago, when it first came out,
it wasn't actually connected to the Internet. But these days
it does search the web, yes.
Speaker 5 (01:23:00):
Yeah, yeah, yeah, And I think almost every AI model
can do can do that as well. So whether you're
using Gemini or Grog or deep Seak or anything like that,
then they will also be able to do the same thing.
Speaker 4 (01:23:13):
And provide you with the links.
Speaker 5 (01:23:15):
They can probably also take it one step further and
give you some exercises and recommended reading from the BBC
or something like that that would compliment it too. So
I would really encourage encourage people to just to play
around and see what they can find.
Speaker 4 (01:23:34):
For themselves.
Speaker 5 (01:23:36):
So let's say I don't know, Let's say you are
going on a you go on holiday to London next month,
and you are going to be and you're an avid
listener of Luke's English podcast, and you'd like to listen
to some to Luke talking about life in London or
(01:23:59):
British customs, London tourist hotspots, whatever it might be. And
you'd also like to have some relevant articles about London
news or things like that, and provide a vocabulary list
of words that might be.
Speaker 4 (01:24:15):
Useful for your time in London. I'm not quite sure
exactly what you might like, but.
Speaker 5 (01:24:24):
Kind of creating a long prompt like that allows you
to search the search all of your archives so you
could find an interesting episode to listen to you can.
It'll also find you interesting articles do whatever you want.
Speaker 4 (01:24:37):
Really, so.
Speaker 5 (01:24:40):
Yeah, I think it's just something that people should switch
switch to when it comes to looking for particular episodes.
Speaker 3 (01:24:49):
I mean, Google is starting to incorporate AI into its
search results these days, so you know, it won't be
long before the old version of Google is replaced by
sort of AI powered version. I suppose, you know, that's
just going to become normal. It'll people will just forget
(01:25:11):
that Google used to be different. They will only eventually
only be familiar with the AI versions of Google. And
I think as we move, you know, forward, AI is
going to become more and more normal and ever present,
to the point where we will kind of even really
notice that it's it's kind of driving things and sort
(01:25:31):
of powering a lot of the things that we do.
Everything's going to become a much easier. Eventually, our smartphones
will have much more powerful AI built into the operating systems,
so that it will just sort of almost intuitively know
what we're trying to get it to do. You know
that maybe it'll be a thing of the past where
(01:25:52):
you know, when you sort of battle with your phone
trying to find certain settings, how do I how do
I reduce the volume of my alarm in this damn thing?
You know, and you're trying to get in there, and
you Eventually you'll just say Siri, reduce the alarm volume
and it'll do it. You know, you know, it'll be
so much it's just gonna reduce so much friction in
(01:26:14):
our lives that will forget what it used to be,
like I expect. Back to that point though, of the
episode archive and finding episodes, what you could do is
just use the u ur l teacher Luke dot co
dot uk slash episodes and give chat GPT that website
address and say search this page for blah blah blah,
(01:26:37):
And that's probably going to be a good way of
focusing it a bit more.
Speaker 5 (01:26:41):
I think I think chat gipt can already do that
with the web search. I think that I think that
is possible as of as of the time of recording.
Speaker 3 (01:26:50):
Yeah. And the other way, of course, is the old
fashioned approach, which is where you open up my episode
archive in your browser and then you do a command
F or control search for keywords, and each episode listed
on my website has a little summary of what's included,
and that can be a way too. For example, if
you're looking for an episode about cheese, where if you
(01:27:11):
just heard me mentioned the cheese episode, you could go
into teacher Lou dot co dot uk, slash episodes and
then control air for or command F and then search
cheese and the words cheese will be highlighted. Every time
that word has been written on that page, it'll be
highlighted and you can skip to it and eventually you'll
find the cheese episode. But I mean, I can just
(01:27:31):
you know, I could just tell you what the number
is if you want, what is cheese is mentioned once
in my episode, archived once in the cheese episode, and
that is the cheese episode, episode number seventy two.
Speaker 4 (01:27:46):
What is what is the cheese episode about? Other than
is it just you talking about cheese? Is there There's
a lot of things to say about cheese, especially lives in.
Speaker 3 (01:27:58):
France, but the hell of a lot to say about chees.
