WEBVTT

00:00:02.139 --> 00:00:04.780
Welcome to the AI Chat Podcast. I'm your host,

00:00:04.860 --> 00:00:07.179
Jaden Schaefer. Every day I cover cutting -edge

00:00:07.179 --> 00:00:09.320
AI news and talk with the leaders behind it,

00:00:09.339 --> 00:00:11.300
breaking down what it means for your life and

00:00:11.300 --> 00:00:17.920
business. Anthropic is building their own custom

00:00:17.920 --> 00:00:21.219
silicone team to design their own AI chips, and

00:00:21.219 --> 00:00:24.640
AMD has just unveiled something called the Helios

00:00:24.640 --> 00:00:27.980
Rack system. Their goal is to challenge NVIDIA

00:00:27.980 --> 00:00:31.739
head -on in AI data centers. Anthropic is upgrading

00:00:31.739 --> 00:00:33.960
Claude's voice mode. It's going to be now running

00:00:33.960 --> 00:00:37.320
on Opus and Sonnet. And Google Cloud Revenue

00:00:37.320 --> 00:00:42.159
jumped 82 % to $24 .8 billion. If anyone has

00:00:42.159 --> 00:00:43.799
been following Google for the last few weeks

00:00:43.799 --> 00:00:45.920
when they've released how much their AI spending

00:00:45.920 --> 00:00:48.299
is and their stock took a massive hit, well,

00:00:48.420 --> 00:00:50.439
evidently, this is the reason why they announced

00:00:50.439 --> 00:00:53.229
that. OpenAI and Anthropic are both lobbying

00:00:53.229 --> 00:00:57.109
Washington to curb Chinese openweight AI. We've

00:00:57.109 --> 00:00:58.869
talked about this a lot on the podcast, but this

00:00:58.869 --> 00:01:02.090
is kind of the latest news on the lobbying front.

00:01:02.310 --> 00:01:05.250
If you want to get a deep dive on these and a

00:01:05.250 --> 00:01:07.989
dozen more stories every single day, I'd love

00:01:07.989 --> 00:01:10.370
for you to check out AIChatDaily .com. That's

00:01:10.370 --> 00:01:12.950
my own sister website, news site that is partnered

00:01:12.950 --> 00:01:15.769
with this podcast. It's essentially like a show

00:01:15.769 --> 00:01:17.689
notes, but it also goes deeper and has extra

00:01:17.689 --> 00:01:20.049
stories I don't cover in here. There's dozens

00:01:20.049 --> 00:01:22.390
of stories. every single day on all the latest

00:01:22.390 --> 00:01:25.469
AI news. And there's a newsletter that you can

00:01:25.469 --> 00:01:27.549
subscribe to on the website. So every single

00:01:27.549 --> 00:01:29.829
morning, if you like, you'll get the top five

00:01:29.829 --> 00:01:32.670
AI news stories straight into your inbox to keep

00:01:32.670 --> 00:01:34.670
you up to date with everything happening in AI.

00:01:34.849 --> 00:01:37.890
And you get some really quick breakdowns of all

00:01:37.890 --> 00:01:39.870
of the facts so that you understand what's going

00:01:39.870 --> 00:01:43.840
on. The website is AIChatDaily .com. The first

00:01:43.840 --> 00:01:45.599
story I wanted to talk about is the fact that

00:01:45.599 --> 00:01:48.299
Anthropic is building their own AI chip design

00:01:48.299 --> 00:01:51.000
team, and it's going to be for creating custom

00:01:51.000 --> 00:01:53.319
hardware for Claude. It's the same thing that

00:01:53.319 --> 00:01:55.400
OpenAI, Google, and Meta, they're all doing it

00:01:55.400 --> 00:01:58.099
right now. And the big reason is everyone wants

00:01:58.099 --> 00:02:00.840
to get off of NVIDIA. These companies are incredibly

00:02:00.840 --> 00:02:03.640
dependent on NVIDIA. NVIDIA basically has quotas

00:02:03.640 --> 00:02:05.719
and they dole out their GPUs and their chips,

00:02:05.760 --> 00:02:07.620
and they only give a certain amount to different

00:02:07.620 --> 00:02:10.099
companies, and nobody wants to feel bottlenecked

00:02:10.099 --> 00:02:12.620
by one player that they're obviously spending.

