Deep Seek AI SMASHES the industry - Error Code 201

WEBVTT

00:00:00.000 --> 00:00:03.080
The entire AI industry got tossed on its fucking

00:00:03.080 --> 00:00:05.960
head this week. A Chinese company released a

00:00:05.960 --> 00:00:10.099
new AI model this week and it appears to be as

00:00:10.099 --> 00:00:13.640
powerful as anything the US AI industry has come

00:00:13.640 --> 00:00:16.660
up with so far and they've done it at a small

00:00:16.660 --> 00:00:18.839
fraction of the computing and, more importantly,

00:00:19.460 --> 00:00:22.500
the energy cost. This sent AI related stocks,

00:00:22.699 --> 00:00:24.480
most specifically Nvidia and a couple of energy

00:00:24.480 --> 00:00:28.829
stocks, just crashing on Monday. This happened

00:00:28.829 --> 00:00:31.809
just after the Trump administration announced

00:00:31.809 --> 00:00:34.490
Project Stargate, which is a $500 billion joint

00:00:34.490 --> 00:00:37.149
venture, to build a whole bunch of AI infrastructure

00:00:37.149 --> 00:00:39.829
that now they're starting to figure out they

00:00:39.829 --> 00:00:44.109
may not need. My question is, is this the greatest

00:00:44.109 --> 00:00:46.689
troll of the American tech industry by China

00:00:46.689 --> 00:00:51.170
ever? Or what do we make of not only the timing,

00:00:51.789 --> 00:00:56.439
but what this could mean with how efficient this

00:00:56.439 --> 00:01:03.640
AI model is supposed to be. Well, I will also

00:01:03.640 --> 00:01:09.459
say that. They were also able to supposedly there's

00:01:09.459 --> 00:01:11.780
a lot of question behind this, but supposedly

00:01:11.780 --> 00:01:14.900
they were able to develop this in like the course

00:01:14.900 --> 00:01:16.859
of I want to say like two months or something

00:01:16.859 --> 00:01:22.000
like in insane like span of time. Yeah. And then

00:01:22.000 --> 00:01:26.219
also a lot cheaper than. They know a lot of the

00:01:26.219 --> 00:01:28.519
American OpenAI or Meta or anybody who's working

00:01:28.519 --> 00:01:31.319
on it. Yeah. Yeah. Which honestly like kind of

00:01:31.319 --> 00:01:34.319
makes sense because like as soon as one person

00:01:34.319 --> 00:01:36.379
kind of paves the way like others are going to

00:01:36.379 --> 00:01:39.519
find like cheaper ways to do it. But and the

00:01:39.519 --> 00:01:41.340
other reason I think that everybody kind of got

00:01:41.340 --> 00:01:44.219
caught off guard is so that the H100 chip that

00:01:44.219 --> 00:01:46.819
Nvidia makes, which is why their stock has been

00:01:46.819 --> 00:01:49.739
skyrocketing the last couple of years, is a primary

00:01:49.739 --> 00:01:52.180
driver behind especially chat GPT and open eyes

00:01:52.180 --> 00:01:55.680
work. Those are export controlled to China. So

00:01:55.680 --> 00:01:57.620
China's not supposed to have those. Now we're

00:01:57.620 --> 00:01:59.480
going to come back to that portion of the topic

00:01:59.480 --> 00:02:04.280
later. But supposedly that the big thing is they're

00:02:04.280 --> 00:02:06.400
saying, hey, look, you guys export control this.

00:02:06.459 --> 00:02:08.900
We found a way to do it without it. Right. The

00:02:08.900 --> 00:02:11.639
other interesting thing is they didn't just release

00:02:11.639 --> 00:02:15.759
it. They put it under MIT's open license. Yeah.

00:02:15.780 --> 00:02:18.580
And said, everybody has access to the source

00:02:18.580 --> 00:02:22.340
code. You can do whatever you want with it. Right.

00:02:22.830 --> 00:02:26.629
which is huge. So they made it open source. They

00:02:26.629 --> 00:02:32.930
gave all the code away for free. Yep. And they

00:02:32.930 --> 00:02:35.969
were like, come at us, bro. Right. And they published

00:02:35.969 --> 00:02:39.349
a big technical document outlining how they did

00:02:39.349 --> 00:02:46.810
it. Obviously, Nvidia stock went way, way down.