So if you open it up that page number seventy two,
the cheese episode, it says why is this called the
cheese episode? Listen to the whole thing to find out,
And it's actually a rambling episode. It was a rambling
episode with a challenge. The rules of the challenge were
I have to speak for thirty minutes. I'm not allowed
(01:28:18):
to stop speaking or stop the recording unless I get
a phone call. If I get a phone call, I
can pause the recording and continue later while speaking. I'm
not allowed to leave any long pauses. Blah blah blah.
I have to try to make sense and keep it interesting,
which is difficult. And I just talked and talked and talked,
and eventually I ended up talking about cheese. And that's
(01:28:38):
why I ended up calling it the cheese episode. So
it's not only about cheese. It's about a bunch of things,
a lot of waffling nonsense, and then some stuff about
cheese at the end. And then I was like, I know,
I'll call this the cheese episode, and everyone will think
why is this called the cheese episode? And then they'll
listen to it and then they'll discover why. That was
episode number seventy seventy two recorded that was published in
(01:29:03):
twenty eleven.
Speaker 4 (01:29:05):
Wow, so that was relatively early in your podcasting.
Speaker 5 (01:29:11):
Journey.
Speaker 3 (01:29:11):
Yes, it was, yeah, you know, a sort of formative
episode where I established the tradition of doing rambling on
the podcast.
Speaker 4 (01:29:22):
Have your opinions on cheese changed in those fourteen years?
Speaker 3 (01:29:27):
My opinions on cheese have matured. You could say I
have more of an appreciation of cheese. I think these
days even more than you know, obviously since moving to France.
Oh yes, I do like cheese. I mean who of course,
I can't imagine anyone not vegans. Of course, vegan cheese.
I've tried it. It wasn't as good as real cheese,
(01:29:49):
I'm afraid, although the principle I do agree with is
just the taste, not so much.
Speaker 4 (01:29:56):
That's a problem.
Speaker 3 (01:29:58):
Yeah, all right. And the other thing I wanted to
ask you about was you mentioned before that you could
make your own sort of you can make your own
GPT within chat GPT and ask you to do certain things.
And am I right in thinking that you've created one
that allows people to chat with a kind of AI
(01:30:18):
version of me? Is this right?
Speaker 5 (01:30:20):
Yes, that's correct. So there are these things called custom
GPTs I think you have. I think you have to
have one of the paid chat GIPT accounts to be
able to create your own, but you can. Anyone can
interact with them. And essentially they are they're kind of
like versions of chat GPT that have some pre existing
(01:30:45):
knowledge or instructions, and they can be about anything.
Speaker 4 (01:30:49):
So you can have a bird.
Speaker 5 (01:30:51):
Watching GPT or GPT about podcasting or whatever it might be.
Or you can have a GPT is the l E
P GPT, which which hopefully which I've tried to create
(01:31:11):
based on your episodes and to act as as you.
So maybe and I should also just put a little
disclaimer here for anyone thinking you know what's going on,
but this is this is meant to be more sort
of uh interesting, fun and fun rather than purely educational.
(01:31:34):
So I thought it'd be fun to make this and say,
I wonder what Luke thinks about interacting with an AI
version of himself and does he think that it's completely
ridiculous or has he got something slightly right? So maybe
if if you wanted to share your screen and try
and you human Luke interacts with AI lou LP GPT
(01:32:02):
and see what happens.
Speaker 3 (01:32:06):
Okay, So I should add that this is not like
a voice emulation of me, it's not actually my voice,
because that is a possible thing of course these days
that you can you know, you could emulate people's voices
and all sorts of other incredible things which are quite frightening,
(01:32:28):
like you can create sort of video versions of a person,
which is on the more terrifying end of the AI
spectrum where you know, you get deep fakes and all
that sort of thing. There's no actual voice emulation in
this GPT that you've created. It's more like just a
text chat that you would have. So I'm going to
(01:32:49):
open it up now.
Speaker 5 (01:32:51):
And it's also been you can given instructions not claim
that it is actually Luke. So there's no sort of
impersonation going on here, but hopefully a bit of fun
where it might tell some jugs or teach some English
or do certain things like that.
Speaker 3 (01:33:09):
Okay, so let me all right, So imagine I'm chatting
with myself. What would I ask myself? Goodness me? This
is a bit like when I'm lying in bed at
night and I can't sleep and I end up sort
of like my thoughts end up bouncing around inside my
own head. I'm just going to keep it simple and
(01:33:29):
say what, Hi Luke. I'm going to start with this
Hi Luke, all right, and it says hello, how lovely
to hear from you? How is it going? Got your
cuppa ready? Meaning have you got your cup of tea ready?
