vessenes

5 ms·

61.
by vessenes 2mo ago
To be clear, I'm not claiming Citadel created double digit drops in SK Hynix / Samsung. I am saying that as market vol hits, vol traders might choose to make it worse. And when word hits the street someone has a liquidity position
62.
by vessenes 2mo ago
...Unless trench warfare your son was living through was on your mind as you wrote
63.
by vessenes 2mo ago
This is everywhere. For reference, former FTXer and OpenAIer raised $225m into a hedge fund structure, went long and short, and reportedly peaked at $40bn of value; leverage bit hard this week and they sold their entire-ish portfolio to Cit
64.
by vessenes 2mo ago
Yeah there’s a mountain of interesting interp work to be done, maybe lifetimes. Looking forward to seeing what you all put out next! In very early gpt-3 beta days, I did some work on whether or not ethical guidance out of GPT varied by lang
65.
by vessenes 2mo ago
You could publish this in the US easily. 145 IQ peeps get tons of stuff wrong, btw, and in many domains my experience is the ‘wrongness’ can intensify as you move up into higher sigma domains.
66.
by vessenes 2mo ago
This is interesting and important work, thank you! Question - has your interp group looked at any of Anthropic’s neuralese-to-words tech? I’d be curious to see thinking traces (as in actual weights thinking not the output thinking) from the
67.
by vessenes 3mo ago
It is a very bad thing. Your friends do not have a solid understanding of geopolitics and are putting many, many people at risk worldwide, possibly for a generation or longer. The reason this petition is a bad thing is if it is effective it
68.
by vessenes 3mo ago
I didn’t say that. I said the US is ahead, and should consider its strategic adversaries and their behavior.
69.
by vessenes 3mo ago
I don’t think it’s clear we’ll get commoditized AI out of this research loop. The only real datapoints we have are that Chinese open weights models seem able to stay a certain amount of distance ‘behind’ released frontier models. It’s clear
70.
by vessenes 3mo ago
… while they address absolutely none of the above points, except for tacitly or overtly agreeing a panopticon surveillance state is needed / desirable to get their desired ‘slow down’, a situation where the cure is infinitely worse tha
71.
by vessenes 3mo ago
Bioweapons are good for .. war. Agentic AI is good for changing your economy, strategic advantage, labor explosion and making more effective Agentic AI. They are not the same category of thing.
72.
by vessenes 3mo ago
This. Will. Never. Work. This research game is a zero-trust iterated game. Any defection carries the benefit of possibly being permanently valuable. There will always be defections. OpenAI defected WHILE THEY WERE pushing to slow. The cost
73.
by vessenes 3mo ago
Seriously. I asked for an elliptic curve math course to be generated last week and Fable instantly opted out and downgraded.
74.
by vessenes 3mo ago
I do not believe that chosen plaintext attacks are obsolete. Any time you have an encrypting oracle setup in which the scheme itself does not require some sort of modification of the plaintext itself, you can reach for a chosen plaintext at
75.
by vessenes 3mo ago
It’s worth thinking through threat assessment and security posture carefully at the design and production phase. First, I’ll say that it’s cool you’re working on security stuff! It’s a hard discipline, but interesting. Next I’ll say that th
76.
by vessenes 3mo ago
Phi was smart (or had a smart training architecture, more precisely), and arguably pushed the conversation forward globally on the value of curated training data.
77.
by vessenes 3mo ago
It's in the constitution, probably a lot. And in the RL training. You can tell Anthropic is really worried about it. I'm interested in just how dishonest the so-called 'helpful only' non-RL model they have floating aroun
78.
by vessenes 3mo ago
Yeah agreed. I have on my long-term project list a hardware 'oracle' that would have everything and a local good model as a librarian/assistant and be solar powered in a pinch.
79.
by vessenes 3mo ago
Are you referring to the META lawsuit? I think the current landscape is not bad for the frontier labs -- it's settled law that it's legal to 'read' and ingest this data. Anthropic went ahead and just settled a licensing
80.
by vessenes 3mo ago
Meh. The (a major?) goal of Rust is this sort of type safety flowing down to memory bugs. Not every possible access pattern was known at the beginning, and generally they developed a philosophy of keeping the "unsafe" memory acces
81.
by vessenes 3mo ago
I think some villainy is implied by "billionaire-backed-woodchipping of a library," but it's just my interpretation, no actual knowledge of Vinge's perspective.
82.
by vessenes 3mo ago
I think it's worth putting a Knuth quote up on your board when you think about verification as an end goal: "Beware of bugs in the above code; I have only proved it correct, not tried it." Or, as Wolfram proposes in another l
83.
by vessenes 3mo ago
I don't see any proof of shredding here. Most book scanners I'm aware of are from Google's scanning days, and those had cameras plus page turning. If we use 'shredding' to mean A book is laid flat, its cover is remo
84.
by vessenes 3mo ago
Perhaps the last great near-term predictor. I often wish he'd written more. To remind us all, he predicted shred and scan would be a short stop over done by villains on the way to nondestructive scanning. That said, supporting Anna&#x2
85.
by vessenes 3mo ago
Presumably lots of stuff wrapped in 'unsafe'. I'm not a rust guy or a rust fan, and the last time I wrote something with Rust was like 7 years ago, but to my memory, because of how Rust manages mutability there are many acces
86.
by vessenes 3mo ago
Really interesting. Upshot: a well trained multimodal video generation model has a world representation model trained inside it. They’ve done some work lifting this world model out and deploying it to robots, where it seems to work well. On
87.
by vessenes 3mo ago
How’s the agent swarm working? Do you have like a sample startup you run using it?
88.
by vessenes 3mo ago
Nerd baited. For a one acre dry-ish property, how many mosquitoes need to be killed on a given day in spring? This seems crazy to me.
89.
by vessenes 3mo ago
This quarter’s clankerisms: “Honest” and “vibes”. The tech here is cool; good reminder to push for more scripts in general from the agentic tools.
90.
by vessenes 3mo ago
I kind of nerd baited myself and have something building for remarkable right now; it's a fork of the riddle diary app for remarkable with spaced repetition and subject tracking. I'm curious how it will feel! drop me a line if you

More ›