geraneum

13 ms·

271.
by geraneum 1y ago
> reasoning models know when they are close to hallucinating because they are lacking context or understanding and know that they could solve this with a question. I’m interested. Where can I read more about this?
272.
by geraneum 1y ago
Hence the flurry of articles, controversial interviews and aggressive marketing and the pitch to replace all white color jobs.
273.
by geraneum 1y ago
Since I can’t reply under you answer for some reason I put it here. We can have a constructive discussion instead. My problem was not actually parsing what you said. I’m questioning the assumption if populace collectively modeling exponenti
274.
by geraneum 1y ago
I believe he's talking to money, to investors. He does it through Axios, CNN, BBC, etc. Their company is not sustainable at this rate. None of the LLM service providers are. They need money for now and that's why they talk like th
275.
by geraneum 1y ago
disagree on what? You have not put forward a coherent statement. I had to fix your sentence. ;)
276.
by geraneum 1y ago
> if they aren't direct LLM users My manager, a direct LLM user, uses the latest models to confirm his assumptions. If they are not confirmed on the first try, he then proceeds to form the question differently until gets what he wan
277.
by geraneum 1y ago
It’s true that there are similarities between what you mentioned and what’s happening in this case. From the article: > The result is a test-time loop that looks less like “chat with a compiler” in the case of sequential revision, and mo
278.
by geraneum 1y ago
“A populace modeling exponential change”. Yeah, that’s just word salad.
279.
by geraneum 1y ago
I’d love a journalist that do their job. For example when someone like this CEO pulls a number out of their ass, maybe push them on how they arrived at this? Why does it displace 50%? Why 70? Why not 45?
280.
by geraneum 1y ago
They are. The audience of this talk is not normal people. He’s excited and is targeting a specific group in his messaging. The author is a person like majority.
281.
by geraneum 1y ago
> horrific disloyalty of both companies and employees There’s no such a thing as loyalty in employer-employee relationships. There’s money, there’s work and there’s [collective] leverage. We need to learn a thing or two from blue collars
282.
by geraneum 1y ago
> when you apply to the populace at large What does this mean? What do you apply to populace at large? Do you mean a populace doesn’t model the exponential change right?
283.
by geraneum 1y ago
> Humans are not familiar with exponential change Humans are. We have tools to measure exponential growth empirically. It was done for COVID (i.e. epidemiologists do that usually) and is done for economy and other aspects of our life. If
284.
by geraneum 1y ago
> It's a perversion of the free market We can, together, overcome such challenges when we accept that "The purpose of a system is what it does".
285.
by geraneum 1y ago
> But that already happened to lots of industries and lots of people, we never cared before about them We did. Why do you think labor laws, unions, etc. exist? Why do you think communism was appealing as an idea in the beginning to many?
286.
by geraneum 1y ago
> LLMs follow instructions. Not really, not always. To anyone who’s used the latest LLMs extensively, it’s clear that this is not something you can reliably assume even with the constraints you mentioned.
287.
by geraneum 1y ago
> Sounds like Claude is doing the right thing, at least for the general problem. No not really. The engineer’s job is work within the constraints first and when a big change is really warranted, pursue it. It’s having a structural issue
288.
by geraneum 1y ago
This has become a common recurrence recently. Have a problem with clear definition and evaluation function. Let LLM reduce the size of solution space. LLMs are very good at pattern reconstruction, and if the solution has a similar pattern t
289.
by geraneum 1y ago
Just saw this today: “US citizen detained by immigration officials who dismissed his Real ID as fake” https://www.theguardian.com/us-news/2025/may/24/us-citizen-d...
290.
by geraneum 1y ago
The funny thing is, the AI that is so good that it can replace 90% of developers, can probably replace the vast majority of software shops as well. What I guess will happen eventually, is that at some point AI (not the current one) will rep
291.
by geraneum 1y ago
> from frontier labs The target market of these companies are software shops. You don’t see them advertising the model capabilities in i.e. Civil Engineering much. They are trying specifically to sell a cheaper replacement of software en
292.
by geraneum 1y ago
> describe some made-up DSL Ironically, for something like the parent suggested i.e. a rules engine, this is the main work.
293.
by geraneum 1y ago
> Writing down the architecture and guidelines more explicitly? Yes, very explicit like “if (condition) do (action)” and get more explicit when… oh wait!
294.
by geraneum 1y ago
> I humbly disagree Both your experience and what the article (research) says can be valid at the same time. That’s how statistics works.
295.
by geraneum 1y ago
> The model definitely iteratively built up (useful and correct even) text that wasn't directly in the training data The text is highly likely in training data, as it’s textbook arithmetic instructions. It’s the number that is proba
296.
by geraneum 1y ago
> it indeed can build up text that is not directly present in its training data. I’m curious how you know that.
297.
by geraneum 1y ago
> But your talking about Copilot makes it seem like you're nowhere near the cutting edge use of AI, so you don't have a well-formed opinion about it You can use Claude, Gemini, etc through Copilot and you can use the agent mode
298.
by geraneum 1y ago
I like this comment because you can frame a lot of other responses here using this GPS analogy. People saying LLMs help me think or help me learn (better my skill) or help me validate my ideas, etc. is like saying I use the GPS to improve m
299.
by geraneum 1y ago
> talking to each other in something other than English WiFi?
300.
by geraneum 2y ago
sometimes I read these sentiments in 2025, and read about “Shanidar 1” who lived ~50k years ago and just feel… sad.

More ›