January 30, 2025 • 8 mins
There's been plenty of volatility in the tech market this week off the back of the recent DeepSeek fiasco.
Despite things picking up after this week's crash, experts are still wondering what's set to happen next.
Sam Dickie from Fisher Funds explains further.
LISTEN ABOVE
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
Bryan Bridge Tech Companies while they've been up and down
like a Yo yo this week since the Deep Seek
thing happened. Interesting timing because Microsoft and Meta have reported
their annual results today. Sam Dicky from Fisher Funds is
with us. Sam. Gooday, good evening, right, how are you
going good? Thank you? Quick throwback? What was the straw
that broke the camel's back here?
Speaker 2 (00:22):
Yeah, that's an interesting question because it wasn't all new news,
so that deep Seak has been openly using these innovative
techniques to drive down the cost to train these AI
models over the last year or so. But it was
kind a threefold. It was the release of its most
recent model R one on twenty fifth January, which showed
the significantly improved results, which you guys have spoken about
(00:42):
this week, plus the talking heads picked it up at
Davos later that week. Plus critically, the deep Sea Gap
went from number thirty one on the Apple lapp store
to number one over the weekend, so that what really
was the straw that broke the camel's back? The consumer
voted with their feet.
Speaker 1 (00:59):
So the significant breakthrough, obviously is that they managed to
do this some well, a lot of questions been asked
about how much US tech was used to make this,
how palatable the platform will be for American users or
America you know, users around the rest of the world
outside of China.
Speaker 2 (01:18):
That's right. I mean one of the breakthrough is it
is important to talk about one of the breakthroughs. There
were many, but the simplest one to understand, the one
that stood out was pure reinforcement learning. So in playing English,
pure reinforcement learning is quite simple. So don't teach one
of these AI models to solve every puzzle on Earth
using every bit of data on earth, like how most
of the US large language models have been trained. Just
(01:42):
reward it if it does a good job with a
digital pad on the back, so a plus one for
the right answer and a minus one for the wrong answer.
And this sort of trial and error way of learning
has proved to be far more efficient and cost effective
as the model, just like we do, loves to reward
and kind of sort them out. In terms of the
controversy and people saying they don't believe the six million
dollar cost to train the model, and the controversy around
(02:05):
open AI saying that deep Seat piggybacked off, it's existing
chat GPT models. We don't know. We just know that
the arms race and the chipboard that you and I
spoke about late last years is really hotting up.
Speaker 1 (02:20):
So basically what you're talking about, they're the pure reinforcement
is like positive reinforcement. How did the big, biggest, you know,
AI companies in the world missed this, Well, that's right,
they were certainly aware of it.
Speaker 2 (02:33):
So Google, for example, used pure reinforcement learning. So the
word pure, by the way, is a lot of the
companies use human feedback reinforcement learning learning, but this is
sort of digital reinforcement learning with no human feedback or oversight.
Google used that technique with one of its much older
models called Alpha Go in twenty seventeen. But this pure
(02:55):
reinforcement learning technique historically had shortcomings and western which we
can go into it in a minute. But wesn AI
companies either weren't quick enough to solve those shortcomings or
probably weren't incentivized to solve them. Given it was an
AI boom, capital was freely available, and they had access
to cutting edge AI chips, which the Chinese did not.
(03:16):
And it seems like perversely the US restricting China from
accessing these cutting edge computer chips to train the AI
models forced deep Seek's hand, force them into a corner
and force them to get there quicker.
Speaker 1 (03:31):
So perverse indeed, so can you just because I'm still
struggling a bit with pure reinforcement earning. Is this a
person is using the AI app and the machine gets
an answer right and you take, yes, you've got that right,
and then understands that it's doing something right or is
this a different process?
Speaker 2 (03:48):
No, So that that's the inferencing, that's when we're using that.
This is the pre training, so the training of the model.
So really simplistically, you know that the models have been
getting bigger and bigger, this is excluding deep Seek, and
they've been getting trained on you know, most of the
text on the internet in the world, and a lot
of the photos and a lot of the numbers and
a lot of the videos, and they've been getting taught
(04:11):
to solve every puzzle on Earth using every bit of data.
