Using Practical Tech | AI - The Good, The Bad, and The Ugly - Today In Space

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Folks, welcome back to anotherepisode of Today In Space. We've

(00:04):
been very busy behind thescenes, doing a lot of 3d
printing and building of thepart detective for this next
phase. So we're doing a lot ofcatch up after a little bit of a
break, because it was a verybusy April for space. Thank you
for joining us and for yourpatience. I'm your space science
podcast host from the EastCoast, Alex G or fanos, in this
quick episode, we're diving intosome practical tech advice.

(00:27):
Since I am, after all, anengineer by training, so I
really like efficiency, and Ilike tech that helps me do more
than I can do as just one humanthis week, let's talk about AI.
I've become a pretty consistentuser of it, and I do talk a lot
to people about it. There's alot we could talk about with AI,
and I do have a lot of deeperthoughts on it, so I can

(00:49):
definitely do that in anotherepisode if you're interested.
But for this week, let's focuson practical Tech with AI, The
Good, the Bad and the Ugly, withmore and more people using AI
and others choosing to opt outentirely whatever your side is
on this, this episode is foryou. So let's dive in.

(01:15):
The good make AI your partner inyour productivity. To me, the
best scenario that humanity andAI are gonna have together is as
partners. For me, the bestexample of that is a game that I
played as a young teenager onGame Boy advanced called Mega
Man Battle Network still livesrent free in my head, and it's

(01:38):
totally informed how I thinkabout AI. And so in the game, if
you haven't played it, humanslive in a 21st century world
where everything is connected tothe internet. And this was in
2001 to 2004 so at that time, itwasn't obvious that that was
going to happen. So it's kind offunny thinking of of it. Back
then, humans would use these netnavies with their devices, and

(01:59):
they would connect to everythingbecause everything's connected
online, you could plug in anduse anything. And these net
navies were basically youravatar to make things happen for
you. So, great way to personifythe coding, basically, you know,
and in the game, you play as amain character called LAN
hilarious. And your net NaVi iscalled Mega Man dot exe, and

(02:23):
then you'd basically the teamwould go together, and you'd
play around, plugging in andinvestigating crimes done by the
group called www. And aside forbeing a great real time tactical
RPG, I really see the net naviesas the AI tools of tomorrow, not
today, because the AI tools arenot there yet, but they could be

(02:43):
like net navies, helping usaccess everything online and
making use of it in unique waysthat we as just one human
couldn't and so that's how Ithink about AI tools. It's my
partner in doing things andgetting things done, not
necessarily replacing me andjust taking the work off my
plate so I can retire and donothing on an island, I would

(03:05):
get bored very easily, so that'swhy I choose the partner aspect
of it. But the key there, aswell as the do everything for me
option, is to not lose myhumanity in the process. Now,
the last episode I did on AItools. I talked about something
very specific, the time savingsand cost savings that it was

(03:25):
doing for my creative processfor the podcast, and a lot of
that's still important today,but nowadays, here's what I use
AI routinely for number onewriting podcast templates,
right? So I'll give grog or chatGPT. I kind of bounce between
both. A bunch of context on theepisode that I want to make,
kind of like my notes on hey, Iwant to do this podcast. Here's

(03:47):
my point. Help me write atemplate for this, but then I
can jump in and start writing.Number two, searching for
articles on space stuff andanything we talk on the podcast.
When I'm researching for anepisode of the podcast, I could
spend a lot of time looking atresources, but these days, I've
started to use grok and chatgptto search for articles, right?

(04:10):
So that way I'm not doing thatwork. I could spend time doing
something else, more beneficialto the creative process. And the
idea there is that I just spendless time and not get distracted
as I'm searching for things. Youknow what? I still have a note
that I will if I'm scrollingonline, especially with all the
space content that I'mfollowing, I will take an
article and put it into thatnote, and then it just kind of

(04:33):
adds in. But this is the speedthat it can do. It is just
amazing. So I love doing it forthat. It helps me focus more on
the big picture of what theepisode is and what story and
messages I'm trying to tell.Number three, I'm transcribing
the episodes time stamps andbasically converting my audio
from complete audio to completewritten in a few minutes. And

(04:54):
that's what otter.ai whichtranscribes the episode and
identifies speakers, if we haveguests and. Notices their
voices. It gives a summary, anoutline, and of course, I've got
a checklist of things that itpicks up on during the meeting,
and I can look back at that. Ithink even if it's only 95%
accurate of transcribing theentire audio of the episode,

