where dreams become reality ⋆˙⟡ 20yo • prev: openai (robotics), a16z, discord, lmarena

Joined March 2020
Include

Only show posts containing:

Exclude

Hide posts containing:

Time range
-
Minimum likes
Replying to @freemanjiangg
so awesome
1
564
they deserve it!
4
444
Some stats on #php2shell. In the 10 hours before it was publicly announced, across the @Cloudflare network we saw thousands of attempts to exploit it. That jumped to tens of thousands in the 45 minutes post announcement, hundreds of thousands within 2 hours, and millions now.
269
3,162
642
21,568
2,083,008
Replying to @lucyjcai
curious if the model can summarize its learnings in a way where u can put it into a new thread and it does just as well (or how well it does)
4
219
Replying to @dimentary
confused. doesnt it all go in the square hole?
1
2
261
Replying to @aidanmantine
let's do it!!!
1
101
Replying to @aidanmantine
congrats!
2
3
540
Replying to @photomatt
awesome!
523
this has always been the case tbf, but vulnerabilities are becoming more and more major tools like cloudflare are incredible, but not widely distributed enough
152
Replying to @Pranavdoma0323
curious, do u think NN won't take care of them?
1
65
Replying to @harinsrikanth
didn't even think of that. hw + architecture codesign (if we can get chip tapeouts not reserved years into the future) will be really cool
312
Replying to @harinsrikanth
yes fs, certainly not for serving at scale yet but cool that you can just do it
2
8
2,235
Replying to @stevens60768
the idea of prompt to real life is pretty compelling. i think people like seeing robots too
1
175
silently, the issue of NVIDIA's CUDA lock-in seems to have gone away as codex can just rewrite the kernels for AMD
14
9
3
540
26,452
Replying to @kir_lev
good counter example
4
370
Replying to @aliteracy
i- in context learning? 😵‍💫
1
73
to be fair, many things in ML that work well don't show up in evolution, and we shouldn't try to encode human intuition of how things work if we r scale pilled(?)
1
10
1,061
Replying to @cdngdev @aliteracy
(the held in place strat is purely something im curious about, this is clearly not the case for robots)
1
132
Replying to @aliteracy
good points! with one eye closed we are used to our environments a bit tho, and we do lots of tricks like moving side to side slightly/viewing different angles. i wonder if u are held in place in a completely novel room with objects you've never seen before if grabbing and doing things would be frustrating this side to side/different angles strat is something that robotics models will probably learn? most failures ive noticed with astra/other models recently are depth estimation problems ive solved by adding a few extra cameras
2
1
155
Replying to @aliteracy
im sure trained models could get really good at this but why dont they show up in evolution? (well, lots of things that work well don't show up in evolution. and maybe we shouldn't make the mistake of encoding human intuition of how this should work re richard sutton)
1
5
473