00:02:47.770 --> 00:02:50.830
As of right now, it's recovered a teeny, teeny

00:02:50.830 --> 00:02:54.280
bit. But starting the week, it was at about one

00:02:54.280 --> 00:02:57.479
hundred and fifty bucks a share. And it's down

00:02:57.479 --> 00:03:02.599
to one twenty four. Oh, no. They lost. They lost

00:03:02.599 --> 00:03:05.039
15 percent in market cap. So they at one point

00:03:05.039 --> 00:03:07.939
had lost six hundred billion dollars this week.

00:03:08.439 --> 00:03:17.340
OK, that's actually. Yeah, that hurts. They gave

00:03:17.340 --> 00:03:22.210
it away for free. Straight. Yeah, free. And that's

00:03:22.210 --> 00:03:32.349
honestly, that's what I love. I so I don't. Coming

00:03:32.349 --> 00:03:35.129
from like the engineering point of view, I must

00:03:35.129 --> 00:03:38.389
say you are engineer. Yeah, I think this is great.

00:03:38.789 --> 00:03:42.370
I love open source stuff. When I was in college

00:03:42.370 --> 00:03:46.150
for my engineering degree, I used open source

00:03:46.150 --> 00:03:50.319
machine learning for my project. for my senior

00:03:50.319 --> 00:03:54.939
design. I, whenever something is open source,

00:03:54.939 --> 00:04:01.319
it kind of feels like, Hey, we are like, we're

00:04:01.319 --> 00:04:04.379
not so much like worried about profits. Like

00:04:04.379 --> 00:04:07.300
we're putting this technology out there to help,

00:04:07.300 --> 00:04:09.639
like, advance the technology. If that kind of

00:04:09.639 --> 00:04:13.370
makes sense. Yeah. Yeah. Like There could be

00:04:13.370 --> 00:04:15.610
some conspiracy theories out there of like, oh,

00:04:15.610 --> 00:04:17.709
they're trying to crash in a video. They're trying

00:04:17.709 --> 00:04:20.709
to crash the US stock market overall, though,

00:04:20.709 --> 00:04:24.930
as an engineer. So far of everything I've read

00:04:24.930 --> 00:04:27.610
and watched, I'm like, this is a good thing,

00:04:27.610 --> 00:04:29.670
like, because here's what's going to happen.

00:04:30.230 --> 00:04:35.449
It's open source. So legit, every single US company

00:04:35.449 --> 00:04:38.569
that's involved in the AI industry is going to

00:04:38.569 --> 00:04:41.170
take their stuff. And in the next, I don't know,

00:04:41.350 --> 00:04:43.730
two or three months going to start using their

00:04:43.730 --> 00:04:47.029
research to advance their own A .I. like this

00:04:47.029 --> 00:04:50.889
is going to move A .I. forward. Like, yes, it's

00:04:50.889 --> 00:04:53.910
going to hurt stocks right now. But in like six

00:04:53.910 --> 00:04:55.709
months from now, like you're going to see some

00:04:55.709 --> 00:05:00.910
sort of leap in A .I. So I agree with you on

00:05:00.910 --> 00:05:03.389
that. And again, the fact that it's open source

00:05:03.389 --> 00:05:10.350
is is huge. Right. The thing that really interests

00:05:10.350 --> 00:05:12.649
me in a couple of ways about this is this. One

00:05:12.649 --> 00:05:15.529
of the big concerns about AI has been everything

00:05:15.529 --> 00:05:17.889
that we've had up to now requires an immense

00:05:17.889 --> 00:05:21.209
amount of power. Yes. All right. Immense amounts

00:05:21.209 --> 00:05:25.870
of energy to make it function. What I like about

00:05:25.870 --> 00:05:29.089
this, in theory, again, all of this grain of

00:05:29.089 --> 00:05:31.129
salt, I think it's going to take months before

00:05:31.129 --> 00:05:34.449
all this is run down. But in theory, this can

00:05:34.449 --> 00:05:39.509
be run on a local server. without the the crazy

00:05:39.509 --> 00:05:41.930
power demands, which, again, was concerning a

00:05:41.930 --> 00:05:43.610
lot of people who were worried about climate

00:05:43.610 --> 00:05:45.250
change and green energy and stuff like that,

00:05:45.269 --> 00:05:48.029
where it's like crap. You need so much energy

00:05:48.029 --> 00:05:51.449
to build this, you know, to do A .I. And now

00:05:51.449 --> 00:05:53.750
they're going and these Chinese guys are going,

00:05:53.750 --> 00:06:02.230
you really don't. Yeah. Like. If this is true,

00:06:02.670 --> 00:06:07.600
if this is legit, real and not a Oh, God, I forgot

00:06:07.600 --> 00:06:10.540
the girl who promised the world that they could

00:06:10.540 --> 00:06:12.860
test for, like, every single disease in the world.