(05:16):
that's still 95% of the workdone of transcribing, which I
tried to do back in the day, andalmost took on some jobs doing
transcription work. Man, thatit's just such a better use of
time. So I love that tool.Number four, generating clips
for social I'm sure if you guysfollow the podcast and follow us
online, you can see how manyclips we're putting out because

(05:39):
of that, it's really nice. I'vetalked about it in the other
episode, removing eight to 12hours a week of what I was doing
per episode to find the clipsmyself after editing the full
episode was a lot of work. Eightto 12 hours, I have been saved a
week because of that, and then Ibasically get eight to 30 clips

(05:59):
every time I generate it. And Ican, I can do better, but
there's captions, right? I canget more specific out if I want
to, which is a cool upgrade.But, you know, adding the
captions was a huge plusoriginally, and now I'm watching
it get better, like every month,which is really, really cool.
But really, for me, the bonusthere is that I'm able to spend

(06:21):
more time on the messaging andhow like put together and
engaging the clips are, becauseit is science, and it can easily
get sucked into a bucket of thisis boring, so anything I can do
to liven it up from the originalthing in those clips is great,
and I just didn't have The timefor that before. And so as far

(06:42):
as the good for AI, you know, Ichoose AI tools that help
enhance a skill that I alreadyhave, or it fills a gap in a
skill set of something I am notgood at, and that's what we're
going to talk about. And so forme, the skill gap that AI is
helping me with is thatstructured writing and I are not

(07:04):
friends, and that's been thecase for a very long time. It's
always been like that, and Ijust in growing up in school, I
hated English class until I hada great teacher about my junior
year of high school that helpedme find my own creative writing,
and basically it's writing inthe same tone as I speak, right?

(07:25):
So in a lot of ways, that's whyI was not good at English,
because I was writing as Ispoke, which I guess is not a
thing, but I guess that's whypodcast and funny enough, one of
the reasons that had me go toschool for aerospace engineering
was that I wasn't gonna have todo all the writing that you'd
have to do with other majors.And boy, was I wrong. There's

(07:47):
still a ton of writing andengineering, but it's at least a
specific set of writing thateventually I got good at, or
good enough, you know. And forthe podcast, I spent years, a
decade, really using the skillthat I do have the gift of gab
to create and put out over 380episodes of a Space Science
Podcast. And for many, manyyears, I would podcast off the

(08:11):
top of my brain, like literallyoff the dome. And it wasn't
because I thought it was cool.It was because I was so bad at
writing that it was actuallyeasier and faster to just do my
research, write a few notes tomyself, basically like a napkin
of notes, put like maybe a wordmap of the big topics that I was
doing, and then I would wing itright. It's not a very efficient

(08:33):
way of doing it, but it worked,and I did learn and gain a ton
of skill in just speaking andcommunicating especially complex
ideas. When I had to, off thetop of my brain, explain
something very complicated inhopefully an easier to
understand simpler way. I had todo that a lot, and so that was a

(08:54):
huge skill to learn. But again,that same skill, the double
edged sword of that was thatsome episodes would just become
a galactic pin the ass to editand produce, because I went off
on some tangent. That was good,but now I've tripled my editing
time for all the stuff I had toadd in. And that was just
breaking me after a while, to becompletely honest. And so when

(09:16):
chat GPT came along, and Irealized I could just ask for a
template, like for an emerging,you know, I would say, you know,
generate me a template for anengaging YouTube video about
five to 15 minutes long, to cutdown on my editing time and just
get the message across fasterand more efficiently. There was

(09:36):
a big eye opening moment for me.That's when I realized how bad
my skill of structured writingwas like if a skill set score
could be out of 10. I have noshame in telling you that on the
best day, my structured writingskill is a three. But you know,
asking for the structure so thatthen I can just share my

(09:58):
thoughts and. It's then edited,and it helps me get the job done
faster. That was huge, and I wasthe weakest link in that part of
the skill set. And chat, GPT andgrok are the bridge that gets me
over the ravine and back intocreative podcast waters way
faster. So the good of AI isthat it can be your partner, and

(10:21):
together, you can accomplishmore than either one of you
could do by yourself. And aspart of how I think about AI and
using it, it's about the humandirecting the progress. And even
if it doesn't generate what youwant, you can immediately know,
like, that's not what I wanted,but if I was looking for