Or maybe you're out and about with your headphones on
doing that classic English podcast walk thing? What brings you here?
(01:33:50):
Today looking for a cheeky language tip, a podcast episode recommendation,
or maybe just a bit of good old English waffle. Okay,
so this is its version of me because I'm always
talking about cups of tea and and things like that. Yeah,
I should.
Speaker 5 (01:34:09):
I should also also add just it sounds like I'm
trying to give myself a whole lot of disclaimers here.
But how it was kind of trained was on your
podcast transcripts rather than me saying this is how Luke
should talk. So it's not me saying Luke would definitely
say got your cupa ready?
Speaker 3 (01:34:29):
What what's done?
Speaker 5 (01:34:31):
Has taken transcripts of the episode and thought, right, this
is the persona that I'm going to build, and then
you know, trying to talk like it thinks this person would,
So that's.
Speaker 3 (01:34:44):
Yes is going on? I got it? So, yeah, you
fed it lots of my transcripts, and this is its
version of me based on that. So, got your cupa ready?
Speaker 1 (01:34:54):
Is?
Speaker 3 (01:34:54):
I suppose because I'm British and of course we drink
tea all the time, and a kuppa is a slang
word for a cup of tea, so I can see
why it's doing that. What am I going to ask it,
so have you got have you got any interesting new
(01:35:15):
podcasts in the pipeline? Mm hm ah, Now that's a
good question. Always nice to know someone's curious about what's
cooking in the LP kitchen. Right, So we imagine me
rubbing my hands together like a mad professor. Let's see
(01:35:38):
what kind of podcast magic we can conjure up next.
Here are a few things in the pipeline. Okay, so
it's just making stuff up now because these are not
It doesn't have some genuine insight somehow by spying on
me what actual episodes I have in my actual pipeline.
But it's got a deep dive into British humor blah
(01:35:58):
blah blah accents of the UK. Part two Lepster Stories
The Lazy Grammar Guide. That sounds interesting. A new grammar
focused episode but delivered Luke style, not like a dusty textbook,
but more like, here's how native speakers actually use this
stuff without thinking about it too much. You'll come away
(01:36:20):
learning loads, but with your brain melting promise. Okay, that's
not actually, it's not a bad idea. The Lazy Grammar
Guide quite like the idea here first. Yeah, okay, well,
that's a bit of fun. I suppose you could. You
could have fun with that. The link will be in
the will be in the PDF for Alistair's lp GPT,
(01:36:40):
which yeah, as he said, it's just a bit of
fun really, and it's not entirely accurate. It's a kind
of a sort of a version of me, a loosely
based a version that's loosely based in a kind of
caricature version of me. Let's say, that's that's quite good fun.
Speaker 5 (01:37:00):
You can probably also this is something I just just
thought about now, but it's probably possible for listeners to
create their own GPT that is a kind of English
learning podcast companion. So kind of takes in lots of
the prompts that we talked about earlier about you know
(01:37:24):
how how they I should behave and so on, and
then you just paste in the transcript and it is
kind of ready to go.
Speaker 4 (01:37:31):
So that's something that people might want to consider doing.
Speaker 3 (01:37:35):
Mm hmm, okay, really good. I said before that we
would discuss some stuff before the end of the episode.
I feel like we could and should. I mentioned issues
earlier on like copyright issues. Have you have you seen
the recent trend for studio Gibli Imagery in chat GPT.
(01:37:57):
Are you aware of the studio Ghibli? Yeah, controversy, can
you tell us what it is?
Speaker 5 (01:38:04):
So it's this animation studio in Japan that has been
going for a lot. I'm not quite sure exactly how long,
but you know, many decades, and as a particular animation
style where I think every single every single kind of
(01:38:25):
image of the of the animation I think is hand drawn.
