Sam Altman is the WRONG man for the job

Now that we've had time to reflect, let's discuss what we can actually learn from the DeepSeek shock. Is OpenAI toast? Is AI over? Are the rest of our lives going to be spent anticipating, navigating, and recovering from pointless tech bubbles?

In order: Maybe. No. Yes.

Most of the commentary on DeepSeek has focused on $NVDA the stock and the myriad silly things Sam Altman has said over the years. Not as many people are thinking about how it exposes an important cleavage between VC hype artists and actual innovators who understand the underlying technology, or why Sam Altman's blind spots could ultimately doom OpenAI from ever delivering on the promise of the future he's sold investors.

00:00 Why did DeepSeek cause Nvidia and OpenAI's valuation to drop?
04:31 How the CHIPS Act and Biden led to Deepseek's disruption
05:33 How Sam Altman's strategy exposes OpenAI to risk
07:57 The venture capital shell game
10:59 The underlying reason tech keeps creating market bubbles
18:51 The lesson we'll refuse to take from this

This video is an excerpt of the podcast Breaking Change:

[zoeherriot]

Yup. I worked in Silicon Valley - the studio went from 50 people to 550 people in a matter of a year. This was based on one project that was making 50 million dollars a year, but only required about 20 people to run. The rest of the studio were sitting around trying to create the next big thing - but... they were really there to help attract more investment money. Mind you - it was great having a chef come in and cook for us and an all you can eat snack bar that was a ten foot long and a foot deep with chocolate bars.

[amerramday]

Never underestimate the ingenuity of Chinese engineers. 😊

[chaz]

Jai Hind. Sam sniggers when told that we Indians also wanted to build our own Indian AI model. In any case, we Indians can never afford the insane training energy usage of these closed Hyperscalers.

[litesp]

Deepseek exposed these AI bro grifters.

[kevinz]

awesome video! great blend of tech, humor, and personality. Look forward to more!!

[beerund]

You got your mic set up the wrong way. The shure sm7b is a “front adres microphone and it’s tuned to pick up sound only at the top.

[ritteradam]

The value depends on 1 thing: if the implements are really exponential and reasoning can be really solved in 1-2 years (and it seems so), even a few months of advantage that comes from spending hundreds of billions of dollars is worth it to be able to solve harder problems with the same algorithms but higher scale.

[DanieleNiero]

And let's not forget the environment cost of this "game".

[nitharsanam]

குறள் 517:
இதனை இதனால் இவன்முடிக்கும் என்றாய்ந்து
அதனை அவன்கண் விடல்.
'This man, this work shall thus work out, ' let thoughtful king command; Then leave the matter wholly in his servant's hand

[johnphamlore]

Nvidia's Jensen Huang explicitly and openly stated exactly what the plan is: The next scale or several orders of scale are to take video from anything and use that to train models. I speculate one idea is to create profiles of how to accomplish tasks that can be downloaded into general purpose humanoid robots.