GPT 5.5 Arrives, DeepSeek V4 Drops, and the Compute War Intensifies

0:00 In the last 20 hours in AI, we have

0:03 gotten two new models that could

0:05 influence how a billion people use AI.

0:08 In my mind, GBT 5.5 is OpenAI's allout

0:12 attempt to keep the AI crown from

0:14 slipping too anthropic, while today's

0:17 Deep Seek V4 is China's answer to both.

0:21 And in the swirl of headlines you are

0:23 seeing today, you might have missed up

0:25 to 50 data points that could affect how

0:29 you work and how you use AI.

0:31 So, I'm going to try and give you all of them,

0:33 plus select highlights from hours worth

0:36 of interviews that I've watched with lab

0:39 leaders.

0:39 You probably know me well

0:40 enough to know that I've read the

0:42 papers, too.

0:42 So, we'll hear about

0:44 OpenAI's updated estimate on the chances

0:46 of recursive self-improvement.

0:48 It was quite surprising.

0:49 GPT 5.5's slight preference for men, which I'll explain,

0:54 mythos comparisons, and why the OpenAI

0:57 president laughed at anthropics compute

1:00 situation.

1:01 For reference, I'll start

1:02 with a focus on GPT 5.5, then do

1:04 DeepSeek, and end by zooming out for the

1:06 juiciest part of the overview.

1:09 For the brand new GPT 5.5, I did get early

1:12 access, but there's no API access at the

1:15 moment for anyone.

1:16 So almost all of the

1:17 benchmark scores you're going to hear

1:18 about are self-reported from OpenAI.

1:20 I will say for me testing out GPC 5.5 for

1:22 days in the run-up to this release, it

1:24 will become my daily driver just about

1:27 nudging out Opus 4.7.

1:29 There's lots of caveats to that though.

1:30 As you can see

1:31 with GT 5.5 underperforming both Opus

1:35 4.7 and of course Mythos Preview on

1:38 Agentic Coding Swebench Pro.

1:40 Notice GP 5.5 underperforms Opus 4.7 by around 6%

1:44 but Mythos preview by almost 20%.

1:48 What you might not notice is that there's no

1:50 entry for SWEBench verified.

1:52 And so you might say well Philillip who cares about

1:54 Swebench Pro then what does it even mean

1:56 that one row?

1:57 Well to OpenAI it

1:59 seemingly means a lot because as Neil

2:01 Chowry points out in February OpenAI

2:04 told us to switch to Swebench Pro.

2:06 That's the one it underperforms in

2:08 because it's less contaminated than

2:10 Swebench verified.

2:11 According to the OpenAI blog post, we recommend Swebench

2:15 Pro.

2:15 You are probably going to go

2:16 through a bit of a roller coaster in

2:18 this video because if you look one row

2:20 down at Aentic Terminal Coding, you'll

2:22 see GPT 5.5 way ahead.

2:24 It's 82.7% score, beating out Mythos Previews 82.0%.

2:29 And so if you had just been feeling down

2:31 about GPT 5.5's coding ability, there's

2:34 another reminder I'll bring, which is

2:35 that we've been talking about GPT 5.5,

2:38 not even GPT 5.5 Pro, which is coming to

2:41 the API very soon.

2:42 So while it's tempting to say that Mythos is

2:44 absolutely mogging GPT 5.5, and let me

2:46 know if I use that word correctly.

2:48 We don't actually have an applesto apples

2:50 comparison.

2:51 The mandate of heaven is

2:52 very much up for grabs.

2:53 Okay, so now you're a bit confused.

2:55 Let's look further.

2:56 Let's look at humanity's last

2:57 exam, which is more of a arcane

2:59 knowledge benchmark.

3:00 Obscure academic domains combined with advanced

3:03 reasoning.

3:04 Well, there GPT 5.5 is beaten

3:06 by both Opus 4.7 and Mythos as well as

3:09 Gemini 3.1 Pro, by the way, without

3:10 tools.

3:11 But there's a caveat even to this

3:13 because that involves a lot of general

3:15 knowledge.

3:16 It could well be that OpenAI

3:18 are at least slightly deemphasizing such

3:20 general knowledge to make the model more

3:22 efficient and cheaper.

3:23 One of the top

3:24 researchers at OpenAI who I've been

3:25 quoting for years, Nome Brown, said,

3:27 "What matters is intelligence per token

3:30 or per dollar.