00:02:12.650 --> 00:02:15.189
a ton of money on. So everyone's building their

00:02:15.189 --> 00:02:18.430
own chips. Right now, they're also scouting Samsung

00:02:18.430 --> 00:02:21.750
as a manufacturing partner, which is ending Anthropic

00:02:21.750 --> 00:02:24.810
status as the basically the last major AI lab

00:02:24.810 --> 00:02:28.409
without a public silicone program. They're like,

00:02:28.490 --> 00:02:30.370
all right, it's time to get serious on that one.

00:02:30.569 --> 00:02:33.409
Anthropic currently relies on a bunch of different

00:02:33.409 --> 00:02:36.169
suppliers. So there's AWS, Google, NVIDIA, AMD,

00:02:36.389 --> 00:02:39.129
all of them give them compute. But of course,

00:02:39.150 --> 00:02:40.550
that means that they don't get to control the

00:02:40.550 --> 00:02:43.210
pricing allocation. or any of the roadmap timing.

00:02:43.469 --> 00:02:46.189
OpenAI shipped their Broadcom designed chip,

00:02:46.310 --> 00:02:49.110
which is called Jalapeno. It's an inference chip,

00:02:49.189 --> 00:02:52.189
and that was in June of this year. Google now

00:02:52.189 --> 00:02:56.129
runs their own TPUs. Meta is building MTIA accelerators.

00:02:56.229 --> 00:02:58.050
And Anthropic basically was the only one that

00:02:58.050 --> 00:03:00.509
wasn't working on this. Custom silicone typically

00:03:00.509 --> 00:03:03.530
takes between 18 to 24 months to design, and

00:03:03.530 --> 00:03:05.310
it's going to cost hundreds of millions of dollars

00:03:05.310 --> 00:03:08.490
per chip version. But every single efficiency

00:03:08.490 --> 00:03:10.949
gain that they can get on this compounds across...

00:03:11.080 --> 00:03:14.379
billions of dollars that Anthropic is using for

00:03:14.379 --> 00:03:17.060
these Claude API calls. So I think this is something

00:03:17.060 --> 00:03:19.319
that is kind of a play on margin. It's not a

00:03:19.319 --> 00:03:21.039
shortcut. This is going to take a long time.

00:03:21.259 --> 00:03:24.039
Anthropic can tune memory and math formats to

00:03:24.039 --> 00:03:27.219
Claude's exact workloads in a way that no off

00:03:27.219 --> 00:03:29.639
-the -shelf GPU can really match. But they're

00:03:29.639 --> 00:03:32.180
also betting on a two -year runway before any

00:03:32.180 --> 00:03:33.719
of that investment is going to get paid back.

00:03:33.840 --> 00:03:35.539
So this isn't something that's very quick, but

00:03:35.539 --> 00:03:37.479
over the long run, this might be a good play

00:03:37.479 --> 00:03:40.199
for them. After announcing a strategic partner,

00:03:40.300 --> 00:03:44.439
with in Anthropic. AMD has just unveiled Helios,

00:03:44.520 --> 00:03:47.479
which is a rack system for AI data centers, and

00:03:47.479 --> 00:03:49.120
they're saying that this is basically a direct

00:03:49.120 --> 00:03:52.240
competitor to NVIDIA's products that they have

00:03:52.240 --> 00:03:55.580
right now. Microsoft, OpenAI, Meta, Oracle, and

00:03:55.580 --> 00:03:58.400
Anthropic have all committed to deploying it

00:03:58.400 --> 00:04:00.979
at a gigawatt scale, and it's going to be starting

00:04:00.979 --> 00:04:05.599
shipments later this year. Anthropic and AMD

00:04:05.599 --> 00:04:07.780
both announced a strategic partnership to deploy

00:04:07.780 --> 00:04:10.900
up to two gigawatts of GPUs on Helios, which

00:04:10.900 --> 00:04:13.800
is, you know, the exact same scale of Nvidia's

00:04:13.800 --> 00:04:15.979
largest AI training cluster. So they're, you

00:04:15.979 --> 00:04:17.519
know, they're really being quite competitive

00:04:17.519 --> 00:04:20.699
in the market. AMD also introduced Venice X,

00:04:20.819 --> 00:04:23.699
which is a data center CPU designed to pair with

00:04:23.699 --> 00:04:26.259
Helios. And that's going to be launching in 2027.

