*people generally might be less skeptical, I should say

Okay, but when I asked what your approach for that was you just said you’re hoping it won’t have the problems that other LLM models will. I might be less skeptical if you’d be willing to describe a methodology beyond just “we reckon this one won’t do that”

So fundamentally what you’re saying is that you don’t have a good response to Ketan’s criticism above

So your approach is that you’re hoping, against the evidence about all existing language models, that this system won’t hallucinate? And you’re planning to publish this solution to one of the major technical challenges of the field not as a groundbreaking paper but as a Dr Carl chatbot?

What’s your approach for that? Because as far as I understand it LLMs aren’t really capable of distinguishing fact from fiction …

Très bien!

Assume a conservative movement toward the South Pole over time. 1cm. 1mm. Whatever. It’s only a matter of time before we reach the moon. FYI, the rate of movement is much higher than you think it is

Thanks! Really interesting way to examine public sentiment on policy - seems like there might be a lot of things where people might be aligning themselves politically on the basis of an inaccurate understanding of current policy settings

That’s pretty interesting . Do you have a source? Genuinely asking here - not trying to question the point.

But that illusion has found an audience that doesn't merely want to be tricked, they have seen a magician pull a rabbit out of a hat and are now dedicating their lives to speculating on the complex ecology that must exist inside the hat to support the rabbits inside, the magician be damned.

Exactly. The thing he’s claiming AI provides to software engineering is actually provided by libraries and APIs. I know that, and I’ve never developed software in my life

I’m a kiwi living overseas and my understanding has always been that it’s an issue but a manageable one. Sounds like it’s gotten worse and is maybe not so manageable any more

The thing is, it’s not even incorrect! Any input is asking the question “what’s an approximation of the words statistically most likely to appear in response to this?” It’s impossible for LLMs to be consistently truthful at an architectural level. No idea how people are trusting them