(10:41):
anything and everything, now Ican say, now that's not how I
want to do it, and that actuallyis very helpful, and can just
help you move forward and getthings done instead of
ruminating over every possibleway that you could do something.
So that's, to me, the great goodof AI today,

(11:02):
and now it's time for the bad Ihave two things, the potential
of AI and the reality of AItools are two very different
things. As a regular user, Ilove finding new ways to use it
and for it to help me out. Butsome projects or tasks are real
black holes of effort, time andenergy, and you can easily waste

(11:22):
a few hours just prompting theright way. And perfectionist
beware, AI is not there yet forevery task, and the more you
learn how to use it, the deeperyou may think about using it.
And that's where it leads usdown the path where many, many
hours can be spent testing. Andmy advice for anyone that is in

(11:42):
that mode of, how do I how can Iuse this? I'm excited about it.
Give yourself a time limit whenyou're testing something new
with AI tools. I think 30minutes is plenty. And you know,
it can be exciting to see whatit can do. But if your goal is
to, like, replace yourselfentirely out of the process, or
have it program something foryou. Be really careful. At that

(12:05):
time, you can spend 40 hoursinto an AI project only to
realize you're only 85% of theway there that you want, and you
still don't have something totake away from your time. So be
careful of that. And number two,not everything that AI generates
is actually useful. This is notreally spoken about by the AI

(12:27):
purists, but take it fromsomeone doing it every day.
There's still a lot for thesellms, large language models and
Agent workflows to figure out.You know, let's take clip
generation for the podcastexample, right? Every time I run
an episode, there's at least 25%of the clips that are just not
good. It's too short. The cutwas weird. The crop is off,
like, I'm not even in the video.It's like the background of the

(12:49):
of the studio here. So that'snot good. And then for image
generation, for the podcastcontent, for B rolls, or even,
like, lead generation for EEG 3dand the part detective, there's
just a lot of junk and thingsthat it just decides to do that
I didn't ask for. So, like, forimage generation, there could be
one thing wrong, like, it'sspelled a word wrong, and you

(13:10):
ask it all right. You know, Ilove this, but you said this
image, this text, I want it toread this. It'll regenerate it,
and sometimes it'll just make acompletely new image and do
something else you didn't like.So then you're just going back
and forth trying to get it towork, where, if you had drawn
just drawn it in the firstplace. It's whatever direction

(13:31):
you want to go in, or maybe youthen take that picture and edit
it right like there are fasterways to do it, instead of just
having the AI do everything.That's where people get caught
and for the part detective andgenerating leads. Like, it
actually can find some reallyinteresting leads that I
wouldn't have thought of, butit'll also generate a list and

(13:51):
just put information in there.And it's like, not right, right.
Like, it's like a differentcompany name and a different
website, or, you know, it justwill put information in there
because it was asked to giveinformation in like a human way.
It's kind of like lying andimplying just to to get the task
done, but to not personify it.It's doing what it was asked to

(14:12):
do, but it doesn't necessarilyknow that it's doing something
wrong, and that's the bad. Nowfor the ugly, when AI gives you
blind spots that you didn't havebefore. So when you first start
using AI tools, it feels a lotlike the singularity, right,
like a massive gravitationaldisturbance of your interest,
because now you realize thepotential, right? And for the

(14:34):
thoughtful of us, been carefulto not use it for things that
are at like an indispensableskill, right? Like for me,
critical thinking, reasoning,and then, like the creative
process, right? The creativeprocess is why I podcast. I
don't want to get rid of that,but even the most disciplined of
us, AI tool users, can getcaught. And I did recently. I
do. I am really careful of howdeep I go into it, like I could

(14:56):
have it create a voice for thepodcast and create the tone.
One, and I can tell it what todo, and then it could just start
doing it, and I could build avoice avatar. It could find the
trending topics, it could makethe episodes. But I wouldn't do
that. That's not that's I don'twant to get rid of myself from
the podcast. I do the podcastbecause I love it, and so that's

(15:16):
a blind spot that I'm aware of,but the one that I wasn't was in
this latest experience, right?So at the start of this month, I
was putting in work on anepisode about the next NASA
Administrator, right? Jaredisaacman. I had done some
research. I put the episodetogether in record time. I was
super busy with 80 stuff, so Ididn't have time to get to it,