00:04:26.399 --> 00:04:29.600
And it's going to compete on kind of this full

00:04:29.600 --> 00:04:31.959
stack alternative to a lot of the different competitors

00:04:31.959 --> 00:04:34.800
that are out there. The CEO, Lisa Su, projected

00:04:34.800 --> 00:04:38.459
the AI accelerator market will reach 1 .4 trillion.

00:04:38.730 --> 00:04:41.370
by 2030, approaching the size of today's entire

00:04:41.370 --> 00:04:43.870
semiconductor industry. So, you know, this one

00:04:43.870 --> 00:04:45.509
segment of the market she's saying is going to

00:04:45.509 --> 00:04:47.430
be as big as the entire semiconductor industry.

00:04:47.709 --> 00:04:49.870
This is the first time that AMD has landed all

00:04:49.870 --> 00:04:52.889
five major frontier AI labs on a single rack

00:04:52.889 --> 00:04:55.050
platform. They have everyone signed up for it.

00:04:55.129 --> 00:04:57.870
I think this is showing that they're really like

00:04:57.870 --> 00:05:00.649
a strong contender for a quote unquote second

00:05:00.649 --> 00:05:03.889
place option for a lot of these massive AI infrastructure

00:05:03.889 --> 00:05:06.829
buildouts that are coming up. The scale of all

00:05:06.829 --> 00:05:08.569
of these different deals. that they've done with

00:05:08.569 --> 00:05:10.970
these five hyperscalers. I think they're all

00:05:10.970 --> 00:05:12.889
measured in gigawatts, and I think that shows

00:05:12.889 --> 00:05:15.089
that the hyperscalers are, all of them are ready

00:05:15.089 --> 00:05:17.649
to diversify away from NVIDIA's almost complete

00:05:17.649 --> 00:05:20.350
dominance. So it looks like we're having a real

00:05:20.350 --> 00:05:22.870
second player in the market, and AMD is gonna

00:05:22.870 --> 00:05:24.709
get some serious market share. Does that, you

00:05:24.709 --> 00:05:26.509
know, is that like super bad for NVIDIA? I think

00:05:26.509 --> 00:05:28.029
their market is just gonna keep growing like

00:05:28.029 --> 00:05:31.329
crazy, but there are other competitors now. Anthropic

00:05:31.329 --> 00:05:33.790
is upgrading Claude's voice mode. It's now gonna

00:05:33.790 --> 00:05:36.269
work with their most powerful models, Opus and

00:05:36.269 --> 00:05:39.040
Sonnet. So I think in the past, it was only working

00:05:39.040 --> 00:05:41.220
with Haiku. And I will say one thing that I actually

00:05:41.220 --> 00:05:43.980
did appreciate about this was I've been going

00:05:43.980 --> 00:05:46.939
back and forth between OpenAI and Anthropic as

00:05:46.939 --> 00:05:49.920
far as using like Cloud Cowork and ChatGPT work.

00:05:50.300 --> 00:05:53.120
And I did find that ChatGPT work after they did

00:05:53.120 --> 00:05:55.279
their upgrade and everything's running on Sol

00:05:55.279 --> 00:05:59.939
5 .6 extra high. When I go and ask it to complete

00:05:59.939 --> 00:06:01.879
a project for me and I'm doing the voice mode

00:06:01.879 --> 00:06:04.379
on it, it takes a while for it to kind of transcribe