But this, this deep Seek technique was, like I said,
like getting a digital pad on the back when it
got an answer right when it was training the model.
So not when we're influencing the model. I kind of
think of it as like, instead of studying for a
maths exam by going through all the mass textbooks, you
(04:32):
just do practice exams and get better and better and
better marks until eventually you're getting sort of ninety eight
percent the practice exams, but you don't bother studying in
any of the source material.
Speaker 1 (04:41):
Wow, and then you're just reinforcing, like teaching your dog
how to set with a piece of every time. Okay,
so let's talk Microsoft and MESA. How did their results
look today? What are they saying about all of this?
Speaker 2 (04:55):
Yeah, so they Microsoft said that this happens every technology
technology cycle, and we see sort of ten x improvements
and costs and efficiency. In other words, costs come down
by ninety percent in most tech cycles. And they also
said Deep Seak have have had some real innovations, so
they're underwriting the innovations, some of this reinforcement learning and
some of the other things we haven't discussed that will
(05:18):
likely get copied almost immediately. Microsoft sophomore or less said
that verbatim, and they also said the big beneficiaries of
these normal sort of big step changes and efficiency are
us Ryan, so the customers as prices come down now
Meta and Zuck he sounded a bit flat footed, so
he said, deep seek to a number of novel things.
(05:40):
We are still digesting, but Meta will copy them. It's
too early to have a strong opinion on what this
means for capital spending levels. So some pretty interesting comments
from big tech there.
Speaker 1 (05:52):
Especially when you consider how much they're planning. I think,
was it eighty seven billion dollars or something they were
planning on spending on AI?
Speaker 2 (05:59):
So well, Microsoft, Yeah, Microsoft's eighty billion dollars of capex total,
but yes, a lot of that will be data sinkers
and AI chips, and met Is underwrote a sort of
sixty billion dollar check a few days ago, even.
Speaker 1 (06:13):
Before its results, because it'll surely have a huge effect
on where you put that right.
Speaker 2 (06:18):
Well, that that's why I do think it's really interesting
that Zuckerberg's saying we are still digesting and it's too
early to have a strong opinion on what this means
for capital spinning levels, despite the fact that two days
or two or three days ago, in the last week,
he underwrote a sixty billion dollar CAPEX.
Speaker 1 (06:33):
Program exactly right, So what are the implications for investors? See,
how should investors think about this?
Speaker 2 (06:39):
Well, I think we do need to answer sort of
four or five critical questions which we don't have the
answer to yet. So firstly, that the model has only
been out for it less than a week. This is
a very simple one. So let's see other reviews go
from all these brand new customers. You know, it went
from number thirty one on the air collapse sort at one.
I've tested it, Ryan, and it's always jammed and tells
me to try again later. So let's see whether the
(07:00):
consumers are loving the model. The second one is does
this mean that companies like in Video and AMD you
guys have spoken about this earlier in the week, have
to cut the prices of their cutting edge accelerated compute
chips and related to that, the third question is if so,
will there be a big enough spike in demand like
we've seen over previous technological cycles to offset the price
(07:20):
custs Like we saw that with flat screen TVs, we
saw producers get way more efficient at producing them. Prices
of flat screen TVs collapsed and demand skyrocketed. And the
final thing is what does all this mean for power
demand which was exploding for AI data centers. But as
far as investors go, the market is shot first and
we'll answer those questions we discussed later, but it's going
(07:42):
to take time to get answers to those questions, and
if Zuckerberg is still digesting, as he said a few
hours ago, it's hard for the average investor to have
clarity yet. But one final point I'll leave you with
Ryan is I think we can expect more volatility, not
just in stop prices, but in how the China versus
us chipwoar. We discussed last year ratchet's up, so we
really do need to buckle up, I.
Speaker 1 (08:03):
Think, Sam. Thanks for that, Sam, Dicky Fisher Funds. Great
to have you on.
