Latest / Best of Business / Sam Dickie: Fisher Funds expert on the ongoing volatility in the tech market

Transcript

Sam Dickie: Fisher Funds expert on the ongoing volatility in the tech market · Best of Business

• 0:09
  Speaker 1: You're listening to a podcast from News Talk ZEDB. Follow
• 0:12
  Speaker 1: this and our wide range of podcasts now on iHeartRadio,
• 0:17
  Speaker 1: Ryan Bridge.
• 0:18
  Speaker 2: Tech companies while they've been up and down like a
• 0:20
  Speaker 2: Yo yo this week since the deep Seek thing happened.
• 0:23
  Speaker 2: Interesting timing because Microsoft and Meta have reported about annual
• 0:27
  Speaker 2: results today. Sam Dicky from Fisher Funds is with us. Sam, gooday,
• 0:31
  Speaker 2: good eating?
• 0:32
  Speaker 3: Right? How are you going good?
• 0:33
  Speaker 2: Thank you? Quick throwback? What was the straw that broke
• 0:36
  Speaker 2: the camel's back here?
• 0:38
  Speaker 4: Yeah, that's an interesting question because it wasn't all new news,
• 0:41
  Speaker 4: so that deep Seak has been openly using these innovative
• 0:45
  Speaker 4: techniques to drive down the cost to train these AI
• 0:48
  Speaker 4: models over the last year or so.
• 0:50
  Speaker 3: But it was got a threefold.
• 0:51
  Speaker 4: It was the release of its most recent model R
• 0:53
  Speaker 4: one on twenty fifth January, which showed that the significantly
• 0:56
  Speaker 4: improved results, which you guys have spoken about this week,
• 0:59
  Speaker 4: plus the talking heads picked it up at Davos later
• 1:02
  Speaker 4: that week. Plus critically, the deep Sea Gap went from
• 1:06
  Speaker 4: number thirty one on the Apple App Store to number
• 1:08
  Speaker 4: one over the weekend, so that what really was the
• 1:10
  Speaker 4: straw that broke the camel's back?
• 1:12
  Speaker 3: The consumer voted with their feet.
• 1:16
  Speaker 2: So the significant breakthrough obviously is that they managed to
• 1:20
  Speaker 2: do this some well, A lot of questions been asked
• 1:23
  Speaker 2: about how much US tech was used to make this,
• 1:26
  Speaker 2: how palatable the platform will be for American users or
• 1:30
  Speaker 2: America you know, users around the rest of the world
• 1:32
  Speaker 2: outside of China.
• 1:35
  Speaker 4: That's right, I mean one of the breakthrough is it
• 1:36
  Speaker 4: is important to talk about one of the breakthroughs. There
• 1:39
  Speaker 4: were many, but the simplest one to understand, the one
• 1:42
  Speaker 4: that stood out was pure reinforcement learning.
• 1:45
  Speaker 3: So in plain English, pure reinforcement learning is quite simple.
• 1:48
  Speaker 4: So don't teach one of these AI models to solve
• 1:51
  Speaker 4: every puzzle on Earth using every bit of data on earth,
• 1:54
  Speaker 4: like how most of the US large language models have
• 1:57
  Speaker 4: been trained. Just reward it if it does a good
• 2:00
  Speaker 4: job with a digital pad on the back, so a
• 2:02
  Speaker 4: plus one for the right answer and a minus one
• 2:04
  Speaker 4: for the wrong answer. And this sort of trial and
• 2:07
  Speaker 4: error way learning has proved to be far more efficient
• 2:10
  Speaker 4: and cost effective as the model just like we do
• 2:12
  Speaker 4: love to.
• 2:12
  Speaker 3: Reward and kind of sort them out.
• 2:14
  Speaker 4: In terms of the controversy and people saying they don't
• 2:17
  Speaker 4: believe the six million dollar cost to train the model
• 2:20
  Speaker 4: and the controversy around open AI saying that deep seat
• 2:25
  Speaker 4: piggybacked off it's existing chat GPT models.
• 2:29
  Speaker 3: We don't know.
• 2:30
  Speaker 4: We just know that the arms race and the chipwoar
• 2:33
  Speaker 4: that you and I spoke about late last year as
• 2:35
  Speaker 4: is really hotting up.
• 2:36
  Speaker 2: So basically what you're talking about, they're the pure reinforcement
• 2:40
  Speaker 2: is like positive reinforcement. How did the big, biggest, you know,
• 2:45
  Speaker 2: AI companies in the world.
• 2:46
  Speaker 3: Missed this, Well, that's right, they were certainly aware of it.
