gio retweeted
She looks so happy to be with him, so proud of him and the life they made together, ready to enjoy their golden years after decades of hard work…. Meanwhile he looks completely annoyed with her existence. This whole story is so incredibly depressing.
gio retweeted
Mate selection theory of feminization undefeated.
With Jensen Huang in the news, it's reminder of how unnatural and opposed to human nature expectations of long-term monogamy are without a religious foundation. So we developed feminization instead. richardhanania.com/p/a-mate-…
gio retweeted
resources are allocated in the world according to supply and demand, and implicitly we have always understood that to be *human* demand. I think the best way to understand gradual disempowerment is the fear that resources will be allocated more and more according to machine demand rather than human. I’m not saying I think this outcome is my prediction, but this seems like the best way to understand the fear people have. It may well require deliberate effort to avoid this outcome.
gio retweeted
Being a black man with an IQ over 130 and forced to interact with median blacks has to be an absolute living hell.
BREAKING 🚨: Kmele Foster: “It's one thing to not think capitalism isn't working, ideally. It's another thing entirely to insist that it is oppressive, that it is bound up with white supremacy.”
Jemele Hill: “How is it not?”
Kmele Foster: “I've always preferred Adam Smith's description of capitalism as the natural system of liberty. The idea that people can own property, that they can exchange it freely with one another for profit, that they can find ways to put their resources to good use and profit in a way that is actually beneficial to everyone else, that there can be gains from trade that are mutually beneficial. These are foundational ideas that have created more wealth than any other system in human history. That is what capitalism is. That is not white supremacy.”
Jemele Hill: “Do you think capitalism has always worked that way?”
Kmele Foster: “I’ve actually just described the basic facts of the economics.”
Hill: “That ain't how it's always worked.”
Kmele Foster: “Slavery is an institution that is older than writing. It predates capitalism. The notion that white supremacy and all these other things are bound up with capitalism in an intrinsic and fundamental way is so absurd and laughable, it barely needs to be defended.”
Jemele Hill: “How was it absurd?”
Kmele Foster: “It barely needs to be defended.”
Jemele Hill: “What do you mean it barely needs to be defended?”
Kmele Foster: “You actually have to articulate what the connections is. How are these two things the same? I'm talking about private property, the right to exchange and sell with whomever you like. What does that have to do with white supremacy?… What is the white supremacy dynamic of private property?”
damn jensen
I have a major result to share.
I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China.
This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model.
Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves.
I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems.
I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used.
It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20.
These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs.
I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at unemploymentlabs.com/checkpo….
There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.
1
72
gio retweeted
really bad. all major Chinese models are being trained on Blackwell class chips
I have a major result to share.
I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China.
This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model.
Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves.
I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems.
I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used.
It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20.
These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs.
I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at unemploymentlabs.com/checkpo….
There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.
gio retweeted
the black leather jacket used to be a signal that this is a wild guy. It used to mean something. Recently, it was worn by any type of guy.
I'm glad Jensen brought back the respect toward it.
