akkad33

I use Gemini instead of Google most of the time and I think most people now go to chatgpt or gemini for things they used to Google.

I think you misunderstand. Yet the model is doing something impressive still by writing code to fix model weights on problems. But that's a whole different ballgame than just reading weights and knowing what to change. That would mean that they have an internal representation of how the model weights map to information or a specific task, which if true, throws all the interpretability research out of the window

I think the hard part of formally verifying something is not writing code. Granted it is hard and LLMs will help with that.You also need to model the user space correctly and this is a bit like causal modeling, there are many confounding factors to be taken into account and if you missed something at modeling time you'll miss bugs

Does that include mortgage/car leasing?

I'll look at it. Did he invent that concept? Control theory is a mathematical discipline with strong mathematical guarantees and conditions. Just adding a feedback loop does not make anything control theory. Anyway in my understanding he does AI evangelism without any original contribution. I see some arxiv papers on related topics.

"in touch" meaning, I don't know if this will work, let me ask Claude?

I think what differentiates efficient use of AI vs bad AI output is not AI skills themselves but comfort in understanding AI output. So domain knowledge and technical knowledge. Everything all these AI gurus preach is like snake oil or old wife's tales. There's no A/B testing done to validate any of the "context engineering" hacks or whatever. It's all based on anecdata.

Our understanding perhaps. But do you think researchers working on it had a similar understanding?

No. Has the underlying technology changed in any way?

It's not a paradigmatic shift though. It is able to do what it was already doing, but better. You can't really claim there has been a staggering improvement since then and anything that it is doing now is wildly unexpected.

If they were benign, why wait till now? Why didn't they say anything when OpenAi first released GPT and Microsoft was the biggest investor? Did something change between then and now?

No it is not forced. I jumped ship the day they force upgraded "selected" users to windows 11 and left others in the dust because their hardware was "incompatible"

People still use Windows? What is it like now? Everything AI? AI in textpad?

Social media is itself the problem, it's not another implementation or version that'll change it. You're online instead of actually interacting with real people. Anything online will be tracked, sold to advertisers or data lobbies. It's not Facebook or Google that's the problem

I switched to vscode because the complexity and runtime overhead of intellij were no longer justified with Claude writing most of the code. I work on a big java code base and every two days I had to wrangle with intellij cache, clearing it or what not. With Claude I'm working in multiple work trees at the same time and this just amplifies every issue intellij had with slow indexing and broken caches. And TBH nowadays I eyeball the code more than read, so a basic code editor with navigation seems enough

I'm reading A scout mindset. It's been an eye opener and really sets out clearly everything I've been noticing in myself and in the world. It's about the lies we tell ourselves

Wow I didn't expect to see this here. It seems in the book she's very open but it seems to have an alienating effect on readers rather than making them sympathetic for her honesty

TLA+ ensures your model is correct, not that the code is correct

Can someone do a ELI5A of how they achieve classification over any user defined list of items? Normal neural networks do a softmax over a known output set to get probabilities

Can someone tell me what is the difference between Jev and a normal neural network that does classification ?

My understanding is: it takes text input and it does one shot classification (no training data)

I don't think they do. I work in a maven shop and half of the people don't even know what to do when maven fails inexplicably

I don't agree. If anything these newer languages have better tooling and new projects are always built from ground up to support open standards like open telemetry

I don't know TLA+ but indeed that seemed to be the limitation when I read about it too. What about languages like Dafny, that allow you to "prove" your programs?

Has anyone tried formal verification with AI generated code? I can't convince my company to use it but I realise it's very easy to ask Claude to add a verification step locally on my own PRs

How do you know this?

I had tried to get long term memory out of Claude by indexing my notes with keywords and putting that in a sqllite database and Claude queries using full text search. Don't know how good it is, it seems to find things alright. My goal was to keep context small and only get Claude to ask for what it needs. Datalog seems like a great idea, will definitely try it out

Then it's not a net speed gain?

What is a problem LLMs can't solve?

Isn't supposed to be faster with LLMs?

Maybe they are dedicated hand coders!