I believe I may be getting this slightly wrong, but
it's an incredibly labor intensive process. It's very distinctive. So
if you see a studio could be image you, you
kind of immediately recognize it. And last week, I believe
(01:38:49):
maybe two weeks ago, relatively recently, at the end of
March or start of April, chatchbt open Ai rather released
a new image generation model that was just way better
than all of the previous ones and allows you to
(01:39:09):
do things like upload an image and say, you know,
create this in studio globally's style, and suddenly everyone was,
you know, millions of people around the world were able
to create their own images in this very particular style
of this animation studio. And of course the animation studio
(01:39:32):
was in no way kind of compensated for it. And
this I can't remember the name of the man who
who's kind of the boss of the whole thing and
the chief animator, but he was very under i think,
in my opinion, understandably cross about it because he, you know,
(01:39:52):
spent his life working on creating this style, and you
know it's incredibly intensive and expensive to do. And then
suddenly anyone around the world, including me, i should say,
was able to generate images like that, and to me,
(01:40:12):
to my kind of untrained eyes, they looked very, very good,
almost unmistakable from from what it would look like if
if someone asks Judighibili did that. So yeah, all sorts
of copyright copyright issues. I also seen I'm not sure
if you've seen this, two league, but people taking some
(01:40:38):
of the transcripts of my episodes and uploading them onto
YouTube with different AI generated voices. So obviously I'm in
a completely different category to kind of the size of
Geo Gibili. But even making a very small podcast, finding
people just take the transcripts and use text generation and
(01:41:01):
then upload it to YouTube to try and pass it
off as their own.
Speaker 3 (01:41:05):
People have taken transcripts of your episodes and they've fed
them into some sort of AI which then goes from
text to speech. Yeah, exactly, And they formed audio episodes
where it's an AI voice reading out your transcript and
they're passing it off as their own content.
Speaker 4 (01:41:24):
Exactly.
Speaker 3 (01:41:24):
Yeah. Right, that's terrible. Yeah, and that just shows how
AI can really make it incredibly easy to sort of
violate people's intellectual property in some way, to breach people's copyright,
and also yeah, just to to create content that they
(01:41:46):
don't have the rights to in some way. It really
makes that easier, I think.
Speaker 5 (01:41:52):
In Yeah, it's with something like a podcast, it's in
a funny category, I guess, because you know, for someone
who is listening to your show, they you know that
they're here for you, right, and to a much much
lesser extent the guests who go on to your show,
but they're here to listen.
Speaker 4 (01:42:12):
To you and to learn from from you.
Speaker 5 (01:42:15):
And so just kind of slapping on instead of Luke,
it's Ted or whatever, and he's going to talk in
a in a funny accent and not have any of
any of your kind of natural charm and quirks and
stuff that is not going to be popular at all.
I would imagine, but something like an image where you know,
(01:42:39):
you can't see the process of it being generated, where
ultimately all you see is the final image. In the
case of the studio Gibi stuff, Yeah, it's kind of
scary because people can't see how that has been made
and the Yeah, the original career of the concept gets
(01:43:02):
no compensation or attribution for their decades of hard work.
Speaker 3 (01:43:07):
Yeah. Yeah, that's the crux of the controversy. Also issues
of slop. This is a kind of a slang word,
a fairly new slang word, which refers to the amount
of AI created content which is everywhere on the internet. Now,
(01:43:27):
you talked about someone using transcripts to make their own podcast,
using your own transcripts to generate their own podcasts with
text to speech software, But there are so many, especially
on YouTube, so many AI generated podcasts. Now, they're very generic.
(01:43:49):
You've noticed these two, right.
Speaker 4 (01:43:52):
Yeah, and so they're huge, some of them, I think
some of them are.
Speaker 3 (01:43:55):
I was surprised actually, because all the ones that I've
ever seen, and they pop up in my use to
recommended videos and stuff, they normally have minimal views, like
only a few views really, maybe a few hundred. But
you've shared a channel with me which has got over
a million and a half views on an episode that
(01:44:16):
was published something like six months ago. Yeah, that's really
a lot.
Speaker 5 (01:44:20):
In I think it's three months it was one and
a half million views, which which is massive, and it's
probably would make it one of the most watched like
English podcast episodes on YouTube. And it's all just just
two AI characters having a having what's clearly a fake
(01:44:45):
conversation with each other.
Speaker 3 (01:44:47):
Do you think it's clearly? You think it's that clear
because I looked through all the comments or as many
as I could under the video that you shared, and
it's like twenty six minute long video. It sounds exactly
like the AI voices that are generated by Google's Microsoft LM,
the Deep Dive function where you put in a text
(01:45:10):
and it generates a sort of roughly ten minute conversation
between two guests where they talk about the thing that
you've put up. It turns it into a podcast episode.
It sounds exactly like that, so it's probably generated with
the same software, maybe Gemini or something. I look through
the comments, I did not see one single comment saying
(01:45:31):
that this is fake, this is AI. Nothing. Everyone is
just saying I love the way you two talk. This
is really useful. People genuinely really loved it, and either
A couldn't tell that it was AI generated or B
just didn't care.