00:06:04.379 --> 00:06:05.839
and write it all down. And it's actually kind

00:06:05.839 --> 00:06:08.199
of annoying. the the leg it's a few second leg

00:06:08.199 --> 00:06:10.500
and i noticed that it doesn't always happen but

00:06:10.500 --> 00:06:13.620
when you have uh soul 5 .6 extra high on it does

00:06:13.620 --> 00:06:16.660
and so i actually appreciated that um anthropic

00:06:16.660 --> 00:06:19.120
when i was using claude it was like instantaneous

00:06:19.120 --> 00:06:21.019
transcription of what i was doing it was all

00:06:21.019 --> 00:06:23.819
on haiku now is it perfect is it the is it the

00:06:23.819 --> 00:06:25.399
best like there's a couple things that i might

00:06:25.399 --> 00:06:27.720
mess up but it was instantaneous and super fast

00:06:27.720 --> 00:06:29.620
well they're going to upgrade it because evidently

00:06:29.620 --> 00:06:31.040
they think that it needs to be a little bit higher

00:06:31.040 --> 00:06:33.480
quality so now they're putting it on opus and

00:06:33.480 --> 00:06:35.680
sonnet i hope they don't sacrifice too much the

00:06:35.680 --> 00:06:37.939
speed because honestly, that was my favorite

00:06:37.939 --> 00:06:40.100
thing. They're going to be able to they're allowing

00:06:40.100 --> 00:06:41.980
you to plug that directly into Gmail, Google

00:06:41.980 --> 00:06:44.439
Calendar, Slack, Canva, Notion. So you can now

00:06:44.439 --> 00:06:47.079
tell Claude by voice to do your emails. I mean,

00:06:47.079 --> 00:06:49.139
everything else that you've probably been doing.

00:06:49.199 --> 00:06:52.620
I even have like a literal I have this device

00:06:52.620 --> 00:06:56.019
from OpenAI. It's like a little keyboard. It

00:06:56.019 --> 00:06:59.500
is, I think, built with work louder. And it literally

00:06:59.500 --> 00:07:01.439
just has a microphone button that I push. So

00:07:01.439 --> 00:07:03.800
I this is the way I've been working a lot lately

00:07:03.800 --> 00:07:06.050
is just talk. to these models and having them

00:07:06.050 --> 00:07:09.110
go and complete tasks for me instead of typing

00:07:09.110 --> 00:07:10.810
it out. I think that's what a lot of people are

00:07:10.810 --> 00:07:12.529
doing. There's a lot of software like Whisperflow

00:07:12.529 --> 00:07:14.110
that's out right now that does that kind of stuff.

00:07:14.290 --> 00:07:17.329
So a lot of different options, but Claude obviously

00:07:17.329 --> 00:07:19.730
decided they wanted to put a big upgrade here.

00:07:20.110 --> 00:07:22.230
Voice mode is now going to default to whatever

00:07:22.230 --> 00:07:25.709
Claude model you last used in text chat. So a

00:07:25.709 --> 00:07:27.370
Sonnet workflow is automatically going to have

00:07:27.370 --> 00:07:30.050
that for the audio, which is kind of what OpenAI

00:07:30.050 --> 00:07:31.509
is doing right now. And I'm not going to lie.

00:07:31.589 --> 00:07:33.540
That's actually. annoying to me. I wish you could

00:07:33.540 --> 00:07:36.819
choose the audio. Honestly, what I find is when

00:07:36.819 --> 00:07:38.500
it does the transcription and when it types it

00:07:38.500 --> 00:07:41.540
out, I would actually rather a faster, worse

00:07:41.540 --> 00:07:44.259
model because you can like many times I'm literally

00:07:44.259 --> 00:07:47.620
using Apple's like voice to text, which is really

00:07:47.620 --> 00:07:51.500
ghetto, not very good. And when I when I do it,

00:07:51.540 --> 00:07:53.060
there's like a bunch of typos, but Claude understands

00:07:53.060 --> 00:07:54.660
and it literally fixes it and does everything

00:07:54.660 --> 00:07:57.240
the way I want. So I find that Claude is very

00:07:57.240 --> 00:07:59.360
good at getting the meaning out. I'd rather just

00:07:59.360 --> 00:08:01.579
have the underlying model better and having.

