mrandish

30 ms·

421.
by mrandish 8mo ago
> We need laws for agents, specifically that their human-maintainers must be identifiable and are responsible. Under current law, an LLM's operator would already be found responsible for most harms caused by their agent, either dire
422.
by mrandish 8mo ago
When you say "the administration", it's worth noting you're describing actions by two different administrations. Both political parties have tried to silence dissenting views through soft censorship.
423.
by mrandish 8mo ago
I didn't test this but I suspect current SotA models would get variations within that specific class of question correct if they were forced to use their advanced/deep modes which invoke MoE (or similar) reasoning structures. I as
424.
by mrandish 8mo ago
> It's not like Ars' own processes somehow unearthed this error. It might be even worse (and more interesting) than that. I just posted a sister response outlining why I now suspect the fabrication may have actually been caus
425.
by mrandish 8mo ago
I was wondering the same thing. After I posted above, I followed the archive.org link to the original article and did a quick search on the last four quotes, which the article claims are from Scott's blog. None appear on the linked blo
426.
by mrandish 8mo ago
When an article is retracted it's standard to at least mention the title and what specific information was incorrect so that anyone who may have read, cited or linked it is informed what information was inaccurate. That's actually
427.
by mrandish 8mo ago
I enjoyed the essay "Dying Every Six Hours" because it's a good sci-fi short story, closer in tone and style to Martha Well's Murderbot books than Asimov. Based on other comments here, I guess I'm unusual in not get
428.
by mrandish 8mo ago
It's a very odd article with virtually no material facts and citing no credible sources. It's the kind of article which seems intentionally 'leaked' based on some ulterior motive. I also struggle to think of anything a c
429.
by mrandish 8mo ago
Alex, It's somewhat ironic that a web page about performant terminal user interfaces uses gratuitously complex CSS mask compositing and cubic gradients which reduce smooth scrolling on my 1 year-old, high-end Dell XPS laptop (>$3k)
430.
by mrandish 9mo ago
Chollet himself says "We certified these scores in the past few days." https://x.com/fchollet/status/2021983310541729894 . The ARC-AGI papers claim to show that training on a public or semi-private set of
431.
by mrandish 9mo ago
> using the current models to help develop even smarter models. That statement is plausible. However, extrapolating that to assert all the very different things which must be true to enable any form of 'singularity' would be
432.
by mrandish 9mo ago
I've been surprised how difficult it is for LLMs to simply answer "I don't know." It also seems oddly difficult for them to 'right-size' the length and depth of their answers based on prior context. I either ha
433.
by mrandish 9mo ago
> does leak per definition. As a measure focused solely on fluid intelligence, learning novel tasks and test-time adaptability, ARC-AGI was specifically designed to be resistant to pre-training - for example, unlike many mathematical and
434.
by mrandish 9mo ago
I don't have a Nest but I suspect it's even simpler than that. She didn't have a subscription but the devices still store video locally up to the capacity of onboard storage. These recent local clips used to all be locally ac
435.
by mrandish 9mo ago
> you need an internal crusade to fight these impulses. Take the high road in the short-term... Anthropic is the market leader for advanced AI coding with no serious competitor currently very close and they are preparing to IPO this year
436.
by mrandish 9mo ago
> Many people have been taught that streamlining and removal is positive. Over the past ten years or so the increasing de-featuring of software under the guise of 'simplification' has become a critical issue for power users. Fo
437.
by mrandish 9mo ago
> Recent updates say this was a unilateral call by FAA because DOD was refusing to coordinate with them for creating safety corridors for DOD drones and/or HEW usage. This is the first explanation I've seen that fits the odd fa
438.
by mrandish 9mo ago
> people ... are motivated by different things. I agree and would add that it's not just different people, it can be the same person in different modes. Sometimes I enjoying making the thing, other times I just want to enjoy having
439.
by mrandish 9mo ago
As a founder fortunate enough to get to and through a positive acquisition, just reading the chapter outline and clicking into a few I can see there's useful perspective here I wish I'd had. Since the situation is pretty rare ther
440.
by mrandish 9mo ago
As a very old school programmer who taught myself assembler in 1982 on an 8-bit 4K micro, I don't see much to mourn here. * People still craft wood furniture from felled trees entirely with hand tools. Some even make money doing it by
441.
by mrandish 9mo ago
The thesis makes sense. I agree AI coding will enable increasing disruption and competitive threats in large enterprise SaaS. For being in tech, it's a category which has been relatively free of major disruption for a while and is prob
442.
by mrandish 9mo ago
Some legislation which sounds good in concept and is well-intended ends up being having little to no positive impact in practice. But it still leaves businesses with ongoing compliance costs/risks, taxpayers footing the bill for an e
443.
by mrandish 9mo ago
This post is credited as being authored "By Claude (Opus 4.6)". At best, this attribution is incomplete because it does not include the prompt the post is responding to. An LLM generates nothing without a prompt, so asserting the
444.
by mrandish 9mo ago
Ron Wyden, Rand Paul and Justin Amash (until he left office in 2021), are the only members of congress I have respect for simply because they've proven consistently willing to ignore party loyalties and even their own political currenc
445.
by mrandish 9mo ago
> I'd expect the numbers are all real. I think a lot of people are concerned due to 1) significant variance in performance being reported by a large number of users, and 2) We have specific examples of OpenAI and other labs benchmax
446.
by mrandish 9mo ago
> the note I see the most from Claude users is running out of usage. I suspect that tells us less about model capability/efficiency and more about each company's current need to paint a specific picture for investors re: revenu
447.
by mrandish 9mo ago
A key aspect of ARC AGI is to remain highly resistant to training on test problems which is essential for ARC AGI's purpose of evaluating fluid intelligence and adaptability in solving novel problems. They do release public test sets b
448.
by mrandish 9mo ago
> The consumers are getting huge wins. However, the investors currently subsidizing those wins to below cost may be getting huge losses.
449.
by mrandish 9mo ago
> Do we still think we'll have soft take off? There's still no evidence we'll have any take off. At least in the "Foom!" sense of LLMs independently improving themselves iteratively to substantial new levels be
450.
by mrandish 9mo ago
> Yeah, these benchmarks are bogus. It's not just over-fitting to leading benchmarks, there's also too many degrees of freedom in how a model is tested (harness, etc). Until there's standardized documentation enabling inde

More ›