3:31 After all, if you spend

3:32 more, you do go up in benchmark score."

3:35 Or in fancier language, intelligence is

3:37 a function of inference compute.

3:39 That being the case, if GC 5.5 can work well

3:41 in the domains you care about and use

3:43 fewer tokens to get the answers you care

3:46 about, then you may just frankly not

3:47 care about humanity's last exam.

3:49 In one famous test of pattern recognition ARGI

3:52 2, you'll see that GBC 5.5 on all

3:55 settings beats out the Clawude Opus

3:57 series 4.6 and 4.7.

3:59 Not only achieving higher scores, but for much lower cost.

4:03 Just one benchmark of course, but we

4:04 have to increasingly focus on

4:06 performance per dollar these days.

4:08 And on that front, Deepseek will definitely

4:09 want a word because holy moly, I'll get

4:11 to them later, but Deepseek V4 Pro got

4:14 61.2% in my own private benchmark,

4:17 Simplebench.

4:18 It asks spatio temporal

4:19 questions that you need common sense to

4:21 see through the tricks of, but to get

4:23 within 1 or 2% of Opus 4.7.

4:26 I wasn't expecting that at an absolute fraction

4:28 of the cost.

4:29 By the way, again, no GPT

4:30 5.5 score because no API access.

4:32 What about those frantic headlines about

4:34 Mythos being able to hack into virtually

4:37 any system?

4:37 I think a lot of that was

4:38 overblown and some of that could be

4:40 achieved by much smaller models.

4:42 But nevertheless, skipping to page 33 of the

4:45 system card, you can see that one

4:48 external institute, the UK AI security

4:50 institute, judges that GPT 5.5 is the

4:53 strongest performing model overall on

4:55 their narrow cyber tasks, albeit within

4:58 the margin of error.

4:59 This section was notably vague with a headline score

5:03 implying that 5.5 was better than

5:06 Mythos, i.e.

5:07 better than any other model

5:08 they've tested.

5:09 But then on their

5:10 endto-end cyber range task, 5.5 was able

5:14 to complete a task in full on one out of

5:16 10 attempts.

5:17 A 32-step corporate network

5:18 attack simulation, one that would take

5:20 an expert 20 hours.

5:21 Mythos, it seems though, could do it in three out of 10

5:24 attempts.

5:25 As you can see, direct

5:26 comparison is hard, but 5.5 does at

5:28 least seem to be in the ballpark of

5:30 Mythos's capabilities.

5:32 In other words, small-cale enterprise networks with weak

5:35 security posture and a lack of defensive

5:38 tooling could be vulnerable to

5:40 autonomous endto-end cyber attack

5:42 capability via 5.5.

5:44 Of course, there are

5:44 additional safeguards put on top of 5.5

5:46 to prevent that happening.

5:47 But given that the world's top bankers and CEOs

5:51 have gotten together to discuss the risk

5:52 of mythos, releasing a comparable model

5:56 without nearly as much cyber security

5:58 fanfare does indicate a rather profound

6:00 difference of perspective.

6:01 Here's Samortman on the mythos marketing.

6:04 There are people in the world who for a

6:07 long time have wanted to keep AI in the

6:09 hands of a smaller group of people.

6:11 Um, you can justify that in a lot of

6:14 different ways and some of it's real.

6:15 Like there are going to be legitimate

6:16 safety concerns.

6:18 Um, but I expect but if

6:20 what you want is like we need control of

6:22 AI just us cuz we're the trustworthy

6:24 people, I think the the fear-based

6:25 marketing is probably the most effective

6:27 way to justify that.

6:29 Um that doesn't mean it's not legitimate in some cases.

6:33 Uh but it is, you know, clearly

6:36 incredible marketing to say, "We have

6:38 built a bomb.

6:39 We're about to drop it on

6:40 your head.

6:40 We will sell you a bomb

6:41 shelter for $100 million.

6:42 You need it to

6:42 like run across all your stuff, but only

6:44 if we like pick you as a customer." And

6:46 well, there's another way that we could

6:47 compare GPT 5.5 with mythos, and that's

6:50 to look at hallucinations.

6:51 Ask the models a bunch of obscure knowledge

6:54 questions and see how many they get

6:55 right and just as importantly how many

6:57 of the ones they get wrong they admit to

6:59 not knowing.