Speaker 5 (01:45:48):
Yeah, I think it's the it's the first one, unfortunately,
But it's that people sort of that it seemed so believable,
And I think if you are maybe if you are
at a lower level of English, you know, you're just
getting started with your English journey, you might not necessarily
(01:46:09):
recognize it. I think any native speaker would immediately recognize it.
And it's also slightly suspicious that you know, there's no
there's no people in the video.
Speaker 3 (01:46:22):
Yeah, there's It's always the same thing. It's always an
AI generated image, which is the video. It's not too
actual humans interacting. It's an AI generated image of two
really either two very attractive looking characters talking to each
other across a desk sort of alarmingly good looking, or
it's it's two sort of cartoon characters with microphones in
(01:46:46):
a nice, happy environment with maybe a sound wave which
is moving and a text transcript appearing on the screen.
But it's always the same kind of generic American accents,
a male and a female. Yeah, but I mean, do
you think that if all those people knew that it
(01:47:07):
was AI, would they care.
Speaker 4 (01:47:11):
I'm not sure.
Speaker 5 (01:47:12):
Maybe if if people are listening to this and they
have been shown I imagine people watching LP on YouTube
will presumably have also been shown you know at first,
but recommended these kinds of videos, and perhaps you've clicked
on them and thought, hmm, that's a I I'm not
(01:47:33):
going to watch that. So maybe yeah, add a comment
and let Luke know about whether you about whether you
have have tried these or not. It seems to me
because also the images they use are pretty clearly AI.
You don't have to be a you know, it doesn't
matter what language you speak for you to be able
to recognize that that is an AI image.
Speaker 4 (01:47:54):
I would think so.
Speaker 5 (01:47:57):
I don't know, but it does seemed to me a
shame because there's so many people, real humans creating human
native contents out there that is much more like by definition,
much more likely to be the kind of thing that
you will encounter when talking to a native speaker. Yet
(01:48:20):
people are still choosing to learn from AI generated text,
whether they know it or not.
Speaker 3 (01:48:28):
I think probably in the scheme of things, it's most
if we calculated all of the listens or views across
all learning English channels, and we're just talking about YouTube
across all the learning English channels on YouTube, and we
worked out the percentage of those views that were from
(01:48:51):
human generated content and the stuff from AI generated content.
At the moment, I think the human generated content is
way more. You know, a channel like English with Lucy
or don't know, one of the other more successful YouTube
channels for learning English. I think there are loads of
channels like that and loads of views that are on
(01:49:13):
those channels. It's rare for me to find an AI
channel that is similarly successful. This one that you showed
me is the first one I've ever seen, But I
wonder how I'll go in the future. And it's really
like the success of these channels really is just based
on I think two factors. One of them is, yeah,
(01:49:35):
like I said, whether people will actually be able to
identify if it's AI or not. And if for the
people listening, if as far as they're concerned, if it's
just English being spoken, then fine, So that's the one
will they identify And second factor, will they actually care
if it's AI or not? And whenever I talk about
(01:49:57):
this and this topic does come up quite a lot
on my pot cast, and I always say the same thing,
which is that I reckon Eventually people will not care,
and people always rush to my defense, which is lovely,
and they say, don't worry. Luke will always choose you
or human teachers over AI. You know, I will if
it's a choice, And I say that's I think to myself,
(01:50:19):
that's very nice. But I think that maybe one day
you'll change your mind, because one day something will happen,
some content will arrive that will make you go, wow,
that actually that's brilliant. I love that, and you might
choose to watch it or listen to it, and that
sense of there being a negative stigma, you know, attached
(01:50:41):
to it will be gone because everyone else is watching it.
Like this YouTube channel is absolutely massive, and when you
watch it you realize why because actually it's great. It's
got all It's just like as good as anything that
you know, Luke or Lucy or or Alistair would come
up with, but even better because of whatever the extra
(01:51:01):
factor is. Because it's sort of like somehow is able
to pinpoint exactly what a human wants in a way
that normal humans are not able to do. You know,
when AI is more intelligent than us, and that apparently
is only around the corner. It will be able to
(01:51:22):
manipulate us better than even the best marketing content creators
or whatever you want to call them, or able to
do you know what I mean. I don't know if
that's a paranoid, sort of slightly pessimistic vision of the future,
but you know, just from what I've read and heard
from the people who really know about this stuff, they
(01:51:43):
do say that eventually, when the AI is more intelligent
than us, that you know, a lot of us won't
stand a chance. You know, it'll be able to do
things that are brilliant, and we will gladly accept them,
you know.