00:08:01.680 --> 00:08:03.379
the voice transcription faster because I absolutely

00:08:03.379 --> 00:08:05.959
hate talking and then waiting five seconds for

00:08:05.959 --> 00:08:08.060
it to type out before I can hit send. But I mean,

00:08:08.079 --> 00:08:10.860
that's just me being impatient. So I guess the

00:08:10.860 --> 00:08:13.000
quality improvement is probably good, but I wish

00:08:13.000 --> 00:08:14.899
they would let you choose in the settings what

00:08:14.899 --> 00:08:17.759
model was powering your voice. Free users are

00:08:17.759 --> 00:08:19.319
going to keep getting haiku. So maybe I got to

00:08:19.319 --> 00:08:20.980
go switch to being just a free user. Just kidding.

00:08:21.459 --> 00:08:24.560
And there's going to be paid users that get the

00:08:24.560 --> 00:08:26.639
Opus and Sonnet and all of the rest of it. They

00:08:26.639 --> 00:08:28.540
have a lot of different languages supported on

00:08:28.540 --> 00:08:30.199
this, by the way, English, French, German, Hindi,

00:08:30.259 --> 00:08:33.350
Japanese. korean spanish um and it you know of

00:08:33.350 --> 00:08:36.610
course auto detects it so that's fantastic Google

00:08:36.610 --> 00:08:42.169
Cloud revenue is up 82 % to $24 .8 billion last

00:08:42.169 --> 00:08:44.529
quarter. That absolutely crushed the expectations

00:08:44.529 --> 00:08:47.190
that Wall Street had. I think it's proving that

00:08:47.190 --> 00:08:50.090
the company's huge AI spending is actually paying

00:08:50.090 --> 00:08:54.970
off. The $514 billion backlog of signed customer

00:08:54.970 --> 00:08:57.029
contracts, I think is showing that there is a

00:08:57.029 --> 00:08:59.990
huge demand for AI infrastructure and service

00:08:59.990 --> 00:09:02.570
is real and it's locked in for years. So a lot

00:09:02.570 --> 00:09:04.049
of these deals that Google Cloud's doing, it's

00:09:04.049 --> 00:09:05.899
not just for this year or for this month. month,

00:09:05.919 --> 00:09:08.600
but these are multi -year deals. Cloud growth

00:09:08.600 --> 00:09:12.620
is accelerating really fast. It's up from 63

00:09:12.620 --> 00:09:15.500
% last quarter to 82 % this quarter. It's mostly

00:09:15.500 --> 00:09:17.700
driven by a lot of different companies renting

00:09:17.700 --> 00:09:20.440
compute power and buying pre -built AI products

00:09:20.440 --> 00:09:23.360
from them. Gemini reached 950 million monthly

00:09:23.360 --> 00:09:26.059
active users. That's up from 750 two quarters

00:09:26.059 --> 00:09:28.720
ago. So the usage of their own products is accelerating,

00:09:28.840 --> 00:09:31.559
but also the amount of companies signing deals

00:09:31.559 --> 00:09:34.360
to power their AI on their infrastructure is

00:09:34.360 --> 00:09:37.100
also increasing. seen. And Alphabet posted 112

00:09:37.100 --> 00:09:40.080
.1 billion in quarterly profit. That's almost

00:09:40.080 --> 00:09:43.620
four times the 28 .1 billion from the same quarter

00:09:43.620 --> 00:09:49.570
last year. That is wild. 112 billion. up to 21

00:09:49.570 --> 00:09:53.029
billion just one year later that's absolutely

00:09:53.029 --> 00:09:56.190
insane google's 180 billion to 190 billion dollars

00:09:56.190 --> 00:09:58.629
in annual spending on data centers and chips

00:09:58.629 --> 00:10:00.850
is now um you know it's something that kind of

00:10:00.850 --> 00:10:03.149
crashed their stock a couple weeks ago when it

00:10:03.149 --> 00:10:04.909
came out and everyone was like oh my gosh google's

00:10:04.909 --> 00:10:07.049
overspending but i think this is showing there

00:10:07.049 --> 00:10:09.389
is serious demand the demand is accelerating

00:10:09.389 --> 00:10:11.789
like crazy um they know they're going to get

00:10:11.789 --> 00:10:14.509
that money back and so i think they expect that

00:10:14.509 --> 00:10:17.190
2027 is going to be when this massive build out

00:10:17.190 --> 00:10:19.870
starts turning into some real cash generation.