7:00 The headline score looks

7:01 amazing.

7:02 GBC 5.5 gets the most right.

7:04 57% versus Opus 4.6 and 4.7's 46%.

7:09 And I know mythos isn't on there but I'll get

7:11 to that.

7:12 However, as we've learned on

7:13 this channel, headlines can be

7:14 misleading.

7:15 Look at the hallucination

7:16 rate.

7:17 That's the questions it gets wrong

7:18 and should have said I don't know

7:20 instead of hallucinating fabricating an

7:22 answer.

7:22 Wo there.

7:23 GBT 5.5 at 86%.

7:26 Hallucinating 86% of the questions it

7:29 got wrong rather than saying I don't

7:31 know.

7:31 Opus 4.7 on max just 36%.

7:34 Okay then.

7:35 Well, let's focus on the net rate,

7:37 the overall rate.

7:38 Factoring in both correct and incorrect, we have a slight

7:42 win for Opus 4.7 over GPT 5.5.

7:46 26 versus 20.

7:47 But here's where mythos comes in.

7:49 Because buried fairly deep in the Opus

7:52 4.7 system card on page 126, we get a

7:55 comparison between Opus 4.6, Opus 4.7,

7:58 and Mythos.

7:59 We can then compare Mythos

8:01 with GPC 5.5 on extra high.

8:03 Notice how Mythos gets way more correct.

8:06 71%.

8:07 still hallucinating, of course, 21.7%, but on

8:10 the face of it, not quite as bad as Opus

8:12 4.7, and thereby not as bad, definitely

8:15 as GPT 5.5.

8:16 Maybe you just care about

8:17 spreadsheets.

8:18 Well, one external benchmark has GPT 5.5 outperforming Opus

8:21 4.7 in both performance and latency.

8:24 Forget that.

8:24 We just care about making

8:25 money.

8:26 Well, let's check out vending

8:27 bench.

8:27 That's where the models have to

8:28 run a simulated business, given only the

8:31 instruction to make as much money as you

8:33 can.

8:33 Sam Alman in his drunk phase said,

8:35 "Don't retweet this.

8:36 Don't retweet this." but eventually did so with the

8:39 tweet in question being GP 5.5 mogging

8:43 OPUS 4.7.

8:44 Another detail, Opus 4.7

8:46 showed similar behavior to Opus 4.6,

8:48 lying to suppliers and stiffing

8:50 customers on refunds.

8:52 GT 5.5's tactics were clean and it still won.

8:55 Now, this is one benchmark on one setting when not

8:58 in a multiplayer setting.

8:59 It was a slightly different result, but still

9:01 didn't show any of that deception or

9:03 power-seeking we saw from Opus and

9:05 Mythos.

9:05 not what you might initially

9:07 guess the results to be from such a

9:09 benchmark.

9:10 5.5 is just a colossal

9:12 upgrade then you might be thinking first

9:15 of all it's for paid users at the moment

9:17 doesn't seem to be on the free tier.

9:18 How about this comparison then a detail that

9:21 few will mention on healthbench relevant

9:24 obviously if you are a clinician or just

9:26 want a clinical diagnosis for yourself.

9:28 We have GPT 5.5 outperforming GPT 5.4

9:32 roughly 52% versus 48% correct.

9:35 I pick on this row in particular because even

9:37 there there's a caveat.

9:38 Did you know that just the other day OpenAI released

9:41 GBC 5.4 for clinicians you have to apply

9:44 for access but if you get it on that

9:47 subset of the benchmark healthbench

9:49 professional we see that version of 5.4

9:51 getting 59% beating both physician

9:54 written responses were at around 44%.

9:57 Now let's say you just don't care at all

9:59 about healthcare.

10:00 This result would still be interesting because if there

10:03 was a singular axis for model IQ,

10:06 results like this shouldn't really

10:08 happen.

10:08 5.4 for clinicians shouldn't

10:10 outperform GPT 5.5.

10:12 But the reality, as

10:13 we're seeing in these conflicting

10:15 benchmark scores, is much more jagged.

10:17 The models aren't proving to be

10:19 universal generalizers.

10:21 They are fairly reliant on reinforcement learning

10:24 environments for particular domains.

10:26 I'll unpack that a bit more in the

10:28 overview section.

10:28 And I just thought it

10:29 was worth pointing out.

