lukeschlather

3 ms·

1.
by lukeschlather 10d ago
That's because you aren't actually rewinding. You're replaying the conversation you just had through the LLM and it's giving you a likely explanation for what it might have said. It is actually possible to rewind LLMs
2.
by lukeschlather 13d ago
I would expect a preference for lutefisk flavored shit, the problem is then you then deduce from that that people prefer lutefisk flavor over cinnamon, rather than that you should stop feeding people shit (and also stop feeding them lutefis
3.
by lukeschlather 13d ago
The idea of publishing null findings is very attractive, but the problem here is that they didn't record what the actual problem they're trying to solve is. It seems a bit like someone prompted Claude to try and make tool calls mo
4.
by lukeschlather 13d ago
I don't think the problem is a lack of previous research, the problem is it looks like the target metrics were selected by an LLM, and it's unclear what the LLM was told to optimize or if it was just told to try and make a better
5.
by lukeschlather 14d ago
I think the speed of improvement is somewhat overstated, and I think this is all cool, valuable fundamental research. Rather than slowing down I would like to talk about how we can make it easier for people to harden their systems using the
6.
by lukeschlather 14d ago
That doesn't really seem true. The HF hack happened with a model that had all the alignment safeguards disabled intentionally. I think there's a good case for that kind of research, but also, OpenAI could just not do that if every
7.
by lukeschlather 14d ago
OpenAI accidentally hacked HuggingFace roughly 2 months ago. I would bet that the danger of them accidentally hacking HuggingFace during a similar attack has increased since then, and will continue to increase for the next 6 months. I don&#
8.
by lukeschlather 14d ago
We really need a hybrid short-haul standard designed for short-range transmission. Something that uses low-frequency spectrum currently allocated to 5G but isn't locked to a single carrier and the protocol can be shared between WiFi an
9.
by lukeschlather 15d ago
Documentation is important. I would say Opus' propensity to write documentation that documents non-features is part of the problem being discussed. And the problem isn't just that it says what the software doesn't do, most of
10.
by lukeschlather 15d ago
I think you aren't recognizing the harm of collecting this information. I might be okay with this if companies were banned from using age for targeted advertising. However, the other harm is the recent IDScan hack which leaked 153M peo
11.
by lukeschlather 15d ago
I would assume an H100 selling for less than $30k is a scam. I see one for 11,329€, and one for $25k but most are over $30k and I don't know how to judge these random storefronts. You've bought an H100 for $20k that was in working
12.
by lukeschlather 15d ago
I would rather take this option than a single H100: > 96 GB Dual NVIDIA RTX PRO 5000 w/ 1000W PSU (non-refundable) +$19,169 Though I don't think there's anything stopping you from trying to stuff an H100 into this machine
13.
by lukeschlather 15d ago
Google is buying the output because they are increasing demand by half a nuclear power plant's worth of power. I actually think it's pretty reasonable to pass a law demanding they build an entire nuclear power plant here.
14.
by lukeschlather 16d ago
For the purposes I use world maps for, correct relative sizing is the paramount thing I'm looking for in a projection. I generally use it to give me a rough idea of distances. Boggs seems inferior, at a glance I have no idea how far it
15.
by lukeschlather 18d ago
LG's terms of service essentially state that they will record any voice commands and store them for 6 months. They explicitly say they may store them in Korea. https://us.lgappstv.com/main/terms Actually I think e
16.
by lukeschlather 22d ago
> Just to get this straight: are you the kind of person to give your 1st grader a calculator I think banning smartphone/computer use before 3rd grade in the classroom sounds like a good idea. Math is incredibly useful. I think the b
17.
by lukeschlather 22d ago
AI detectors do not work. There are passages in this that have some odd structures that don't feel human to me. I'm sure a human edited this and refined it with some prompting, it's not just rough output from an AI. But a lot
18.
by lukeschlather 23d ago
Putting IDs in mobile wallets is such a minefield without explicit bans on phone searches by law enforcement. Law Enforcement being able to rifle through all your email/bank/medical records because you had a traffic stop is just n
19.
by lukeschlather 23d ago
I'm not talking about future capabilities, I'm saying it will cease to be arguable within 5 years. We can't just bury our heads in the sand. Kids are going to need to engage with this directly, and it cannot wait until after
20.
by lukeschlather 23d ago
Frontier LLMs with Internet search are different beasts than what you're describing as an average LLM. In 5 years I don't think you're going to be able to get away from LLMs if you care about correctness, and we're alrea
21.
by lukeschlather 24d ago
This sounds like a swarm of agents with particular prompting that happens to guide the LLMs toward doing a lot of fuzzing, so it's not either/or; you're getting all the compute requirements of both.
22.
by lukeschlather 24d ago
I'd be interested in some legal/GDPR takes on PII handling in prompts. If someone enters PII into a prompt, and Mistral retains it for training, is it sufficient for them to say "don't enter PII into prompts?" Of co
23.
by lukeschlather 1mo ago
I'm strictly talking about developed countries where we have rule of law and it's trivial to house and feed people if you have free housing and free food. I'm not suggesting AI can solve warlords, I'm not suggesting anyo
24.
by lukeschlather 1mo ago
Ending poverty is a purely mechanical problem, with very straightforward solutions if you have robots that can do any mechanical task. You don't need to give anyone a job, you just need to give them food and shelter. (Ideally also medi
25.
by lukeschlather 1mo ago
If AI is really that powerful, shouldn't someone like Gates be able to end hunger and homelessness with a one-time capital expenditure? Somehow "give people jobs" is the primary problem he wants to solve, not ending poverty.
26.
by lukeschlather 1mo ago
My muscle memory usually presses ctrl-backspace when I mistype a word, which typ.ing actually handles better than most typing software, which is to say ctrl-backspace does what I expect, it deletes the word and I can retype it, which is wha
27.
by lukeschlather 1mo ago
The human isn't the one responsible for the traffic violation, Waymo is. Could you give an example where a Waymo committed a traffic violation and they weren't cited?
28.
by lukeschlather 1mo ago
What does it mean to be held responsible for the consequences? OpenAI helped remediate the damage done by the model and took steps to make sure it wouldn't happen again. In what way were they not responsible? Nobody said they were supe
29.
by lukeschlather 1mo ago
They could put everything in some kind of nonprofit book vault/archive that includes all the source pages as long as they didn't use or distribute it. They could even provide a mechanism for rights holders to recover the text for
30.
by lukeschlather 1mo ago
It's only been at most the past year where AI has been unambiguously helpful and not a hindrance. 3 years ago it gave the appearance of being helpful but it tended to be more of a hindrance.

More ›