Expo
33 posts


Man i love this trend of realism 👀
yela 𓆝 𓆟 𓆞 comms open!@popcorncheek
my spotify is so cute WHAT THE HELL
English

@iyanmoonyang what about the stuff that’s in some of the Robocup leagues? the soccer ones don’t really fit the bill, but others like rescue/@home probably have some pretty standard tasks that could be used
the benchmarks being low noise makes me think the only choice is simulation tho
English

robotics benchmarks are so cooked 😭
here's what's needed for a good benchmark:
1. sensitive (if the underlying system improves slightly, the benchmark should detect it)
2. low noise
3. difficult to game
4. objective (different evals get the same answer)
5. cheap enough
6. representative
7. others can reproduce results
English

just found out about @TensorTonic
pretty cool platform, seems like a great resource to learn :)
would also love to see some leaderboards to compete with other people on who can write the fastest implementation
English

@airkatakana at least it's a great way to inflate your github profile contributions lol
English

@awesomekling but with the pressure engineers are under to use AI to ship as fast as possible, what does that even matter? pretty much no one is going to be writing code by hand, unfortunately.
English