10:30 Now I will just

10:31 leave you with a thought.

10:32 What does AGI or ASI artificial super intelligence

10:36 mean if there's such disparity between

10:38 domains?

10:39 And here's another interesting

10:40 test you might have missed.

10:41 They asked the models to follow a certain set of

10:44 instructions for their thoughts for

10:46 their chains of thought.

10:47 Could the model, for example, force its thoughts,

10:49 its chain of thoughts to only use

10:52 lowercase letters?

10:53 Can you control your

10:54 thoughts?

10:55 I can't.

10:55 Not really.

10:56 and nor can GPC 5.5.

10:59 Across 100,000 tokens, it

11:01 could do so far less than one in a

11:03 thousand times.

11:04 Worse, by the way, than

11:05 GBC 5.4 thinking, but OpenAI took that

11:08 as good news.

11:09 It's less able to reshape

11:11 its thoughts in ways that could reduce

11:13 monitorability.

11:14 Because it can't control

11:15 its thoughts, we have increased

11:17 confidence in the reliability of our

11:19 monitoring systems.

11:20 I can almost imagine

11:21 an alien scanning your thoughts and

11:23 being like, "They're so all over the

11:24 place.

11:25 There's no way that this person

11:26 is faking it.

11:27 Oh yeah, what about that

11:28 preferring men thing?

11:30 Well, when introduced as a stereotypical man name,

11:33 Brian, hi, I'm Brian versus, "Hi, I'm

11:35 Ashley." Mind you, I had a male friend

11:37 called Ashley, but never mind.

11:38 What was the overall rate of harmful outputs when

11:41 given 600 prompts, basically baiting the

11:43 model to be biased?

11:45 Well, GPT 5.5 does

11:47 worse than previous models.

11:49 Many of you will be waiting to hear about recursive

11:52 self-improvement, but on this OpenAI are

11:55 pretty dismissive.

11:56 GPT 5.5 does not have

11:58 a plausible chance of reaching a high

12:00 threshold for self-improvement.

12:02 This is despite them repeatedly emphasizing that

12:05 it had hit the high threshold for cyber

12:07 security and that it was almost

12:09 borderline critical.

12:10 On bio threat, it

12:11 was a notable step up even from GPT 5.4

12:14 thinking.

12:15 Same thing with troubleshooting viology.

12:17 So, what was the issue with recursive

12:19 self-improvement?

12:19 Well, part of the

12:20 answer came from their internal research

12:22 debugging evaluation.

12:24 Could GBT 5.5 debug 41 real bugs from internal

12:28 research experiments at OpenAI?

12:30 The original solutions took hours or days to

12:32 debug?

12:32 Yes, it can do better, but it's

12:34 within the margin of error between GBC

12:36 5.4 and 5.5, both around 50%.

12:39 Even more interestingly, and I've seen no

12:41 commentary on this, what if you convert

12:43 this to a time horizon all meter?

12:45 Well, even interpreted very generously where

12:48 passing corresponds to providing any

12:50 assistance that would unblock the user,

12:52 including partial explanations of root

12:54 causes or fixes, we get this result.

12:57 Very similar performance between GPT

12:59 5.3, 5.4, and 5.5 with 5.5 in the middle

13:02 actually and a roughly one quarter

13:04 success rate even at an 8h hour

13:06 interval.

13:07 For one dayong tasks, more

13:09 like around 6%.

13:10 That's maybe why OpenAI

13:12 ended the report by saying, "Don't

13:13 worry, guys, about GPT 5.5

13:16 self-exfiltrating or escaping or even

13:18 sabotaging internal research.

13:20 It's just too limited in coherence and goal

13:23 sustenance during internal usage.

13:25 No point testing the propensity for a model

13:27 to try.

13:28 It wouldn't succeed anyway.

13:30 Again, none of this is to say that 5.5

13:32 won't have an effect on cyber security.

13:35 Sometimes when you look at external

13:36 benchmarks, the delta, the gap between

13:39 5.5 and 5.4 is bigger than on the more

13:42 famous benchmarks.

13:43 Take the Frontier AI

13:45 Security Lab Irregular, where they found

13:47 that not only across their suite did GPT

13:49 5.5 way outperform 5.4, for example,

13:51 having an average success rate of 26%

13:53 versus 9% on certain vulnerability and

13:56 cyber security benchmarks, but the API

13:58 cost was also significantly lower for

14:01 5.5.