Speaker 2 (08:06):
As always, for more from Heather Duplessy Allen Drive, listen
live to news talks it'd be from four pm weekdays,
or follow the podcast on iHeartRadio
Bryan Bridge Tech Companies while they've been up and down
like a Yo yo this week since the Deep Seek
thing happened. Interesting timing because Microsoft and Meta have reported
their annual results today. Sam Dicky from Fisher Funds is
with us. Sam. Gooday, good evening, right, how are you
going good? Thank you? Quick throwback? What was the straw
that broke the camel's back here?
Speaker 2 (00:22):
Yeah, that's an interesting question because it wasn't all new news,
so that deep Seak has been openly using these innovative
techniques to drive down the cost to train these AI
models over the last year or so. But it was
kind a threefold. It was the release of its most
recent model R one on twenty fifth January, which showed
the significantly improved results, which you guys have spoken about
(00:42):
this week, plus the talking heads picked it up at
Davos later that week. Plus critically, the deep Sea Gap
went from number thirty one on the Apple lapp store
to number one over the weekend, so that what really
was the straw that broke the camel's back? The consumer
voted with their feet.
Speaker 1 (00:59):
So the significant breakthrough, obviously is that they managed to
do this some well, a lot of questions been asked
about how much US tech was used to make this,
how palatable the platform will be for American users or
America you know, users around the rest of the world
outside of China.
Speaker 2 (01:18):
That's right. I mean one of the breakthrough is it
is important to talk about one of the breakthroughs. There
were many, but the simplest one to understand, the one
that stood out was pure reinforcement learning. So in playing English,
pure reinforcement learning is quite simple. So don't teach one
of these AI models to solve every puzzle on Earth
using every bit of data on earth, like how most
of the US large language models have been trained. Just
(01:42):
reward it if it does a good job with a
digital pad on the back, so a plus one for
the right answer and a minus one for the wrong answer.
And this sort of trial and error way of learning
has proved to be far more efficient and cost effective
as the model, just like we do, loves to reward
and kind of sort them out. In terms of the
controversy and people saying they don't believe the six million
dollar cost to train the model, and the controversy around
(02:05):
open AI saying that deep Seat piggybacked off, it's existing
chat GPT models. We don't know. We just know that
the arms race and the chipboard that you and I
spoke about late last years is really hotting up.
Speaker 1 (02:20):
So basically what you're talking about, they're the pure reinforcement
is like positive reinforcement. How did the big, biggest, you know,
AI companies in the world missed this, Well, that's right,
they were certainly aware of it.
Speaker 2 (02:33):
So Google, for example, used pure reinforcement learning. So the
word pure, by the way, is a lot of the
companies use human feedback reinforcement learning learning, but this is
sort of digital reinforcement learning with no human feedback or oversight.
Google used that technique with one of its much older
models called Alpha Go in twenty seventeen. But this pure
(02:55):
reinforcement learning technique historically had shortcomings and western which we
can go into it in a minute. But wesn AI
companies either weren't quick enough to solve those shortcomings or
probably weren't incentivized to solve them. Given it was an
AI boom, capital was freely available, and they had access
to cutting edge AI chips, which the Chinese did not.
(03:16):
And it seems like perversely the US restricting China from
accessing these cutting edge computer chips to train the AI
models forced deep Seek's hand, force them into a corner
and force them to get there quicker.
Speaker 1 (03:31):
So perverse indeed, so can you just because I'm still
struggling a bit with pure reinforcement earning. Is this a
person is using the AI app and the machine gets
an answer right and you take, yes, you've got that right,
and then understands that it's doing something right or is
this a different process?
Speaker 2 (03:48):
No, So that that's the inferencing, that's when we're using that.
This is the pre training, so the training of the model.
So really simplistically, you know that the models have been
getting bigger and bigger, this is excluding deep Seek, and
they've been getting trained on you know, most of the
text on the internet in the world, and a lot
of the photos and a lot of the numbers and
a lot of the videos, and they've been getting taught
(04:11):
to solve every puzzle on Earth using every bit of data.
But this, this deep Seek technique was, like I said,
like getting a digital pad on the back when it
got an answer right when it was training the model.
