@Deepusleepy Neon is in my brain to a degree that is hard to fathom
Opus, meanwhile, keeps fumbling. it proposes fixes that aren't to my taste, or makes worse judgement calls, or it explains something badly and I have to ask it to explain better because I'm confused
Google cooked
@Deepusleepy ask neon to explain GRH to you and tell it how much math you know (i told it that I was an undergrad student with background up to advanced calculus)
blew me away
@thinkingshivers I believe this is precisely the reason why Neon is such a generational leap
It being better at writing is not just an improvement in a vacuum, it's because the model got better at larping as a person and simulating an ego
1. LLMs are just stochastic parrots.
2. Fine, articulate parrots. Still just predicting the next word.
3. They canโt even multiply.
4. Fine, they can multiply. Try olympiad math.
5. Fine, they can compete. Try winning gold.
6. Fine, gold. Someone already knew the answers.
7. Fine, an open problem. Try a Millennium Prize Problem.
8. Fine, but it wasnโt Riemann.
9. Fine, Riemann. but did it understand the proof?
10โฆ
@iPuneetSingh@theo Fuckk no, what are you talking about? First of all, that's really impossible to predict unless you train a model, and even then it won't be accurate.
Second of all, I'm gonna burn 100% of my usage either way, so I might as well just have more of it.
@Deepusleepy I was messing with Gemini Tasks, actually, and even though Gemini Tasks is kind of bad, just because Neon is such a good model, it was kind of working.
That said, I've never used Grokbot, so I might not have a good comparison point for what this is supposed to be like.