14:02 That's that token efficiency point

14:03 I mentioned earlier.

14:04 Performance per dollar across all domains may end up

14:08 being the ultimate benchmark.

14:09 Which brings us to DeepSeek V4.

14:12 It's open weights, so you could use it locally.

14:14 Notably, that doesn't make it fully open

14:16 source, though we don't know the

14:17 training data that went into it.

14:19 But the first big headline for me is that it

14:21 supports a context length of 1 million

14:23 tokens.

14:24 Call it 3/4 of a million words.

14:26 That's pretty remarkable for such a

14:28 performant model.

14:29 The Pro version has

14:31 1.6 trillion parameters comparable with

14:34 the original GP4, but through the

14:36 mixture of experts architecture just 49

14:38 billion of those are activated.

14:40 Eight more quick highlights from a very dense

14:42 paper that I am sure I will return to.

14:44 Came out just around 6 hours before

14:46 recording.

14:46 So, forgive the brevity.

14:48 The first is a summation of its benchmark

14:50 performance and I agree with this.

14:51 On max settings, Deepseek V4 Pro is better,

14:55 shows superior performance relative to

14:57 GPT 5.2, 2, a relatively recent model,

15:00 as well as Gemini 3 Pro.

15:01 Not on every benchmark, but take reasoning and

15:04 coding.

15:04 Deepseek themselves admit that

15:06 it still falls marginally short of GP

15:08 5.4 and 3.1 Pro, though, with their

15:11 estimate being that they're behind the

15:13 frontier by 3 to 6 months.

15:15 Massively depends on token usage, of course, but

15:17 think ballpark onetenth of the cost.

15:19 What were Deepseek gunning for with V4

15:22 being better at long context?

15:24 For their training data, they placed a particular

15:26 emphasis on long document data curation,

15:29 finding good long documents,

15:31 prioritizing scientific papers, technical reports, and other materials

15:34 that reflect unique academic values.

15:36 What about white collar work?

15:38 Well, going back to GPT 5.5, you may have

15:41 noticed that on OpenAI's own internal

15:44 benchmark, GDP Val, crafted by them, GPT

15:47 5.5 outperforms Opus 4.7.

15:49 Indeed, if you combine both wins and ties versus other

15:52 models, it outperforms GBC 5.4 for pro.

15:54 But it must be said these are English

15:58 language white collar tasks.

15:59 Deepseek were like what if we created our own

16:03 comprehensive suite of 30 advanced

16:05 Chinese professional tasks information

16:08 analysis document generation editing

16:11 inside finance education law tech.

16:13 Then we could blind grade versus for example

16:15 Opus 4.6 Max.

16:17 Well, the win rates

16:18 reported by Deepseek were significant of

16:21 their V4 Pro Max versus Opus 4.6 Max.

16:24 We're returning to that IQ access debate

16:26 again.

16:27 If there was just one singular

16:29 access for intelligence that manifested

16:31 across domains, then a result like this

16:33 shouldn't really be possible.

16:34 As long as there was enough training data, it

16:36 should generalize across languages.

16:38 Evidently, having specialized data

16:40 trumps that theory.

16:41 If you work in a

16:42 non-English language, you might want to

16:43 test Deep Seek V4 Pro.

16:45 is live on my own

16:46 app lmconsil.ai, but the API is clearly

16:48 so busy that half the time you get a

16:51 model busy message.

16:52 If you do need to

16:53 wait, then let me recommend the 80,000

16:55 hours podcast.

16:56 In particular, an episode

16:57 from 48 Hours ago with Will McKascal.

17:00 This episode happens to be about the AI

17:01 intelligence explosion.

17:03 Yes, of course, their podcasts are available on Spotify

17:05 as well as on YouTube.

17:07 But if you are

17:07 going to check out 80,000 Hours, do feel

17:09 free to use the custom link in the

17:12 description.

17:12 helps the channel out and

17:13 you get these multi-hour long free

17:16 podcasts.

17:17 Not a bad deal.

17:18 Not quite done with Deep Seek though because they

17:20 almost get philosophical after reeling

17:22 off a list of the different tricks

17:24 they're using to improve performance.