00:10:20.149 --> 00:10:22.110
But in order to get that money, like they have

00:10:22.110 --> 00:10:23.809
to spend, they got to spend the money to make

00:10:23.809 --> 00:10:26.110
the money, but they can see a lot of demand.

00:10:26.269 --> 00:10:29.029
So evidently they're going all in. OpenAI and

00:10:29.029 --> 00:10:31.830
Anthropic are both lobbying US regulators to

00:10:31.830 --> 00:10:35.529
restrict Chinese openweight AI models like Moonshot's

00:10:35.529 --> 00:10:37.950
Kimi. You know, K3 came out a while ago and it

00:10:37.950 --> 00:10:39.570
was all the rage. And then we have the latest

00:10:39.570 --> 00:10:42.950
Quen model came out a few days ago. What they're

00:10:42.950 --> 00:10:44.950
saying here is that there are national security

00:10:44.950 --> 00:10:47.909
concerns. They, of course, are saying this at

00:10:47.909 --> 00:10:49.970
a time when it would be quite convenient. It

00:10:49.970 --> 00:10:53.169
would push these companies towards, you know,

00:10:53.190 --> 00:10:55.690
being the winners in this space. No one could

00:10:55.690 --> 00:10:57.490
really dethrone them. None of these open source

00:10:57.490 --> 00:11:00.110
models could go scorched earth on them. Opening

00:11:00.110 --> 00:11:02.649
eyes, Dean Ball argued the U .S. should create

00:11:02.649 --> 00:11:05.629
regulatory fear uncertainty and doubt around

00:11:05.629 --> 00:11:07.669
open weight models to weaken their competitive

00:11:07.669 --> 00:11:10.909
position um and then he kind of walked that argument

00:11:10.909 --> 00:11:12.769
back but that was like his first thing we've

00:11:12.769 --> 00:11:15.230
heard this in crypto a lot back in the day the

00:11:15.230 --> 00:11:17.629
fud um that he's you know openly saying that

00:11:17.629 --> 00:11:19.029
they should create some fud and he's like no

00:11:19.029 --> 00:11:21.070
no actually it's serious it's not it's not just

00:11:21.070 --> 00:11:25.850
fud for for fud's sake Kimmy's launch at the

00:11:25.850 --> 00:11:28.049
end of last month, I think a lot of people were

00:11:28.049 --> 00:11:30.190
panicking. There was a viral demo showing it

00:11:30.190 --> 00:11:35.649
generating a macOS -like UI mockup in 30 minutes.

00:11:35.889 --> 00:11:38.370
I think a lot of the hype faded after that. A

00:11:38.370 --> 00:11:39.909
lot of these demos are like, oh my gosh, look

00:11:39.909 --> 00:11:42.350
at this crazy, incredible thing. And then once

00:11:42.350 --> 00:11:43.629
it actually comes out and people actually play

00:11:43.629 --> 00:11:44.990
with it, they're like, okay, it's not as crazy

00:11:44.990 --> 00:11:47.570
as it seems, which is why I'm actually a little

00:11:47.570 --> 00:11:50.289
skeptical. I mean, whatever, you can say I'm

00:11:50.289 --> 00:11:52.409
dumb, but it's why I'm a little skeptical of

00:11:52.409 --> 00:11:55.350
like, uh anthropic always hyping up their models

00:11:55.350 --> 00:11:57.409
to an insane degree only giving them to enterprises

00:11:57.409 --> 00:11:59.490
not to the regular people and saying look it

00:11:59.490 --> 00:12:02.120
can do all these crazy things I don't know, it

00:12:02.120 --> 00:12:03.600
makes me skeptical when they won't give it to

00:12:03.600 --> 00:12:05.820
you, but maybe it's truly dangerous. I don't

00:12:05.820 --> 00:12:08.440
know. This is the second time that we've had

00:12:08.440 --> 00:12:11.799
in the last under a year, basically, where we've

00:12:11.799 --> 00:12:14.220
had this kind of cycle where there's a huge cheaper

00:12:14.220 --> 00:12:16.059
model that comes out and it's super capable.