Speaker 5 (01:52:01):
And I guess also just with the example of this,
like AI generated podcast, it clearly is tapping into something
that people like because the dialogue, you know, there's two
different voices, the dialogue is clearly at a level that
makes it relatively easy to understand. And because it's you know,
(01:52:21):
because it's all AI generated, they're they're not you know,
arms and rs and people accidentally going and changing what
they were going to say, so it's probably pretty easy.
It's like it's like reading a text but just listening
to it, which is very different to a normal conversation
where people kind of think and say, hang on, actually this,
(01:52:42):
or they go and they make a slight mistake like
like I'm doing now, so it's actually easier.
Speaker 3 (01:52:49):
So yeah, you say that, But I've listened to some
of these things, and they do incorporate those sorts of
natural features where someone appears to be making up their
sentence as they go along, like the rhythm of it.
Like in the previously AI generated audio, like let's say,
(01:53:12):
even text to speech or something, it did sound very
robotic and horrible, awful, unlistenable, but they've found ways to replicate.
They've really found a way to analyze natural human speech
and replicate it so that it sounds like someone is
thinking while they're talking, you know, the way that like
(01:53:33):
for I'll do give a demonstration of it now. So
when you're making a cup of tea, one of the
most important things you have to do is first of all,
consider whether you're going to make the tea in a
cup or whether you're going to make it in a teapot. Okay,
And that's the difficult choice that you have to make
when you're making a cup of tea and then the
other person goes, exactly, yeah, that's the thing. I'm always
(01:53:56):
having to deal with it in the cup or in
the teapot. It's one of those, isn't it. Yes, it is.
So let's have a deep dive into what making a
cup of tea is really like. From a British perspective,
it's always that kind of thing where they're sort of
laughing a little bit while they talk, they interrupt each
other slightly. They you know, it's incredible what they have
(01:54:17):
managed to develop. Hello everyone, I'm just interrupting here because
I'd like to demonstrate what the two of us are
just talking about by playing you a sample of an
AI generated audio podcast which I just created using Google's
(01:54:37):
Notebook LM software. So what I've done is I went
into Google's Notebook LM, which is free. Anyone can use
it as long as you have a Google account, and
I inputed this podcast episode that you're listening to, this one.
I took the transcript from the conversation that you're listening to,
(01:54:58):
and I put it into Notebook LM and generated what's
called the audio overview otherwise known as the podcast Deep Dive,
and it within a few minutes produced this eleven minute recording.
I'm just going to play you the first minute or
so just to give you an idea of what I'm
talking about. Listen to the way that the speech sounds natural,
(01:55:22):
and all those little features I was talking about before,
about how it sounds like they're thinking while they speak,
the way their voices slow down sometimes, the way they
react to each other, the way they almost interrupt each other,
and certain things are always repeated, like the word exactly.
There's always someone saying exactly always in these sorts of recordings,
(01:55:46):
and other similar things like that. Anyway, let's play that
sample so you can hear what it sounds like.
Speaker 8 (01:55:53):
Welcome to the deep Dive. Today, we're looking at a
well pretty clever way you can get even more out
of listening to podcasts for learning English. We're talking about
using AI tools, you know, the chat, GPT, that kind
of thing to really boost your progress exactly.
Speaker 9 (01:56:11):
Lots of you are already listening to podcasts, which is well,
it's great for immersion, yeah, definitely, but what if that
listening could be more active, more targeted, and actually feed
directly into your speaking and writing.
Speaker 8 (01:56:25):
Okay, I see where you're going AI is a sort
of helper.
Speaker 9 (01:56:28):
Sort of like a study buddy. Yeah, one that's listened
to the episode right there with you.
Speaker 8 (01:56:33):
And the cool thing is this doesn't have to be
some huge time commitment, right, not at all.
Speaker 9 (01:56:36):
Even just you know, a few minutes with an AI
tool after you listen, it can make a real difference.
Speaker 3 (01:56:42):
Okay, So that was a sample of Notebook LM's audio
overview automatically created podcast Deep Dive, And that's exactly the
sort of thing that you find now all over the place,
especially on YouTube. All of these AI generated podcasts, they
all sound like that, and no doubt they're just going
(01:57:04):
to get more and more realistic as we go forwards.