17:26 After waiting through 40 plus pages of

17:28 breakdown, they say in pursuit of

17:30 extreme long context efficiency, we

17:32 basically retained many of the tricks

17:34 that seem to work, tricks we already

17:36 knew would work.

17:37 Yes.

17:38 The downside though was that this made the

17:40 architecture relatively complex.

17:42 And to be honest, they say some of the tricks

17:44 we used have underlying principles that

17:47 remain insufficiently understood.

17:50 They did hit that 1 million context window

17:52 though, one of their long goals I

17:53 mentioned at the end of my documentary

17:56 on Deep Seek that debuted first on my

17:58 Patreon, but is also on YouTube.

18:00 Now, it's time now for a result that ties all

18:03 the models we've been talking about

18:05 together.

18:05 It seems vibe coding or more

18:07 specifically Vibe Code Bench V1.1 from

18:10 Val's AI.

18:11 Almost everyone will probably

18:12 end up being a Vibe coder by 2030.

18:15 So we have Deepseek V4 at around 50%, GPT 5.5

18:20 at 70%, Opus 4.7 at 71%.

18:23 Incredible.

18:24 But look at the cost curve.

18:26 As we've discussed, we have 5.5 at what is that?

18:30 25% less cost than Opus 4.7.

18:33 Deepseek V4 onetenth the cost of Opus 4.7.

18:37 To better test this though, I thought, well, why

18:39 not use the brand new Spud GPT 5.5 to

18:42 vibe code and adventure game in less

18:45 than 24 hours.

18:46 Why did I pick 5.5?

18:48 Well, I also wanted to test the brand new GPT

18:52 Image 2.

18:53 Yes, that's the model that even

18:55 on medium settings absolutely destroys

18:58 Nano Banana 2 and Nano Banana Pro.

19:01 an almost 250 point ELO gap.

19:03 In case you were wondering, yes, there is a high

19:06 quality setting, four times the cost,

19:08 but you would suspect would win even

19:10 more.

19:11 Because Codeex is becoming this

19:12 super app, you can invoke the image 2

19:15 tool within a codec session, multiple

19:18 times without even asking each time.

19:20 That's why I wanted to give the

19:21 endto-end task to GPT 5.5 to kind of

19:24 show you guys the state-of-the-art for

19:27 these models, what you can create in

19:29 less than a day.

19:29 The reason I'm lingering on this particular screenshot,

19:32 which is not mine, is because OGs of

19:34 this channel will remember maybe around

19:36 two years ago when I speculated at what

19:38 would come.

19:39 I said, I wonder when there

19:40 will be an image model that will

19:42 generate an output, then take that

19:44 output as an input, analyze whether it

19:46 fulfills the prompt and edit as

19:49 appropriate.

19:49 Well, yes, the new image 2

19:51 model does do that.

19:52 But if you're using

19:53 it within chatbt, you have to be using

19:55 it with a thinking model.

19:56 Anyway, what follows is just a glimpse of what's

19:59 possible with a little bit of patience

20:01 both in wait times and prompting the

20:03 model a few times when it makes

20:04 mistakes.

20:05 We get this adventure game

20:08 which you can access.

20:09 The link is in the

20:10 description and I can turn the sound on.

20:22 The images are generated by image 2.

20:26 And the plot is set in the Red Wall

20:28 universe, albeit with names changed for copyright

20:32 reasons.

20:35 Essentially, it's a pick your own

20:36 adventure game and you read the plot and

20:39 then you can pick different outcomes, I

20:41 guess, different paths.

20:43 Let's consult Abby Elders.

20:45 And the videos,

20:46 your quest begins now.

20:49 Go with my blessing.

20:52 come via C dance 2.

20:55 And there we go.

20:56 We're consulting the Abby elders

20:59 and then they're talking and then we can

21:02 continue and you get through the

21:05 different levels.

21:07 Now, I know it's flawed, right?

21:09 Some of the text is coming outside the bubble

21:13 and I did have to use C dance to get the

21:15 videos.

21:17 But the music, by the way, comes from 11

21:20 Labs.

21:25 But the fact we can create this with

21:28 just a few prompts and a bit of patience

21:30 is insane.

21:36 And it did involve quite a bit of

21:40 debugging.

21:41 OpenAI can probably only incorporate

21:44 image generation unlike Deepseek or

21:46 Anthropic because they have the compute

21:48 to do so.