00:06:13.040 --> 00:06:14.259
Oh, yeah, yeah, yeah, yeah, yeah. I know what

00:06:14.259 --> 00:06:18.399
you're talking about. Elizabeth, what was the

00:06:18.399 --> 00:06:20.379
name of her company? I don't remember. I don't

00:06:20.379 --> 00:06:24.000
know. But like, if it's legit, like this is real,

00:06:24.019 --> 00:06:27.279
like this is huge. I, you know, I'll just say

00:06:27.279 --> 00:06:33.500
it. I'm an electrical engineer. I primarily focused

00:06:33.500 --> 00:06:37.360
on power generation. It's not what I'm currently

00:06:37.360 --> 00:06:41.720
doing, but. All my classes are around power generation.

00:06:43.220 --> 00:06:47.420
No, this is huge like power. It's going to be

00:06:47.420 --> 00:06:51.519
one of the it's like the largest. I would say

00:06:51.519 --> 00:06:54.399
issue with electrical engineering right now is

00:06:54.399 --> 00:06:58.759
how do you keep up with the? I don't know ever

00:06:58.759 --> 00:07:03.050
increasing demand of power, you know. And so

00:07:03.050 --> 00:07:05.870
the fact that, like, China was able to find a

00:07:05.870 --> 00:07:10.850
way to do AI with less power is huge to me. I

00:07:10.850 --> 00:07:13.509
love it. I'm like, hell, yeah, let's go. And

00:07:13.509 --> 00:07:17.209
then, like, I read like 60 articles and it's

00:07:17.209 --> 00:07:20.110
like how horrible this is. But I'm like, I don't

00:07:20.110 --> 00:07:22.350
understand why it's bad. A lot of this, I think,

00:07:22.370 --> 00:07:24.870
is a lot of a lot of CNBC, Wall Street Journal

00:07:24.870 --> 00:07:29.470
trying to fluff up. Right. The the U .S. market.

00:07:30.050 --> 00:07:34.819
Yeah. I got to imagine that that Monday morning,

00:07:34.819 --> 00:07:38.060
you know, Sachin Adela and Mark Zucker, all these

00:07:38.060 --> 00:07:41.980
guys were like, crap. Right. And I got to imagine

00:07:41.980 --> 00:07:44.220
they had their engineering teams grabbing the

00:07:44.220 --> 00:07:47.519
source code Monday morning going, you better

00:07:47.519 --> 00:07:49.560
figure out if this is actually working, because

00:07:49.560 --> 00:07:53.459
if it is, we're in, you know, necessity is the

00:07:53.459 --> 00:07:55.160
mother of invention. And it looks like the Chinese

00:07:55.160 --> 00:07:56.959
were like, hey, look, you you expert control

00:07:56.959 --> 00:08:02.230
these chips. You you tried to screw us. But between

00:08:02.230 --> 00:08:04.689
and the thing that I drew in my tick tock when

00:08:04.689 --> 00:08:06.990
I made a tick tock about this the other day was.

00:08:08.170 --> 00:08:11.410
Between the tick tock ban immediately being reversed

00:08:11.410 --> 00:08:13.810
and everything that happened there and then this

00:08:13.810 --> 00:08:16.990
immediately happening, it seems like every time

00:08:16.990 --> 00:08:20.269
the U .S. is like, look at us, look at us. They're

00:08:20.269 --> 00:08:25.649
like the Chinese right now are like. Double birds,

00:08:25.810 --> 00:08:31.370
baby, we got you. And I don't think our leadership

00:08:31.370 --> 00:08:34.970
is currently prepared for what's happening right

00:08:34.970 --> 00:08:38.990
now. I don't think we have the political or social

00:08:38.990 --> 00:08:41.750
will to deal with this. Adam Schiff was like,

00:08:41.830 --> 00:08:43.909
this is our Sputnik moment. I was like, you know

00:08:43.909 --> 00:08:46.389
what? You might be right. But if it is, we're

00:08:46.389 --> 00:08:52.970
screwed. We are screwed. Because if this is our

00:08:52.970 --> 00:08:55.250
Sputnik moment, we don't have the political leadership

00:08:55.250 --> 00:08:59.360
and social leadership to meet the moment. Right.