What did you think of that?
Speaker 10 (01:57:09):
Were you?
Speaker 3 (01:57:10):
Was that obviously fake? Did you think that was obviously fake?
Or did it sound pretty realistic? Did that sound appealing
to you? Is that the sort of thing you would
want to listen to? Now maybe you are thinking, well,
it sounded very American. They just sound really really generically American.
And you know, Luke, I listened to Luke's English podcast
(01:57:31):
because I want British English at the moment. Notebook LM
doesn't do different accents. It also doesn't do different languages.
I should say, it's just in English, but it's always
that generic standard American accent, So they don't do a
British English version, but they will eventually they. I mean,
you know, there's probably other ai that will do that.
(01:57:52):
But anyway, what did How did that sound to you?
Maybe it was a bit too American or something if
you're looking for British English. But regardless of that, is
that the sort of thing you could find yourself listening to.
Did it really strike you as being artificially generated or
did it sound like a conversation that you could have
continued to listen to? Let us know in the comments. Anyway,
(01:58:16):
back to the conversation.
Speaker 4 (01:58:19):
And clearly it is. It is better than not listening
to anything at all.
Speaker 5 (01:58:24):
So this is not to say to people who are
listening to that stop doing it, but it's Yeah, it's
fascinating to see that this is a choice that people
are making, either consciously or subconsciously.
Speaker 3 (01:58:39):
Yeah, indeed it is, and we've got to shape up. Yeah, yeah,
we're going to have to compete. I suppose.
Speaker 5 (01:58:47):
No.
Speaker 3 (01:58:47):
I was thinking about that this morning as I was
walking down the road. I was thinking, well, what do
what do we have to do. We have to sort
of really maybe lean into the human side of things
and always keep things sort of really natural and make
(01:59:08):
sure that it's really distinctive, and make sure that we're
including humor and feelings and things like that, which I
think that AI struggles to do at the moment. It's
not I've never really been moved by anything that AI
has come up with. I've been disturbed by it, but
(01:59:30):
I've never really felt a genuine sort of emotional reaction
to it.
Speaker 5 (01:59:38):
Yeah, I would agree. I'd also say that you know,
your lived experience and what you've shared is about your
life as a human being. It's something that AI cannot replicate.
So if someone's listening to an AI generated podcast and
they're saying, oh, yeah, this one morning when I was
(02:00:00):
walking back from taking my children to nursery in Paris,
So if an AI podcast is saying that it is
clearly just invented. Whereas there are probably people who've been
listening to you since episode number seventy two, what is
the cheese or sorry geez, who have been following you
(02:00:21):
and your life and its developments since then, and will
that stuff that cannot be replicated by anyone else, and
certainly not by AI.
Speaker 3 (02:00:35):
We will see. We'll just have to see, but it's
certainly a sort of very interesting time. We could talk
about other stuff, you know, like the things I mentioned
before about academic the effect on the academic world and
those things, but I think that's probably enough for now. Alistair,
thank you very much for sharing your specific tips and
(02:00:57):
advice on how people can really use AI as a
learning tool with a podcast, for example. I think that's
been very interesting. I'll remind everyone to get the PDF.
The link will be in the description to get it.
There'll be a lot of the prompts that Alistair has
mentioned included there, so you can copy paste them, adapt
(02:01:18):
them as you see fit. A link to the l
EPGPT which is a bit of fun, and some other
bits and pieces that you'll find on there. Great Alistair,
just remind us how we can sort of follow your
work and get more from you online, your podcast, your
website and stuff like that.
Speaker 5 (02:01:40):
Sure as I have a website which is Leonardo English
dot com Leonardo like Leonardo da Vinci, and I also
have a podcast which is English Learning for Curious Minds,
available on all respectable podcast apps and also now available
on YouTube.
Speaker 3 (02:01:58):
Okay, great.
Speaker 5 (02:02:00):
I want to also say end with one more thing,
which is why people shouldn't blindly trust anything that AI
tells them to do. As part of thinking about what
we should talk about in this episode, Luke, I had
some fun talking to chatchupt about our conversation. And I'll
(02:02:20):
tell you what Chatchipt said I should that I should
end our conversation with. And I'm going to read this
out word for word, which is probably gonna explain why
you shouldn't do it. So I should end with try
these tricks with l ep, folks, drop your chatchypt results
in the comments let's turbo charge your English together?