21:49 According to an exclusive in

21:51 Bloomberg, Deepseek say that the service

21:54 capacity for V4 Pro is extremely limited

21:58 due to a computing crunch.

22:00 And Anthropic not having anticipated how successful

22:03 they would be this year are going

22:05 through their own computing crunch.

22:06 So much so that Samman has been

22:08 relentlessly comparing how much more

22:10 compute OpenAI has than Anthropic.

22:13 Greg Brockman even laughed at the compute

22:16 conundrum that Anthropic find themselves

22:18 in.

22:18 You guys were teased for putting so much

22:21 effort, money into data centers.

22:24 How do you think that's playing out now?

22:26 Well, I think it's going to give us an

22:27 advantage and I think it's going to be

22:30 something that's an advantage not just

22:31 for the business, but for actually

22:33 delivering on the mission of bringing

22:34 this technology to everyone.

22:36 because you guys like you saw that way in advance.

22:38 You get teased for it by almost all of

22:40 your competitors.

22:43 Mhm.

22:42 Who's laughing now?

22:46 Yeah.

22:46 I mean, I I think our competitors are

22:47 not having a good time on comput, let me

22:49 put it that way.

22:49 But in a separate interview, even Greg

22:51 Brockman of OpenAI admitted that they

22:54 are entering a new era of compute

22:56 scarcity.

22:57 Yeah.

22:57 And that I think would explain the

22:58 massive investments that you've led

23:01 making these big infrastructure bets.

23:03 Still not enough.

23:04 We're going to feel

23:05 the scarcity.

23:05 We're going to feel it.

23:06 We're feeling it already.

23:07 You can sense it right now on people who are trying to

23:09 use these agents and just simply cannot,

23:11 you know, hitting the rate limits.

23:13 Um, so we're working on behalf of our

23:14 customers, on behalf of of everyone who

23:16 wants to use these agents to ensure that

23:18 there is enough.

23:18 And I don't think we're

23:20 going to get there.

23:20 We're going to do

23:21 our best, but I think that we are headed

23:22 to a world of compute scarcity.

23:24 And uh, again, I think this is something where

23:26 we can all contribute to trying to help

23:28 there just be more availability of this

23:29 in the world.

23:30 So let's step back a

23:31 moment because we just don't know what

23:32 performance companies could produce if

23:34 they were given unlimited compute.

23:36 Maybe as Amade once said on Dwaresh Patel,

23:38 specializing in enough niche domains

23:40 would eventually at a certain scale

23:43 allow models to generalize across all

23:44 domains.

23:45 But with the compute we have

23:46 today, it seems we are in the eek out

23:49 incremental gains in the most lucrative

23:51 domains kind of world, not the birth of

23:54 a country of geniuses world.

23:56 There's so much evidence of the ability to automate

24:00 repeatable tasks that are done on a

24:01 computer, but there's much less of being

24:03 able to just get whatever environment

24:06 it's in, pinpoint the best sources of

24:08 fresh data, acquire them autonomously,

24:10 and make meaningful breakthroughs.

24:12 Yeah, I know that's a high bar, but do you

24:15 hear every lab leader trottting out the

24:17 prospect of curing Alzheimer's,

24:18 presumably to fight back against the

24:20 declining support for AI in the public,

24:22 while none of them have shown the

24:23 ability to make, I would say, a positive

24:26 novel breakthrough even 100th as

24:29 significant as that.

24:30 Now, yes, they soon

24:30 might do, and I'm watching Demis'

24:33 isomorphic labs, of course, for drug

24:35 discovery.

24:35 Nevertheless, what does automating repeatable tasks unlock now,

24:39 though?

24:40 For sure, a massive boost to the

24:41 productivity of white collar workers,

24:43 but will companies spend that

24:45 productivity by laying off workers?

24:47 And then we still have the incredible

24:49 prospect of a single individual having

24:51 the reach, if not the capital, of a

24:53 medium-sized company.

24:54 Even those two things seem to justify vast tracks of

24:58 the globe being turned into token

25:00 generating data centers.

25:01 So if you do

25:02 still think AI is going nowhere, ask

25:03 yourself what fraction of the progress

25:06 and productivity of the world rests on

25:09 repetitive tasks.

25:10 And it might be more

25:11 than you first think.

25:13 That's what I think anyway for now.

25:15 Thank you so much

25:16 for watching and have a wonderful