I’m curious why you use this language of “appears to match or even beat”, seemed pretty clear from the benchmarks that it’s well ahead of Fable. Do you doubt the reporting?
I'm using both extensively at the moment and I've not yet found that either is demonstrably better at specific tasks than the other. They're both beasts!
(I've not had Fable do as much work in Blender as Astra yet though, and it's quite possible that Astra has Fable beat on 3D modeling.)
Maybe I can help with the curiosity - his own experience is probably that astra is close to fable, not miles ahead of it. Everyone has an opinion, and some people don't trust the benchmaxxed stats. Some of us just try it for our use cases and figure out which one we like best!
its important to read it anyway because there have been and will continue to be errors in the construction of the proof software itself. which leads ai and humans alike to prove things that arent true
Totally agree. People have become totally dependent on AI even if they don't really need it.
As an analogy, I think about my dependency on Google Maps. Salt Lake City is probably the easiest city in the world to navigate because the streets are laid out in a Cartesian grid, and addresses are just literally those Cartesian coordinates (i.e. 500 South 450 East means 5 blocks south and 4 and a half blocks east of the center point, which is the SLC Mormon temple). It's trivial to know how to get to any address, but I reflexively enter in to Google Maps whenever I drive.
The premise of the story is silly. If AI gets cheaper and faster, then it certainly will get smarter (just not in the raw sense) as harnesses, tool access, memory, etc. improve. You can’t hold intelligence as a fixed variable but then imply other aspects of it improve
Not an expert, but I believe humanist would be somewhat antithetical to brutalist. Where brutalism has hard edges/corners and are "cold" humanist designs have fewer right angles and are more warm/inviting.
I make an app called Reflect [1] which is geared towards tracking these kinds of things. Admittedly, sticking with the habit and forming it in the first place can be hard. If you’re interested in talking about it, let me know! We have a discord that we link to in the app and that’s the fastest way to get a hold of me.
this is the first time ive heard of beam search. i would have reached for a genetic algorithm of some sort, although it seems like some stochastic versions of beam search exist to avoid local minima. i wonder if there are any good frameworks for building these that agents can construct and use.
I think the question is how do you keep track of the ideas that the agent is pursuing - like I was working on this for implementing a fft, and doing the optimizations, but I held its hand and was like - hey let's go back and retry this older thing you discounted because of a 3 % slowdown.
reply