Speaker 3 (02:02:43):
It just sounds so generic, right, it does sound quite generic, cheesy.
It's not that it's not the way you would say.
You wouldn't say, let's turbo charge our English. Don't forget
to smash the like button, folks. It's not the way
that you would have put out.
Speaker 4 (02:02:56):
It's not it's not really, it's not me. I think
I just say thank you very much for having me, Luke,
and goodbye.
Speaker 3 (02:03:02):
All right, brilliant, thank you, thank you, Alistair, have a
lovely day, eat it bye. So that is basically the
end of the episode, thank you again to Alistair for
taking the time to prepare so much for this episode
and to share it with us on the podcast today. Yeah,
(02:03:23):
just like Alistair said, you can get the pdf. You'll
find a link in the description in the show notes,
you'll find a PDF. And on that pdf is all
the stuff you need, really, all of those prompts that
Alistair talked about, all the ones that he showed us,
plus loads of other ones that we didn't have time
to demonstrate. But you'll find loads of prompt on it,
prompts on his pdf, lots of stuff that you can use.
(02:03:46):
And yes, you can copy those things and paste them
into chat GPT and use them. You can adapt them
as you see fit, use them as a starting point.
And yeah, try it all out, Okay, try to see
if those prompts will help you. Try using chat GPT
or indeed other forms of AI, other forms of generative AI,
(02:04:08):
to help work on your English and push it, push
it further, you know, try to push it further and
try new ways, more sophisticated ways of using it as
a sort of learning assistant or learning tool. Don't just
use chat GPT to do all the work.
Speaker 4 (02:04:27):
For you.
Speaker 3 (02:04:27):
I mean, don't just you know, say I want to
write a comment in reply to this message, can you
do it please? And then you just copy paste what
it does? You know, don't use it as a replacement
for your own ability to use English. Don't use it
as a as a crutch, because then your your English
will not actually develop properly, you won't develop those true
(02:04:49):
genuine English abilities English skills. But you can use chat
GPT to really help you build your skills, as has
been demonstrated in the episode. So try stuff out, Okay,
try different things and let us know. Okay, I'd love
to know what the results were like for you, what
you tried to do, what happened, what the results were.
(02:05:12):
And you know, if you've got any specific sort of
study techniques or study routines or anything that you have
found is useful for you, you know, share those things
with us. Let us know how it goes for you.
All Right, that's the end of the episode. I don't
want to talk longer further here. There's a million things
to say as usual, but I think that's enough for now.
(02:05:35):
I look forward to reading your responses and comments. And yes,
link in the description for the pdf thanks again to Alistair.
Have a lovely afternoon, evening, morning, night, whatever time it
is wherever you are in the world, and I will
speak to you again on the podcast soon, but for
now it's just time to say goodbye, bye bye boo.
(02:06:01):
Thanks for listening to Luke's English podcast. For more information,
visit Teacher Luke dot co dot uk.
Speaker 11 (02:06:22):
We're serving up summer at Dune Stores and it's barbecue season.
Mix and match three for ten euro across a fantastic
range of meat and fish. Try something vibrant like succulent
King Bronze or enjoy boardbe quality assured meats like peery
peery chicken rings or new hot honey pork belly slices
and many more. Any three for just ten euro plus
(02:06:43):
save with the till with a ten or fifty grocery voucher.
Dune Stores always better value. Terms of conditions apply actually
be used on ex. Grocery shop of fifty or more.
Speaker 1 (02:06:53):
Before you listen to Oliver, Oliver listens to you dot E.
Speaker 10 (02:06:57):
Now it might have been my name of the show,
but it really is all about you. We want to
chat about what you want, whether it's a series or
just a bit of crack. We go from deep conversations
to laughing so hard you spill out your tea. That's
exactly what we're after, because everything in life is welcome
here kat.
Speaker 2 (02:07:11):
Oliver Callen every weekday at nine am on Ortie Radio
one and the Orte Radio app.
Speaker 3 (02:07:21):
If you enjoyed this episode of Luke's English Podcast, consider
signing up for Luke's English Podcast Premium. You'll get regular
Premium episodes with stories, vocabulary, grammar and pronunciation teaching from me,
and the usual moments of humor and fun. Plus with
your subscription, you will be directly supporting my work and
(02:07:41):
making this whole podcast project possible. For more information about
Luke's English Podcast Premium, go to teacher Luke dot co
dot uk slash Premium info