johnnienaked

4 ms·

1.
by johnnienaked 2mo ago
My apologies, I didn't see that
2.
by johnnienaked 2mo ago
They actually produce weak outputs with extremely high effort
3.
by johnnienaked 2mo ago
More often than it's worth most of the time
4.
by johnnienaked 2mo ago
Or you have extremely simple taxes
5.
by johnnienaked 2mo ago
It has very little to do with lag time on tax policy
6.
by johnnienaked 2mo ago
I can attest to this. It's so odd that a technology can solve a millennium math problem but completely messes up pretty basic financial calculations.
7.
by johnnienaked 2mo ago
They won't go to jail either
8.
by johnnienaked 2mo ago
Anyone else think we need a pause? Anyone? Bueller?
9.
by johnnienaked 2mo ago
A bicameral mind cool
10.
by johnnienaked 2mo ago
One of the failings of modern society is its inability to consider that not everything can be quantified.
11.
by johnnienaked 2mo ago
Anthropic's profit estimate they released a while ago was a masterclass in deceptive reporting. They didn't include training costs, stock based compensation, or liabilities. Token gross margin is not profit.
12.
by johnnienaked 2mo ago
So we're arresting the people at Anthropic that made it possible, right? Right?
13.
by johnnienaked 2mo ago
How can you not believe it? Neither of them are even close to making a profit and the market is saturated with cheaper alternatives. This is a ploy for regulatory capture.
14.
by johnnienaked 2mo ago
It's a ridiculous explanation that makes no sense.
15.
by johnnienaked 2mo ago
I especially loved the part where he claims they spent $60m on compute to push on N-S because of a Twitter rumor, and they totally didn't steal the idea from mathematicians using their tools.
16.
by johnnienaked 2mo ago
Sounds like a great time for open mathematical research is upon us. /s
17.
by johnnienaked 2mo ago
Token Bill Hunting
18.
by johnnienaked 2mo ago
It's too late. Sub models are deployed at every major organization in the United States and all it will take is turning off the option to improve the model for them to train directly on your own personal workflow, which CEOs will greed
19.
by johnnienaked 2mo ago
They don't need to train a whole model. They can feed it new information and fine tune it.
20.
by johnnienaked 2mo ago
How is it unclear? The entire point of deploying models across corporate America is to train on your workflows. Eventually replacing you with digital you is why they're doing it!
21.
by johnnienaked 2mo ago
It wouldn't be shocking at all. They stole human data to train the first models and they've been stealing it ever since to train new models. Stealing mathematicians private chats and private research and taking credit for it would
22.
by johnnienaked 2mo ago
They stole it.
23.
by johnnienaked 2mo ago
LLMs do nothing but steal, and the companies that own them are fully aware and eager to do it.
24.
by johnnienaked 2mo ago
LLMs are nothing but giant theft machines.
25.
by johnnienaked 3mo ago
Because, like 98% of people in this space, you don't mention or even consider cost. Improving the lower bound of Riemann is impressive, but how impressive would it remain if it was announced that training and inference cost $1 billion
26.
by johnnienaked 3mo ago
I wouldn't be surprised if half these proofs turn out to be well crafted hallucinations, barring of course the ones actually verified in Lean
27.
by johnnienaked 3mo ago
Yet a significant percentage of people simultaneously believe it "reasoned" its way to novel mathematical proofs. Something doesn't jive
28.
by johnnienaked 3mo ago
I wonder how much it cost(s)
29.
by johnnienaked 3mo ago
Net good for Meta, they hope
30.
by johnnienaked 3mo ago
It's literally just brute forcing lol

More ›