• 2:49
  Speaker 4: So Google, for example, used pure reinforcement learning. So the
• 2:53
  Speaker 4: word pure, by the way, is a lot of the
• 2:55
  Speaker 4: companies use human feedback reinforcement learning learning, but this is
• 3:00
  Speaker 4: sort of digital reinforcement learning with no human feedback or oversight.
• 3:06
  Speaker 4: Google used that technique with one of its much older
• 3:08
  Speaker 4: models called Alpha Go in twenty seventeen. But this pure
• 3:12
  Speaker 4: reinforcement learning technique historically had shortcomings and Western which we
• 3:16
  Speaker 4: can go into in a minute. But Western AI companies
• 3:18
  Speaker 4: either weren't quick enough to solve those shortcomings or probably
• 3:21
  Speaker 4: weren't incentivized to solve them.
• 3:23
  Speaker 3: Given it was an.
• 3:23
  Speaker 4: AI boom, capital was freely available, and they had access
• 3:28
  Speaker 4: to cutting edge AI chips, which the Chinese did not,
• 3:33
  Speaker 4: and it seems like perversely, the US restricting China from
• 3:36
  Speaker 4: accessing these cutting edge computer chips to train the AI
• 3:40
  Speaker 4: models forced Deep Seek's hand, force them into a corner
• 3:44
  Speaker 4: and force them to get there quicker.
• 3:48
  Speaker 2: Perverse. Indeed, so can you just because I'm still struggling
• 3:51
  Speaker 2: a bit with pure reinforcement eLearning. Is this a person
• 3:55
  Speaker 2: is using the AI app and the machine gets an
• 3:58
  Speaker 2: answer right, and you take, yes, you've got that right,
• 4:01
  Speaker 2: and then understands that it's doing something right or is
• 4:03
  Speaker 2: it a different process.
• 4:05
  Speaker 4: No, So that's the inferencing, that's when we're using that.
• 4:07
  Speaker 4: This is the pre training, so the training of the model.
• 4:10
  Speaker 4: So really simplistically, you know that the models have been
• 4:14
  Speaker 4: getting bigger and bigger, this is excluding deep Seek, and
• 4:17
  Speaker 4: they've been getting trained on you know, most of the
• 4:20
  Speaker 4: text on the internet in the world, and a lot
• 4:22
  Speaker 4: of the photos and a lot of the numbers, and
• 4:24
  Speaker 4: a lot of the videos, and they've been getting taught
• 4:27
  Speaker 4: to solve every puzzle on Earth using every bit of data.
• 4:32
  Speaker 4: But this, this deep Seek technique was, like I said,
• 4:36
  Speaker 4: like getting a digital pad on the back when it
• 4:38
  Speaker 4: got an answer right, when it was training the model,
• 4:40
  Speaker 4: so not when we're influencing the model. I kind of
• 4:43
  Speaker 4: think of it as like, instead of studying for a
• 4:46
  Speaker 4: mass exam by going through all the mass textbooks, you
• 4:49
  Speaker 4: just do practice exams and get better and better and
• 4:51
  Speaker 4: better marks until eventually you're getting sort of ninety eight
• 4:53
  Speaker 4: percent the practice exams. But you don't bother studying in
• 4:56
  Speaker 4: any of the source material.
• 4:58
  Speaker 2: Wow, and then you're just reinforcing, like teaching your dog
• 5:01
  Speaker 2: how to set every time. Okay, so let's talk Microsoft
• 5:07
  Speaker 2: and MESA. How did their results look today? What are
• 5:09
  Speaker 2: they saying about all of this?
• 5:12
  Speaker 4: Yeah, so they Microsoft said that this happens every technology
• 5:15
  Speaker 4: technology cycle, and we see sort of ten x improvements
• 5:19
  Speaker 4: and costs and efficiency. In other words, costs come down
• 5:21
  Speaker 4: by ninety percent and most tech cycles. And they also
• 5:26
  Speaker 4: said deep Seak have have had some real innovations, so
• 5:28
  Speaker 4: they're underwriting the innovations. Some of this reinforcement learning and
• 5:32
  Speaker 4: some of the other things we haven't discussed that will
• 5:34
  Speaker 4: likely get copied almost immediately. Microsoft sophomore or less said
• 5:38
  Speaker 4: that verbatim, and they also said the big beneficiaries of
• 5:41
  Speaker 4: these normal sort of big step changes and efficiency are
• 5:46
  Speaker 4: us Ryan so the customers as prices come down. Now
• 5:49
  Speaker 4: Meta and zuck he sounded a bit flat footed, so
• 5:53
  Speaker 4: he said, deep seek to a number of novel things.