00:08:59.700 --> 00:09:02.100
You don't have a JFK in there going, we're going

00:09:02.100 --> 00:09:04.139
to go to the moon by the end of the decade. You

00:09:04.139 --> 00:09:12.179
just don't have it. So from a political point

00:09:12.179 --> 00:09:15.059
of view, I do agree with you 100 percent. I do

00:09:15.059 --> 00:09:21.940
agree with you. From an engineering point of

00:09:21.940 --> 00:09:26.659
view. I would say like, yeah, hopefully it is

00:09:26.659 --> 00:09:29.980
our spunk spunk Nick. Right. Like, do you think

00:09:29.980 --> 00:09:34.940
we have? Actually, let me rephrase this a different

00:09:34.940 --> 00:09:38.179
way. Let's presume for a moment that our engineering

00:09:38.179 --> 00:09:40.980
talent is on par or maybe slightly better than

00:09:40.980 --> 00:09:45.779
that of mainland China. Yeah. Do we have the

00:09:45.779 --> 00:09:49.419
situation from a political and social standpoint?

00:09:49.539 --> 00:09:51.460
I talk about the fact that, you know, we're in

00:09:51.460 --> 00:09:54.480
a capitalist society and that changes how you

00:09:54.480 --> 00:09:58.990
run business. Do we have the ability to put that

00:09:58.990 --> 00:10:02.769
engineering talent to work in an efficient manner

00:10:02.769 --> 00:10:05.549
where we can catch up if we are as far behind

00:10:05.549 --> 00:10:12.009
as we may be. This is this is a little tricky.

00:10:14.409 --> 00:10:17.289
Because honestly, like the the engineering side

00:10:17.289 --> 00:10:23.870
of me wants to be like. essentially like fuck

00:10:23.870 --> 00:10:27.450
our political leadership. It has nothing to do

00:10:27.450 --> 00:10:30.509
with it because we're still going to pay these

00:10:30.509 --> 00:10:33.169
people. We got to pay the engineers. You got

00:10:33.169 --> 00:10:38.529
to get them to do the work. But however, I will

00:10:38.529 --> 00:10:48.450
say, though, engineers. I'm trying to figure

00:10:48.450 --> 00:10:52.649
out how to phrase this. I, so I used to be asked

00:10:52.649 --> 00:10:56.470
like, Hey, is engineering hard? And I would say,

00:10:56.830 --> 00:10:59.870
no, engineering is not hard. Anyone can be an

00:10:59.870 --> 00:11:02.149
engineer like Alex. You could be an engineer.

00:11:02.389 --> 00:11:05.889
Uh, your wife could be an engineer. Um, anyone

00:11:05.889 --> 00:11:08.570
can be an engineer. The thing about engineers

00:11:08.570 --> 00:11:14.409
though, is their dedication. So you have to put

00:11:14.409 --> 00:11:19.129
everything aside. And you have to find a solution

00:11:19.129 --> 00:11:25.610
period point blank. So There were there were

00:11:25.610 --> 00:11:28.190
days where you essentially you have to say there's

00:11:28.190 --> 00:11:34.330
a problem problem a whatever I'm not going to

00:11:34.330 --> 00:11:38.970
go to sleep tonight until I solve it period and

00:11:38.970 --> 00:11:41.409
With that you have to tell your friends like

00:11:41.409 --> 00:11:43.850
hey, I can't I can't talk right now. I can't

00:11:43.850 --> 00:11:49.220
hang out. I have to figure this out There were

00:11:49.220 --> 00:11:53.659
many nights which. And again, you have to you

00:11:53.659 --> 00:11:58.340
kind of make a trade off of like. You know, I

00:11:58.340 --> 00:12:00.960
will do whatever it takes to solve this problem.

