naasking

15 ms·

301.
by naasking 8mo ago
I think people being murdered is pretty good intimidation frankly.
302.
by naasking 8mo ago
RSS directories perhaps, instead of relying on search engines.
303.
by naasking 8mo ago
Boot sector lisp? https://github.com/jart/sectorlisp
304.
by naasking 8mo ago
They haven't said a peep since CES over a month ago where they very quietly talked about their battery, no big fan fare. They seemed to be keeping a pretty low profile, it was all the YouTubers that started a huge hype train. That drov
305.
by naasking 8mo ago
> I'm not sure being hesitant to accept claims without evidence is a mark against credibility. Some skeptics have gone much further than that, calling the Donut battery impossible. The fact that they're taking their time addres
306.
by naasking 8mo ago
> Do you separately now create a state flag for that one render object? That can be a reasonable choice sometimes. Note that the point is that you introduce state where necessary , rather than stateful UI being the default as with retai
307.
by naasking 8mo ago
If they did actually achieve something remarkable, giving the skeptics time to bury their own credibility has some advantages.
308.
by naasking 8mo ago
> Is there any reason to not just switch to 1-based indexing if we could? Seems like 0-based indexing really exacerbates off-by-one errors without much benefit You'd just get a different set of off-by-one errors with 1-based indexin
309.
by naasking 8mo ago
I was thining of something like LLaDa that uses a Transformer to predict forward masked tokens: https://arxiv.org/abs/2502.09992
310.
by naasking 8mo ago
IIRC, some researchers are working on mixed AR+diffusion models for this sort of thing.
311.
by naasking 8mo ago
> Not only has USAID's destruction permanently destroyed US reputation in many place and will be responsible for the deaths of millions, including children, but many US farmers were USAID farmers. 100% of their crop and all of their
312.
by naasking 8mo ago
Yes, some kind of spec is always needed, and if the human programmer only has the spec in their head, then that's going to be a problem, but it's a problem for teams of humans as well.
313.
by naasking 8mo ago
> since the AI is writing the tests it's obvious that they're going to pass That's not obvious at all if the AI writing the tests is different than the AI writing the code being tested. Put into an adversarial and critical
314.
by naasking 8mo ago
Publishing something publicly means readers can do anything they want with the knowledge they gained from it. They just can't redistribute it verbatim due to copyright.
315.
by naasking 8mo ago
Nobody gets to have unbounded information about others. It's weird that you think there should be no privacy constraints.
316.
by naasking 8mo ago
Historical record, as one example. We gain considerable value from official records from the past, why would our descendents be any different?
317.
by naasking 8mo ago
Some pretty compelling evidence is history: we had dynamic and interactive web pages 20 years ago that were faster on computers that were an order of magnitude slower.
318.
by naasking 8mo ago
> Volunteering does not create a duty. Volunteering to maintain a project literally does create entail accepting duties, that's what taking on the role of maintainer entails. They are of course free to give up that role at any time,
319.
by naasking 8mo ago
Hence why I said "voluntarily".
320.
by naasking 8mo ago
If they're doing this and bothering to interact with tickets at all, presumably they've willingly taken on a duty to the software's quality and all that that entails.
321.
by naasking 8mo ago
Google is still behind the largest models I'd say, in real world utility. Gemini 3 Pro still has many issues.
322.
by naasking 8mo ago
That doesn't make sense with subscriptions.
323.
by naasking 8mo ago
> Once again LLM defenders fall back on "lots of AI" as a success metric. That's not implied by anything he said. He simply said that it was fascinating, and he's right.
324.
by naasking 8mo ago
> So you think that enough of the complexity of the universe we live in is faithfully represented in the products of language and culture? Math is language, and we've modelled a lot of the universe with math. I think there's st
325.
by naasking 8mo ago
I agree LLMs are inefficient, but I don't think they are as inefficient as you imply. Human brains use a lot less power sure, but they're also a lot slower and worse at parallelism. An LLM can write an essay in a few minutes that
326.
by naasking 8mo ago
I think it's correct to say that LLM have word models, and given words are correlated with the world, they also have degenerate world models, just with lots of inconsistencies and holes. Tokenization issues aside, LLMs will likely also
327.
by naasking 8mo ago
Intelligence must be sigmoid of course, but it may not saturate until well past human intelligence.
328.
by naasking 8mo ago
Sure, instruction tuned models implicitly plan, but they can easily lose the plot on long contexts. If you're going to have an agent running continuously and accumulating memory (parsing results from tool use, web fetches, previous his
329.
by naasking 8mo ago
Unless I'm misunderstanding what they are, planners seem kind of important.
330.
by naasking 8mo ago
But language does have structure, as does logic and reasoning. Universal approximation is great when you don't know the structure and want to brute force search to find an approximate solution. That's not optimal by any stretch of

More ›