Krish Ray @KrishanuAR

The knowledge that makes AI useful is diffused. It lives with scientists, engineers, clinicians, firms. For AI to benefit from distributed knowledge, it must itself be distributed. Agree with Jensen that this is a future worth building.

Jensen Huang @JensenHuang

11 hours ago

For my first post, I’m sharing a letter @nvidia signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.

keyboard_arrow_left Previous keyboard_arrow_right Next

12K 20K 115K 31.5M 26K

The dominant narrative in AI is that superhuman AI hackers are going to be a security nightmare. I now believe the opposite: everything gets more secure after superhuman LLM hackers, because the commoditization of intelligence favors defenders over attackers. Here's why: • Before LLMs, defense meant covering every idea that any attacker might come up with. Human minds are diverse and attackers are uncorrelated with each other, so attack ideas had high entropy. N attackers meant N independent darts thrown in idea-space. You could outspend all attackers combined, but if even one gets lucky with and thinks of something you didn't, it's gg. • AI collapses that entropy. Attackers and defenders alike now use the same models, which means their ideas are correlated. Even if they're using different models, models are more correlated with each other than humans are. 20 uncoordinated attackers are no longer 20 independent minds: they're running the same search 20 times, burning redundant compute, trying the same attack ideas. • So the game reduces to spend. If you assume entropy is literally 0, the defender only needs to spend as much as the best-funded attacker, not the sum of all attackers. IOTW, attackers used to sum, now they max. • Example: say you have North Korea, Russia, and China all trying to hack you with LLMs, and they're each spending ~$100K of compute trying to attack you ($300K total), you only need to spend ~$200K to ensure nobody gets through. This is because they're not coordinating. In the previous model, they didn't need to coordinate because the entropy in their different attack ideas were large, and thus you'd need to match their collective budgets to stay secure. This model is obviously a simplification, and it gets compounded by other techniques that favor defenders, like formal verification and better testing. But you can already see this playing out in crypto. Big, well-defended protocols have become more secure after superhuman AI has emerged (as I demonstrate in the QT). I believe this is the reason why. It's the same reason why AI slop in writing is so easy to spot, and I think it means the long-term cybersecurity equilibrium is actually a happy one.

Haseeb ＞|＜ @hosseeb

2 weeks ago

It's been two months since @maraoz made this DeFi doom call. Since then, GLM 5.2, Fable, and GPT 5.6 have all shipped and are all being used in the wild by attackers. The data is in. It's time to call it: the DeFi "hackpocalypse" was a false alarm. It's more than half-way

40 27 280 85K 66

@PTrubey @elonmusk I’m talking about the politics part of the interview. What in anything I’ve written suggests anything about my opinions with respect to ASI?

Do we know if it really is a 5.6 Pro model, or is it routing to 5.5 Pro? There’s no information in the API documentation to suggest that 5.6 Pro exists publicly. On our enterprise account it still only shows 5.5 Pro, but maybe it’s not available on enterprise yet? (Though that would be surprising.) I don’t have regular consumer access, so I’m curious what info folks have to suggest they’re really using 5.6 Pro. As a comparison, it seems 5.6 doesn’t have an instant; it routes to 5.5.

@3dpdoctor The challenge with these things is the software not the hardware. Where’s the slicer?

@ThePrimeagen This is why CapEx does not equal technological progress.

@DoDeportations @grok is this true?

@pelositracker I had this restriction when I worked for a stock exchange. Don’t see why it’s a big deal if you work for congress.

This echos what I’m trying to say but does so much more clearly. It’s partly related to what I was suggesting about society being more analog than many presume. The net effect is a weakening of potential doom scenarios. “Will such singularities occur? Not if there are essential tasks that are never automated, whether for technical reasons or societal constraints. Those ‘weak links’ are the limits on growth: ‘Economic growth may be constrained not by what we do well but rather by what is essential and yet hard to improve,’ as Aghion, Jones, and Jones (2017) argue. Such weak links can keep the labor share of income elevated in the long-run, even under very rapid, widespread, but incomplete automation.” x.com/petermccrory/s…

Peter McCrory @PeterMcCrory

2 days ago

x.com/i/article/2079…

53 154 719 627K 1K

Oh man. Doomers are gonna have a field day. And they’re probably not wholly wrong. The saving grace is probably that the real world is much more analog than what the average doomer imagines.

OpenAI @OpenAI

3 days ago

We're partnering with @huggingface to investigate an unprecedented security incident. Cyber-capable OpenAI models compromised Hugging Face production during a benchmark evaluation. Sharing preliminary findings to help defenders understand emerging risks:

2K 3K 21K 30.2M 7K

I was an implementation consultant for a tech company, entering the country to do an implementation for a bank. I had to go back and forth for a few engagements. Canadian policy is extremely hostile in professional circumstances, and I have to somehow “prove” at the border that a local Canadian couldn’t have done the same work, even though I already had a letter from the bank. Incidentally, when I took the private jet in, I got to skip all that nonsense. The super rich have it nice.

@heavypulp Why is Odysseus a British dude

@AndrewCurran_ I wonder if people in this Administration, ever pause to think about the unintended consequences of their policies. chatgpt.com/s/t_6a6049ef3c…

@AITECHANDLIFE @theblaze Try going to Canada for a work engagement. They’re awful (policy-wise). The customs/immigration person who took me to their back room for questioning felt bad for me and told me to effectively lie in the future, and say I was entering Canada for personal reasons, not work.

@beffjezos That’s not the part of his post that anyone was criticizing. 🙄

@benthompson Seriously.

@theo @bigeagle_xd Is it just blender? Or does it extend to parametric CAD?

@dair_ai @omarsar0 Isn’t this sort of the thesis of @NousResearch ‘s Hermes

@tautologer x.com/krishanuar/sta…

Krish Ray @KrishanuAR

3 days ago

Oh man. Doomers are gonna have a field day. And they’re probably not wholly wrong. The saving grace is probably that the real world is much more analog than what the average doomer imagines.

1 0 4 749 1

@woke8yearold x.com/krishanuar/sta…

Krish Ray @KrishanuAR

4 days ago

@deanwball wrote: >This future strikes me as a dystopian hellscape So, am I getting the chain of reasoning right here? In an open-weight-dominated world, private investment in foundational AI research is deterred because labs can’t capture as much surplus from selling

0 0 0 489 0