00:12:01.379 --> 00:12:07.059
I. There were many nights where I had no sleep

00:12:07.059 --> 00:12:09.559
and I stayed awake from chocolate chip cookies

00:12:09.559 --> 00:12:14.000
and monster energy drinks. Keeping myself awake.

00:12:14.440 --> 00:12:18.460
I went like two days with no sleep, literally

00:12:18.460 --> 00:12:20.720
just trying to figure stuff out. And I think

00:12:20.720 --> 00:12:25.039
like that, that's engineering is not hard, but

00:12:25.039 --> 00:12:28.279
most people do not have the dedication to do

00:12:28.279 --> 00:12:30.519
that. Most people want to go to college to party.

00:12:30.700 --> 00:12:38.299
Most people will give up after a while. And honestly,

00:12:38.440 --> 00:12:41.740
like my, my father like passed away while I was

00:12:41.740 --> 00:12:45.320
in college. I took one day. Away from college

00:12:45.320 --> 00:12:48.200
till I deal with that and then was like back

00:12:48.200 --> 00:12:51.279
to college like you have to have that mindset

00:12:51.279 --> 00:12:54.960
of like no matter what comes up. I do not care.

00:12:55.000 --> 00:12:59.000
I do not care if the world's on fire. I don't

00:12:59.000 --> 00:13:02.419
care if Elon Musk deploys a Death Star into outer

00:13:02.419 --> 00:13:05.659
space. I am going to sit here. I'm going to figure

00:13:05.659 --> 00:13:08.919
it out. I'm going to find a way and kind of tying

00:13:08.919 --> 00:13:14.409
this back to your question Alex. That kind of

00:13:14.409 --> 00:13:17.990
mindset and there's people that take it way more

00:13:17.990 --> 00:13:20.690
like to the excrete to the extreme than I do

00:13:20.690 --> 00:13:23.970
I don't consider myself like some sort of genius

00:13:23.970 --> 00:13:27.789
You know person there's there's way like smarter

00:13:27.789 --> 00:13:30.529
people than me. There's way more dedicated people

00:13:30.529 --> 00:13:32.649
than me They would probably laugh and be like,

00:13:32.669 --> 00:13:35.750
oh you went without sleep for two days That's

00:13:35.750 --> 00:13:40.779
funny The thing about engineering is that you

00:13:40.779 --> 00:13:45.120
will find people that will just They will stay

00:13:45.120 --> 00:13:47.379
up and they will solve what like other people

00:13:47.379 --> 00:13:50.100
will say is like a pointless problem just to

00:13:50.100 --> 00:13:55.179
do it. So like Yes, it does suck it a hundred

00:13:55.179 --> 00:13:56.899
percent. I agree with you It does suck that we

00:13:56.899 --> 00:13:59.799
do not have the political leadership to back

00:13:59.799 --> 00:14:04.539
us on this. However Engineering is going to proceed.

00:14:04.539 --> 00:14:09.080
It's going to push forward You're going to have

00:14:09.080 --> 00:14:12.299
companies and also just random engineers that

00:14:12.299 --> 00:14:15.059
are going to just take on the task themselves.

00:14:15.059 --> 00:14:17.320
They're not going to wait for the right moment

00:14:17.320 --> 00:14:22.779
to kind of present itself. I hope you're right.

00:14:23.860 --> 00:14:28.720
I have very little faith in us right now. That's

00:14:28.720 --> 00:14:33.299
my thing. I have so little faith in us. I just

00:14:33.299 --> 00:14:38.799
do. It sucks. But that's that's kind of where

00:14:38.799 --> 00:14:43.539
we're at. Anyway, what do you all think? Is this

00:14:43.539 --> 00:14:45.799
a big sea change? Is it the sea change we think

00:14:45.799 --> 00:14:48.860
it's going to be? Does this change the whole

00:14:48.860 --> 00:14:51.820
thing? What's going to be the next big troll

00:14:51.820 --> 00:14:53.820
coming out of China? Because I think that this

00:14:53.820 --> 00:14:57.600
is just their opening salvo. I think TikTok and

00:14:57.600 --> 00:15:03.000
deep sea is just the this. Just the surface of

00:15:03.000 --> 00:15:05.259
what they're about to hit us with. So let us

00:15:05.259 --> 00:15:07.779
know what you think. down in the comments below.