• 5:57
  Speaker 4: We are still digesting, but Meta will copy them. It's
• 6:02
  Speaker 4: too early to have a strong opinion on what this
• 6:04
  Speaker 4: means for capital spending levels. So some pretty interesting comments
• 6:08
  Speaker 4: from being take there.
• 6:09
  Speaker 2: Especially we can when you consider how much they're planning
• 6:11
  Speaker 2: I think was it eighty seven billion dollars or something
• 6:13
  Speaker 2: they were planning on spending on AI so well Microxoft.
• 6:18
  Speaker 4: Yeah, Microsoft's eighty billion dollars of capex total, but yes,
• 6:20
  Speaker 4: a lot of that will be datas and AI chips
• 6:23
  Speaker 4: and Mets underwrote a sort of sixty billion dollar check
• 6:29
  Speaker 4: a few days ago, even before.
• 6:30
  Speaker 2: Its results, because it'll surely have a huge effect on
• 6:33
  Speaker 2: where you put that right.
• 6:35
  Speaker 4: Well, that that's why I do think it's really interesting
• 6:37
  Speaker 4: that Zuckerberg's saying we are still digesting and it's too
• 6:41
  Speaker 4: early to have a strong opinion on what this means
• 6:43
  Speaker 4: for capital spending levels, despite the fact that two days
• 6:46
  Speaker 4: or two or three days ago, in the last week,
• 6:47
  Speaker 4: he underwrote a sixty billion dollar CAPEX.
• 6:49
  Speaker 2: Program exactly right, So what are the implications for investors.
• 6:54
  Speaker 2: See how should investors think about this?
• 6:56
  Speaker 4: Well, I think we do need to answer sort of
• 6:58
  Speaker 4: four or five critical questions which we don't have the
• 7:00
  Speaker 4: answer to yet. So firstly, that the model has only
• 7:03
  Speaker 4: been out for less than a week. This is a
• 7:04
  Speaker 4: very simple one. So let's see other reviews. Go from
• 7:06
  Speaker 4: all these brand new customers you know, went from number
• 7:08
  Speaker 4: thirty one on the air collapse sort at one. I've
• 7:11
  Speaker 4: tested it, Ryan, and it's always jammed and tells me
• 7:14
  Speaker 4: to try again later. So let's see whether the consumers
• 7:17
  Speaker 4: are loving the model. The second one is does this
• 7:20
  Speaker 4: mean that companies like in Video and AMD you guys
• 7:22
  Speaker 4: have spoken about this earlier in the week have to
• 7:24
  Speaker 4: cut the prices of their cutting edge accelerated compute chips?
• 7:28
  Speaker 4: And related to that, the third question is if so,
• 7:30
  Speaker 4: will there be a big enough spike in demand like
• 7:33
  Speaker 4: we've seen over previous technological cycles to offset the price cuts?
• 7:37
  Speaker 4: Like we saw that with flat screen TVs, we saw
• 7:40
  Speaker 4: producers get way more efficient at producing them. Prices of
• 7:43
  Speaker 4: flat screen TVs collapsed and demand skyrocketed. And the final
• 7:47
  Speaker 4: thing is what does all this mean for power demand
• 7:49
  Speaker 4: which was exploding for AI data centers. But as far
• 7:53
  Speaker 4: as investors go, the market is shot first, and we'll
• 7:55
  Speaker 4: answer those questions we discussed later, but it's going to
• 7:58
  Speaker 4: take time to get answers to those questions, and if
• 8:01
  Speaker 4: Zuckerberg is still digesting, as he said a few hours ago,
• 8:05
  Speaker 4: it's hard for the average investor to have clarity yet.
• 8:07
  Speaker 4: But one final point I'll leave you with Ryan is
• 8:09
  Speaker 4: I think we can expect more volatility, not just in
• 8:11
  Speaker 4: stop prices, but in how the China versus US chip
• 8:15
  Speaker 4: board we discussed last year ratchets up, so we really
• 8:18
  Speaker 4: do need to buckle up, I think Sam.
• 8:20
  Speaker 2: Thanks for that, Sam, Dicky Fisher Funds. Great to have
• 8:22
  Speaker 2: you on as always for.
• 8:24
  Speaker 1: More from news Talks at b Listen live on air
• 8:27
  Speaker 1: or online, and keep our shows with you wherever you
• 8:29
  Speaker 1: go with our podcasts on iHeartRadio