lukev

9 ms·

121.
by lukev 11mo ago
This is an issue of content, not transmission technology. Have you read the transcripts of any of these chats? It's horrifying.
122.
by lukev 11mo ago
LLMs are an interesting, useful technology. The "chatbot" format is a cognitive hazard, and places users in a funhouse mirror maze reflecting back all sorts of mental and conceptual distortions.
123.
by lukev 11mo ago
I very much agree. I've been telling folks in trainings that I do that the term "artificial intelligence" is a cognitohazard, in that it pre-consciously steers you to conceptualize a LLM as an entity . LLMs are cool and usef
124.
by lukev 11mo ago
"Form without substance" is quite possibly the most accurate way to describe LLMs in general. Extremely powerful (and extremely useful) pattern finders and replicators... and little else.
125.
by lukev 11mo ago
Well, that's what a LLM is . The problem is if one's mental model is built on "AI" instead of "LLM." The fact that LLMs can abstract concepts and do any amount of out-of-sample reasoning is impressive and in
126.
by lukev 1y ago
The bottleneck is TSMC, who also make chips for almost every other hardware vendor. TSMC is indeed increasing their production capability as fast as possible, but it's not easy... chip foundries are extremely expensive, complex, and ta
127.
gpt-oss-safeguard: A new milestone for open source safety infrastructure(roost.tools)
4 points by lukev 1y ago|1 comments
128.
by lukev 1y ago
Great, sounds awesome. I'm actually 100% in favor of this vision. So when are we going to start pivoting towards a more socialist economic system? Where are the AI leaders backing politicians with this vision? Because that's absol
129.
by lukev 1y ago
Agreed 100%, except that this is not a new development. This process has been ongoing, with automation has been taking over certain classes of jobs and doing them faster and better since the industrial revolution. And capitalism has flouri
130.
by lukev 1y ago
Knowing what I know about LLMs, from their internal architecture and from extensive experience working with them daily, I would find this kind of result highly surprising and in a clear violation of my mental model of how these things wor
131.
by lukev 1y ago
The Gell-Mann Amnesia effect. And you're absolutely right, it's extremely pronounced in LLM users.
132.
by lukev 1y ago
Framing this question as "AI good" OR "AI bad" is culture-war thinking. The real problem here is that there's clearly a strong incentive for the big labs to deceive the public (and/or themselves) about the actu
133.
by lukev 1y ago
The cool part is that none of any of this is actually that big or difficult. You can master it on-demand, or build your own substitutes if necessary. Yeah, if you chase buzzword compliance and try to learn all these things outside of a part
134.
by lukev 1y ago
I agree 100%, except for the logo, which persistently looks like something they... probably did not intend.
135.
by lukev 1y ago
How this reads to me is that a skill is "just" a bundle of prompts, scripts, and files that can be read into context as a unit. Having a sub-agent "execute" a skill makes a lot of sense from a context management, perspec
136.
by lukev 1y ago
I agree with your conclusion not to sweat all these features too much, but only because they're not hard at all to understand on demand once you realize that they all boil down to a small handful of ways to manipulate model context. Bu
137.
by lukev 1y ago
This may have been fixed as of yesterday... Version 2.0.17 added a built in "Explore" sub-agent that it seems to call quite a lot. Helps solve the inherent tradeoff between reading more files (and filling up context) and keeping t
138.
by lukev 1y ago
It definitely can and should be a factor when choosing what hardware to set your relatives up with in the first place, though.
139.
by lukev 1y ago
Generative AI for coding isn't your new junior programmer, it's the next generation of app framework.
140.
by lukev 1y ago
Think about it this way; they are encoding whole "thoughts" or "ideas" as single tokens. It's effectively a multimodal model, which handles "concept" tokens alongside "language" tokens and &quot
141.
by lukev 1y ago
> But Apple’s justification went further. “Information provided to Apple by law enforcement shows that your app violates Guideline 1.1.1 because its purpose is to provide location information about law enforcement officers that can be us
142.
by lukev 1y ago
I guess you and I define "actual churn" differently.
143.
by lukev 1y ago
As someone who has been doing frontend dev since "AJAX" was the hot new stuff... I am deeply, deeply disappointed in the field. It simultaneously has an extremely high rate of churn and an extremely low rate of actual innovation.
144.
by lukev 1y ago
There's two things potentially at stake here: 1. Whether there is an effective legal framework that prevents AI companies from generating the likenesses of real people. 2. The shared cultural value that, this is not cool actually, not
145.
by lukev 1y ago
I've been waiting for someone to build a Disqus alternative on ATProto (the protocol behind Bluesky, Blacksky, Skylight, etc.) Posts could show both under the article (like "normal" comments) and for aggregated discussion on
146.
by lukev 1y ago
It is, according to the benchmarks. I'm just taking the materials they provided at face value. If you have run your own benchmarks or have convincing anecdotes to the contrary, that would be an interesting contribution to the discussio
147.
by lukev 1y ago
No. Plan mode still works fine. They just changed it so you can't set it to use Opus in planning mode... it uses Sonnet 4.5 for both. Which makes sense Iif it really is a stronger and cheaper model.
148.
by lukev 1y ago
It means that the benchmark isn't useful anymore and we need to build a harder one. edit: as far as what the numbers mean, they are arbitrary. They are only useful insofar as you can run two models (or two versions of the same model) o
149.
by lukev 1y ago
How about we monitor all internal business communications of publicly traded companies and audit for business irregularities there? It is estimated that white collar crime causes far more economic damage than any other type.
150.
by lukev 1y ago
Alternatively, perhaps it is the duty of the executives to defend their company, its viewers and employees, when confronted with unlawful and unreasonable government demands.

More ›