(15:37):
and I put the episode out recordtime. I felt really good about
it, which sometimes is hard todo, especially when it's just
like a creative process, and Iam the way that I am, but I
really was happy with it, andwhen I put it out, luckily, I
have a friend that reached outand said, Hey, this hasn't
happened yet. I had jumped thegun and put the episode out

(15:58):
before the actual Senate votehad happened, the confirmation
vote had happened, but becauseI'm not like a US political
expert, I missed it. And maybe Iwould have missed it without AI,
but I would have had more timeto maybe pick up on. Oh, hey,
the whole vote hasn't happened.Oh, the votes that are counted
there are not the total amountof votes. So maybe I would have

(16:20):
picked it up, but maybe Iwouldn't have, but that was not
a blind spot that I thought wasgoing to be there. And so for
me, that that was the ugly thatit it gave me this error that I
didn't want to do, and I'm goingto adapt and change the process
later. I'm going to add in sometime after I get the final

(16:40):
script of the episode done. Oflike, what I want to do, I need
some time to just review it andjust check for any stuff that
I'm missing. And it makes methink about, you know, in this
low risk scenario of releasing apodcast, right, which is not
that important in the scheme ofthings, but what about things
that are way more important.Right? Where these these blind

(17:03):
spots could be happening thatnobody are looking for? What if
an AI tool in the medicalindustry is diagnosing and it
misses something that a doctoror a nurse could have picked up
in two seconds? What if AI isbeing used for weather changes
right as as they're flying fornavigation, and it doesn't
account for something that asimple pilot could have done in
a second. That's where we humansneed to be careful in the gray

(17:27):
zone of AI use cases, and it'svery powerful, but we need to be
really mindful of how we'regrowing together in the
partnership. Right? If AI isgrowing to a point where humans
don't grow anymore. That's bad,and the same is opposite, right
where it's like if these AItools are not developed

(17:47):
properly, maybe with somesafeguards for logic and
reasoning, or going against justthe greed of a company wanting
to have the best AI tool outthere. These are things that we
need to be aware of and think ofso I wanted to share my ugly
case of it, because people arenot talking about those. I
wanted to add that to theconversation, and if it can help

(18:09):
you avoid a situation like that,then I think we're, we're going
to be in a better place for allof us. So that's it. That's the
episode. Thank you for joiningus. This was a practical tech
edition of Today In Space.Please let us know what you
think and what you're doing ifyou're introducing AI into your
world, are you completelyagainst it? Are you curious but

(18:30):
apprehensive? Are you just allin like we want to hear from
you. Email us at Today In Spacepodcast@gmail.com you can also
comment below or DM us at TodayIn Space pod on Instagram, and X
Today In Space on tick tock andat El Greco on x. That's my
personal account. That's E, L,G, R, the number three Co. And

(18:50):
hopefully soon, we should beseeing another starship orbital
test flight, which will be veryexciting. Wishing the team over
their luck as they get ready forit. Starship is an absolute
beast of a machine, and if theycan figure out how to tame it,
then we're going to have areally fantastic feature ahead
of us, specifically for humanspace flight. Make sure to check
out AG, D printing.com to seewhat we're doing over there at

(19:13):
AG, 3d in the printing lab. Lotsof projects for the part
detective are going out verysoon, and I'm going to be
sharing them, and I'm pumped toshow you what we've been making.
If you're looking for some coolspace gifts you want to support
the podcast, check out the linksin the episode description, our
ly printed James Webb SpaceTelescope model and the starship
and new Glenn rocket pens areout now. They make great gifts

(19:34):
for the space nerd in your lifeor yourself, and you help
support the podcast andeverything we're doing so we can
grow and as always, the easiestthing you can do to support the
podcast is to subscribe to thepodcast, leave a five star
review and like and share withothers. It's free. It helps a
ton. We really appreciate you.Thank you for joining us for
another episode of Today InSpace. Spread love, spread

(19:56):
science. Stay curious. We'll seeyou in the next episode. You.
See ya

All Episodes

Using Practical Tech | AI - The Good, The Bad, and The Ugly

Episode Transcript

Popular Podcasts

NFL Daily with Gregg Rosenthal

On Purpose with Jay Shetty

Dateline NBC

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}Using Practical Tech | AI - The Good, The Bad, and The Ugly