00:12:16.639 --> 00:12:19.779
DeepSeek had something that kind of was along

00:12:19.779 --> 00:12:21.259
these lines. And then afterwards, there's a bunch

00:12:21.259 --> 00:12:22.899
of policy restrictions that people were calling

00:12:22.899 --> 00:12:25.840
for after it came out. So I think we're seeing

00:12:25.840 --> 00:12:27.639
basically the same pattern play out over and

00:12:27.639 --> 00:12:29.659
over again. Chinese labs are going to ship like

00:12:29.659 --> 00:12:32.090
a super cheap open. model. It's going to be really

00:12:32.090 --> 00:12:33.830
competitive in the benchmarks. A lot of the American

00:12:33.830 --> 00:12:35.710
executives are going to start warning about security

00:12:35.710 --> 00:12:38.570
and regulation and we need to have restrictions.

00:12:39.029 --> 00:12:41.409
And of course, the people that are really winning

00:12:41.409 --> 00:12:44.250
are those big closed model companies, OpenAI,

00:12:44.330 --> 00:12:47.549
Anthropic, etc. So I think whether those security

00:12:47.549 --> 00:12:51.049
concerns are legitimate or not, it doesn't change

00:12:51.049 --> 00:12:54.190
the fact that all of the people asking for these

00:12:54.190 --> 00:12:56.090
policy changes happen to be the people that are

00:12:56.090 --> 00:12:58.509
going to be winning from it. So I am a little

00:12:58.509 --> 00:13:00.750
bit skeptical of it, if I'm being honest. And

00:13:00.750 --> 00:13:02.809
I actually think open weight models are fantastic

00:13:02.809 --> 00:13:05.210
for innovation. And like for my own company,

00:13:05.269 --> 00:13:07.850
being able to use a model that's 90, you know,

00:13:07.850 --> 00:13:12.029
90 % cheaper, 90 times cheaper than an open or

00:13:12.029 --> 00:13:13.710
than a closed source model, if it can do some

00:13:13.710 --> 00:13:15.309
of the same tasks that we needed to get down

00:13:15.309 --> 00:13:17.509
on a repeated basis. To me, that makes a lot

00:13:17.509 --> 00:13:20.039
of sense. But There's a lot to debate on this

00:13:20.039 --> 00:13:21.919
topic for sure. Guys, thank you so much for tuning

00:13:21.919 --> 00:13:23.940
into the podcast. If you enjoyed this episode,

00:13:24.120 --> 00:13:25.700
make sure to leave a rating and review wherever

00:13:25.700 --> 00:13:27.720
you get your podcasts. If that's over on Spotify,

00:13:28.000 --> 00:13:29.720
you got to hit the About tab. If it's on Apple,

00:13:29.860 --> 00:13:32.000
you can leave a little comment, drop some stars.

00:13:32.139 --> 00:13:33.779
I know people have complained that Apple's not

00:13:33.779 --> 00:13:36.039
intuitive for leaving reviews, but go to the

00:13:36.039 --> 00:13:38.879
main podcast page, scroll down, hit some stars,

00:13:38.940 --> 00:13:40.840
leave a review. It helps us show out so much.

00:13:40.899 --> 00:13:43.500
I would be incredibly grateful. Go check out

00:13:43.500 --> 00:13:46.179
AIChatDaily .com if you want to get all of these

00:13:46.179 --> 00:13:48.429
news stories and articles. format and if you

00:13:48.429 --> 00:13:50.909
want to get a news brief sent to you every single

00:13:50.909 --> 00:13:53.190
day you can subscribe to our newsletter i will

00:13:53.190 --> 00:13:54.850
catch you guys all in the next episode