So not when we're influencing the model. I kind of
think of it as like, instead of studying for a
maths exam by going through all the mass textbooks, you
(04:32):
just do practice exams and get better and better and
better marks until eventually you're getting sort of ninety eight
percent the practice exams, but you don't bother studying in
any of the source material.
Speaker 1 (04:41):
Wow, and then you're just reinforcing, like teaching your dog
how to set with a piece of every time. Okay,
so let's talk Microsoft and MESA. How did their results
look today? What are they saying about all of this?
Speaker 2 (04:55):
Yeah, so they Microsoft said that this happens every technology
technology cycle, and we see sort of ten x improvements
and costs and efficiency. In other words, costs come down
by ninety percent in most tech cycles. And they also
said Deep Seak have have had some real innovations, so
they're underwriting the innovations, some of this reinforcement learning and
some of the other things we haven't discussed that will
(05:18):
likely get copied almost immediately. Microsoft sophomore or less said
that verbatim, and they also said the big beneficiaries of
these normal sort of big step changes and efficiency are
us Ryan, so the customers as prices come down now
Meta and Zuck he sounded a bit flat footed, so
he said, deep seek to a number of novel things.
(05:40):
We are still digesting, but Meta will copy them. It's
too early to have a strong opinion on what this
means for capital spending levels. So some pretty interesting comments
from big tech there.
Speaker 1 (05:52):
Especially when you consider how much they're planning. I think,
was it eighty seven billion dollars or something they were
planning on spending on AI?
Speaker 2 (05:59):
So well, Microsoft, Yeah, Microsoft's eighty billion dollars of capex total,
but yes, a lot of that will be data sinkers
and AI chips, and met Is underwrote a sort of
sixty billion dollar check a few days ago, even.
Speaker 1 (06:13):
Before its results, because it'll surely have a huge effect
on where you put that right.
Speaker 2 (06:18):
Well, that that's why I do think it's really interesting
that Zuckerberg's saying we are still digesting and it's too
early to have a strong opinion on what this means
for capital spinning levels, despite the fact that two days
or two or three days ago, in the last week,
he underwrote a sixty billion dollar CAPEX.
Speaker 1 (06:33):
Program exactly right, So what are the implications for investors? See,
how should investors think about this?
Speaker 2 (06:39):
Well, I think we do need to answer sort of
four or five critical questions which we don't have the
answer to yet. So firstly, that the model has only
been out for it less than a week. This is
a very simple one. So let's see other reviews go
from all these brand new customers. You know, it went
from number thirty one on the air collapse sort at one.
I've tested it, Ryan, and it's always jammed and tells
me to try again later. So let's see whether the
(07:00):
consumers are loving the model. The second one is does
this mean that companies like in Video and AMD you
guys have spoken about this earlier in the week, have
to cut the prices of their cutting edge accelerated compute
chips and related to that, the third question is if so,
will there be a big enough spike in demand like
we've seen over previous technological cycles to offset the price
(07:20):
custs Like we saw that with flat screen TVs, we
saw producers get way more efficient at producing them. Prices
of flat screen TVs collapsed and demand skyrocketed. And the
final thing is what does all this mean for power
demand which was exploding for AI data centers. But as
far as investors go, the market is shot first and
we'll answer those questions we discussed later, but it's going
(07:42):
to take time to get answers to those questions, and
if Zuckerberg is still digesting, as he said a few
hours ago, it's hard for the average investor to have
clarity yet. But one final point I'll leave you with
Ryan is I think we can expect more volatility, not
just in stop prices, but in how the China versus
us chipwoar. We discussed last year ratchet's up, so we
really do need to buckle up, I.
Speaker 1 (08:03):
Think, Sam. Thanks for that, Sam, Dicky Fisher Funds. Great
to have you on.
Speaker 2 (08:06):
As always, for more from Heather Duplessy Allen Drive, listen
live to news talks it'd be from four pm weekdays,
or follow the podcast on iHeartRadio
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Special Summer Offer: Exclusively on Apple Podcasts, try our Dateline Premium subscription completely free for one month! With Dateline Premium, you get every episode ad-free plus exclusive bonus content.