"Essential to these methods is that they can find good approximations, but the search for them should be by our methods, not by us."
Producing Entropy
Joined May 2019
- Tweets34.7K
- Following894
- Followers928
- Likes57K
Sam retweeted
It was never a real issue. The paper claiming to demonstrate it had an unrealistic setup of indiscriminately training on just synthetic data, discarding available human data. Keeping that data in was all you had to do to prevent collapse.
Sam retweeted
imagine an open model with astra level cyber capabilities gets released to everyone to use/abiliterate.
how exactly does this make the world more secure? like, i genuinely don't understand. you do realize not everyone is just going to burn tokens and magically secure themselves, right?
i will die on this hill:
you need to build models even more capable than astra before open models catch up, and invest more in cyber programs that give the good guys access to these models first.
open models reaching astra level cyber capabilities before defenders are ready(whatever that means) would be dangerous for the world.
@_deven___ for picture credits
Sam retweeted
Deep expertise in any STEM field will increasingly become an infohazard over the next few years. Watching my friends in the field process the ETH result really cemented this. During this Weltbilderschütterung, be there for people who thought they built their houses upon stone.
Sam retweeted
I been saying this for a long time, but tate used to be very open about how to manipulate women into sex work … would just brag about trafficking openly.
All that stuff got deleted from his website when getting fame and eyeballs … now he acts like he’s been setup
Sam retweeted
My takeaway from things I see internally is that even if frontier LLMs "paused" here indefinitely, we'd continue to figure out ways to get substantially more value, and our token demand for frontier tokens would continue to increase.
Maybe needless to say, its still very very early for the world at large - there is still a lot of demand for frontier tokens to unlock out there beyond my glass-clad/ivory tower.
I guess on the other hand, while the demand for tokens will go up, I'm still really unsure how total spend will change. In the hypothetical world of "locked at this level of intelligence", I could imagine spend increasing but also serious margin compression. I guess it goes back to something I've seen @tszzl (and others) muse about - AI will generate a lot of value, but to who does that accumulate? You've got the foundry level, the chip designer level, the "lab" level, the consumer level... will be split in some ratio. You can count the revenue at every part of the chain, but the profit is not double countable. I don't know what I think about that. I find it hard to believe its all to the labs.
Sam retweeted
Congrats to @AlexTamkin et al; surgical research design and concerning (if intuitive) results showing just how readily experts adopt AI agents' interpretations of evidence (reasonable to expect this'll increase with AI capabilities). @ezraklein I hope your interviews on AI safety and societal impact scope this in
Sam retweeted
‼️ Anthropic and Google researchers found a flaw that lets one crafted packet hijack a server during network boot, before Secure Boot checks the OS.
Almost all servers on earth are affected.
CVE-2026-97631 sits in TianoCore EDK II, code in the UEFI firmware of most servers.
Readers added context they thought people might want to know
The vulnerability is limited to servers actively PXE network-booting with vulnerable EDK II firmware; most servers boot from local disks and are unaffected in normal use, with Intel noting limited impact on supported products.
github.com/tianocore/edk2…
arstechnica.com/security/2026/…
Sam retweeted
Some entertaining highlights from wheresyoured.at/wheres-the-m… 20 months ago:
- "Anthropic has projected (made up) that it will make at least $12 billion in revenue in 2027"
- "This company is not worth $60 billion. Anthropic has raised $14.7 billion to create an also-ran Large Language Model company"
- "Anthropic, much like OpenAI, has categorically failed to productize its Large Language Model"
- "Its CEO is also a sleazy conman who, like Sam Altman, continually promises that his company's AI systems will become powerful and autonomous in a way that they have never shown any possibility of becoming."
- "Any investor in Anthropic needs to seriously consider what it is they're investing in. Its products are both unpopular and commoditized, and it lost $5.6 billion last year! Stop dancing around this fact! Stop it!"
- "there’s a lot of demand to put AI in stuff and some demand to buy stuff with AI on it by enterprises buying software, but little evidence to suggest significant user adoption or usage, I’d argue because Large Language Models do not lend themselves to features that provide meaningful business returns."
- "When a company uses the term “agent,” they are intentionally trying to be deceitful, because the term “agent” means “autonomous AI that does stuff without you touching it.” The problem with this definition is that everybody has used it to refer to “a chatbot that can do some things while connected to a database,” which is otherwise known as a chatbot."
- "It isn’t working. The users aren’t there. The revenue isn’t there. The best time to stop this was two years ago, and the next best time is as soon as humanly possible. I have said that generative AI is a group delusion in the past, and I repeat that claim today."
- "we are approaching the limit of what the transformer-based architecture can do, if we haven’t already reached it. No amount of beating off about test-time compute and connecting Large Language Models to other Large Language Models is going to create a new use case for this technology"
- "generative AI is OpenAI. The consumer adoption of this software has completely failed, and appears to be going nowhere fast. ChatGPT is sustained entirely on deranged, specious hype drummed up by a media industry"
... and my favorite ...
- "I will keep writing this stuff until I’m proven wrong."
Sam retweeted
"University researchers built a satellite that tracks people from low Earth orbit and dispatches a 5 kg payload to within .1 meter of their position, traveling at Mach 25 to anywhere on Earth in under 5 minutes. They suggest it could be used to deliver greeting cards."
Purdue University's IDEAS Lab built an autonomous drone that tracks runners through dense forests completely without GPS or cloud networks.
By running 3D LiDAR mapping and obstacle-avoidance AI entirely on-chip, it can weave under tree canopies and predict target movement through blind spots in real time.
Massive implications for next-gen search & rescue and edge AI.
Sam retweeted
They're taunting us. They're stunting on us.
Sam retweeted
This is like a hundred CVEs for mathematicians
Announcing three major scientific discoveries, one each in mathematics (harmonic analysis), computer science (complexity theory), and physics (quantum field theory).
AI starting to help us improve our understanding of the universe! openai.com/index/building-sc…
