It's the first time in the history of the world.
Ilya Sutskever has brought exciting news!
Safe Superintelligence (SSI), which he founded, is set to release its first model this month.
Gavin Baker, a billions-dollar public-market investor who is recognized as one of the "most influential" top investors, has been recognized as Nvidia's "First Big Institutional Investor".
He just made it clear that the SSI is releasing the model.
If the news is true, the AI circle will definitely welcome heavyweight players.
When he founded SSI, Ilya Sutskever expressed his ambition:
The company's special feature is its first product, a secure super-intelligence.
Until then, it will do nothing else.
If Ilya can live up to the promise of "direct communication superintelligence" and what the SSI has to offer, the expectations of netizens have been filled.
AI-related stocks rebounded sharply last month, with Baker describing it as "compressing the bear market into a month in 2022".
When he released the SSI message, the U.S. AI market was down about 35-40% from its peak in June.
Despite this, Silicon Valley is still in high spirits and accelerating.
In the conversation, Baker repeatedly stressed that his task this week was "pressure testing" to try hard to find any negative quantitative data, but almost nothing.
The theme naturally extends to "what happens that can really break the current stock market's optimism", and Baker moves to the technical level: continuous learning and sample-efficient learning.
Many current laboratories feel that they are very close to solving these two problems.
He used specific numbers to illustrate the effect: early models trained only about 20 billion tokens, whereas today's models train up to 30 trillion tokens.
If in the future "only pre-train about 10 trillion tokens and then put the model into the real world to learn efficiently on its own", the demand for large-scale training will decrease dramatically, causing a steep drop in demand for British Wi-Fi cards.
In fact, the proportion of training computing power in the overall semiconductor demand will be close to zero (or become extremely small).
After explaining this logic, Baker explains:
"This is the most interesting direction at the moment, and no one knows if it's a long-term problem or a short-term breakthrough.SSI says they will launch the model in August.Now a whole new generation of labs is focusing on this".
Gavin Baker is the long-time chief investment officer of Atreides.
Since joining Fidelity in 1999, he has lived through the Internet bubble, mobile Internet, cloud computing, social networking and generative AI.
He anticipated the value of companies like Nvidia, SpaceX, Meta, Uber, Tesla and Cerebras.
It's really worth noting that since 2000, he's been following Nvidia and the entire semiconductor ecosystem for more than 20 years. From gaming GPUs, CUDA, cryptocurrencies, to deep learning training, reasoning, and AI factories, he's basically gone through several transformations of Nvidia's business model in its entirety.
In a GTC campaign in 2026, Huang said he was "Nvidia's first big institutional investor".
Investing podcast host Patrick O'Neill Shaughnessy thus rated his understanding of the world of technology almost "encyclopedic".
He has a really scarce resource: blood.
As early as 2017, the Wall Street Journal reported that Baker was able to enter startups like Uber, Snap, and others years in advance because of "access", meaning that the company's founders and financiers were willing to get him into the franchise.
Although the news has not yet received an official response, he said SSI could release a new model in August, which is worth listening to.
Cranquet on Wall Street: When Algorithms Monopolise ConsensusIn addition to the sharp clash between technology and capital, Gavin Baker makes a deeply sociological observation:Claude has become the "Walter Cronkite" on Wall Street.
Walter Cronkite was one of the most famous American legendary news anchors of the last century and has been called "the most trusted man in America".
When he proclaimed something to the whole of America, it became the only truth in the whole of America.
Today, global capital markets are falling into an unprecedented state."Breakdown in Diversity"。
In the past, thousands of analysts across the U.S. have come up with thousands of different interpretations of complex semiconductor export-restricted news or misleading profits.
This view of "biodiversity" is the cornerstone of open market liquidity and price discovery.
Now, however, the first reaction to the news, whether from retailers, buyers or seller analysts, is to copy a large portion of the text into Claude and then wait for its probability output.
Claude is like a never-ending Walter Cranquet. Because of its powerful logical attribution, its conclusions are often extremely clever, rigorous, and difficult to argue.
The result is:Almost everyone is trading according to the same logic of the same AI.
The high homogenization of this consensus has led to the extreme and cyclical contraction of asset price fluctuations.
In this highly formatted marketplace driven by unified algorithms, Gavin Baker was able to reveal publicly on August 4 that "SSI is launching a model this month" as a spectacular rebellion against the "Clankett mentality".
Prior to this, SSI had a long-term partnership with Intel to gain access to the Vera Rubin platform, improving computing power by 10 times, perhaps providing hardware support for model releases.