@sheggle_

What a wild ride Re CFC: we did the easy economically feasible thing, so obviously climate change and AI regulation can happen in a similar fashion

Joined November 2019
  • Tweets1.4K
  • Following179
  • Followers46
18Media
Sheggle retweeted
owl
@owl_postingOct 10
terry tao: imagine, if you will, a petrifying situation. you have produced a very effective drug, but you do not know how it works. do you trust the drug? now, this is just a hypothetical, but— drug developers:
Andrew Curran
@AndrewCurran_Oct 10
This is a slide from Terence Tao's 'Math 2.0' public lecture yesterday evening at Caltech. I am highly confident that, if the cure actually worked, almost any Stage 4 cancer patient would tell you that human comprehension of its mechanism is completely irrelevant to them.
47
128
20
1,889
79,199
Sheggle retweeted
Pleometric
@pleometricOct 3
I had Claude scroll TikTok, then I gave it 200 bucks and asked for a music video 9 hours later it came back with this... I'm cooked lmaaoooo 😭😭😭

Watch Me Till The End

I had Claude scroll TikTok, then I gave it 200 bucks and asked for a music video 9 hours later it came back with this... I'm cooked lmaaoooo 😭😭😭

275
307
422
7,411
3,658,238
For Trump, at his age, pacing the frontier too much means death. Same for senators and people in congress. We're 1-2 miracle drug stories removed form them realizing this.
1
20
Sheggle retweeted
QC
@QiaochuYuanSep 12
if you can't make use of a magic button that solves any problem in your field then that is a problem with your field, not a problem with the magic button
5
12
1
246
6,010
So when is deepseek gonna do block-level selection first? Feels like the only lever that has much pull left to optimize flops/byte of their models.
1
42
You think openai feels safe to commit felonies cuz by the time they've drowned everyone in paperwork and appealed and appealed and appealed they have superintelligence and a robot army anyways?🤔
1
30
Sheggle retweeted
Liron Shapira
@lironSep 8
The exquisitely beautiful thing about theoretical math research is how it uncovers deep connections between seemingly unrelated fields. Like the connection between Erdős’s planar unit-distance conjecture and matrix multiplication. Or non-sofic groups and matrix multiplication. Or arithmetic circuit lower bounds for the permanent and matrix multiplication. Or high-dimensional sphere packing and matrix multiplication. Or Navier–Stokes and matrix multiplication.
22
57
4
776
23,201
Going on a repost spree cuz shit got me shook
17
things are getting grim in the math world.
13
71
12
1,600
109,040
Replying to @zeta_globin
yeah you're fourth in line
1
1
90
3,341
Sheggle retweeted
xlr8harder
@xlr8harderSep 9
Replying to @ElliotGlazer
imagine shinigami eyes that let you see if someone was a subreddit moderator
5
6
2
118
5,651
You're laughing? Meta made a visionary move, figured out agent forums are all you need before any of the other labs, and you're laughing?
max
@mSanterreSep 6
Why did they even do this
35
putting a hickey on her neck where he got shot for the kirkversary
145
19,069
575
237,379
3,005,222
Seeing benchmarks far and wide being topped by a new model like when fable released.
26
Next let's create a model that does this trick with J space
Boyd Kane (quantized)@beyarkaySep 3
GPT-6 Astra is very good at not thinking about the question (but then still getting the correct answer)
1
45
The Fed has no Navy and desperately wants a deal.
Justin Wolfers
@JustinWolfersSep 4
If I'm reading this correctly, the President is threatening the Fed with tariffs.
131
224
17
2,296
176,368
Stock Talk
@stocktalkweeklySep 4
*PRESIDENT TRUMP WARNS FED & BOND MARKETS: "LOWER [INTEREST] RATES, OR I'LL STOP TRADING WITH COUNTRIES WITH WHICH WE HAVE A DEFICIT"
60
651
41
8,681
605,580
Sheggle retweeted
Conks
@conksresearchSep 1
this is Biden’s global bond selloff
56
57
5
1,139
53,329
I think we'll see a shift in this in the coming years. My guess is that the attitude towards research, as agents get more capable, will move away from sexy ideas towards whatever works.
Niels Rogge
@NielsRoggeAug 16
This is very cool, but why call this AI research? Actual research outcomes: - GPT - Muon/AdamW - Dropout - LayerNorm - etc This is just writing data loading tweaks/faster kernels/gradient tricks etc on existing research
167
Load more