Pinned Tweet
"This may be a graduate class, but it's still called Algebra, so we're doing the quadratic formula on the first day!" --me, apparently
ALT Derivation of the quadratic formula by completing the square.
Noah Snyder retweeted
Me: I need to cancel my AT&T service.
AT&T: May I ask why?
Me: I’m moving.
AT&T: Where are you moving to?
Me: Somewhere else. Why do you need to know?
AT&T: It’s required for documentation purposes. It’s part of our cancellation process.
Me: Well, it’s not part of MY process. I’m asking you to cancel my service, not help me move.
AT&T: I’ll need your new address to proceed.
Me: You know what? Forget I said I was moving. I’m not moving anywhere. I simply want to cancel my service.
AT&T: Can you provide a reason for cancellation?
Me: Absolutely. Please document that AT&T is nosy, intrusive, unnecessarily complicated, and apparently believes canceling a service requires surrendering my future whereabouts.
AT&T: …
Me: …
AT&T: Okay, your service will be cancelled.
Me: So you CAN cancel it without knowing where I’m going?
AT&T: …
WHY DO COMPANIES MAKE IT SO DIFFICULT TO STOP PAYING THEM? 😭
Sometimes you just want to cancel a service without being interrogated like you’re applying for witness protection.
I don’t owe a company my life story just because I no longer want to be its customer. 😂
CFSG has some pretty unique features from the viewpoint of formalization:
1. It’s interesting and useful as a bare fact.
2. No complete correct proof exists in the literature.
3. It was genuinely plausible that the headline claim was not true.
AGMAI is just there so that OAI can make the point even harder that OAI's goal is to humiliate humanity.
Noah Snyder retweeted
The fact that OpenAI will categorically not release the reasoning traces for their new results, because the process of arriving at the result is the real product, seems quite close to the heart of the problem.
I posted a paper to Hexagon!
Historically math papers were written for two somewhat contradictory reasons:
1. For other people to read.
2. To carefully verify that the claims are true.
With AI these don't have to be done in the same paper anymore!
Both of these are slightly disappointing "negative" results. I was hoping to actually find a new tensor category! But this also means I think the second paper is useful even if you don't read it, because it tells you not to bother looking in this particular direction.
Two things Claude does (that ChatGPT didn’t do) that I find very charming:
1. Suggesting that I email or talk to specific mathematicians about questions. (“The simplest way to find out is to ask him.”)
2. Saying “my memory is” if it hasn’t checked something against the text.
If the weak positive form of NS problem is true (Statements A/B) *and* the weak negative form is true (Statments C/D, as claimed by OpenAI), can you still get $1m by solving the former?
Noah Snyder retweeted
You Can See Everything has some elements that feel like a Knives Out film from a different universe.
*chooses words carefully*
Right down to the magician-detective having a female assistant to think with of uncertain complicity.
Noah Snyder retweeted
"Elizabeth Holmes isn't interesting, she's just a scam artist" is totally BS, scam artists have always been fascinating
Noah Snyder retweeted
>even sam altman would provide global UBI if openai reached singularity
if you believe this you have a level of gullibility and lack of reality testing in the domain of political economy that if it applied to activities of daily living would qualify you for a conservatorship
