Alex Meiburg

Quantum info, Magic the Gathering jokes, Lean, whatever else. | Assistant Professor, Applied Math @ UWaterloo IQC, Perimeter Institute.

[Alex Meiburg @Timeroot]

@mccorvie @jevonduve moment

[Alex Meiburg @Timeroot]

@elon_lit It's mathematically neat that you can do this, but I don't see any way in which is this allows "malware" or an "exploit", and presenting it as such seems intellectually dishonest.

You're threat model assumes the user is downloading and running untrusted code -- or .so's!

[Alex Meiburg @Timeroot]

Seems like some automated system saw heavy GitHub Actions use and concluded it was abusing Actions for general purpose computation and so 'flagged' (i.e. shadowbanned) me.

But this flow for recovery is broken!
@github's support could use some work!

[Alex Meiburg @Timeroot]

GitHub
-Flags my account -> low rate limits
-I file ticket asking to have it re-enabled
-They respond that it violated ToS but don't elaborate where so I can't correct them
-"This ticket is archived, to respond open new ticket"
-Try to open new ticket
-Can't, rate limit 1/day🙃

[Alex Meiburg @Timeroot]

@eiselems @sudoingX Wait, can I use my VRAM for swap? Obviously only good for large paging in/out but extra 16GB at that speed is nice

[Alex Meiburg @Timeroot]

@Eggiari69 As someone who is very shit at drawing, yeah, I think I hate people like you. Because this person IS trying. The "just try it! You can do it!" is super unhelpful and genuinely harmful. 90% of people who are shit at it aren't going to post how much they hate it because why?

[Alex Meiburg @Timeroot]

@nickcammarata @csvoss @QiaochuYuan I like the post. Reads easy. :)

[Alex Meiburg @Timeroot]

@mariyaivasileva Lowkey that's a crazy reference list at the bottom. Good essay!

[Alex Meiburg @Timeroot]

I question Google's recommendations here

[Alex Meiburg @Timeroot]

@lacker I think this is a joke but I had so so many people tell me this as serious advice in undergrad....? It's a pretty common argument to math students towards finance

[Alex Meiburg @Timeroot]

@plumsirawit Between this and your other https://t.co/hQfCJsEsRs, it seems that are you confused about how a lot of math happens. Math is a lot more than problem-solving, and nothing you wrote addresses that.

[Alex Meiburg @Timeroot]

@noahlt @iwsfutcmd are you familiar with "corm"? Many things we call root vegetables are actually corms, not roots. Similarly potatoes are not roots, they're stem tubers.

[Alex Meiburg @Timeroot]

@rougeteaviolet Please do, that is ridiculous!

[Alex Meiburg @Timeroot]

@atnoire I can't empathize with the message (male only child) -- the song isn't made for me -- and evenso the song literally still gives me CHILLS every time, it's so good!!

[Alex Meiburg @Timeroot]

@littmath Do you feel like Grothendieck was "philosophically" wrong here (about how mathematics should be done), or merely "factually" wrong (he thought Deligne's trick would be limited in use and didn't appreciated how it could grow; he might have updated had he seen math today)?

[Alex Meiburg @Timeroot]

@atlanticesque ...the math is "solved" because they picture solving problems from a textbook or homework, and totally ignore the other parts of math. But if LLMs do all the problem solving, it's really fucking hard to do the other parts adequately.

[Alex Meiburg @Timeroot]

@atlanticesque AI has really only displayed any success at one of those three ("identifying problems, solving those problems, putting those solutions to use"), because that is something that can be readily trained to do. A lot of mathematicians' concern is that people will say...

[Alex Meiburg @Timeroot]

@Sauers_ This is a much more thorough analogy and I like it, but I think the marginal explanation rate on people who misunderstood Tao will not be very high 😞

[Alex Meiburg @Timeroot]

@makeitagoodone1 But it's "easy" to make nanobots that kill all the cancer, wait four years, and then [kill you/torture you/take over your brain]. It'd pass clinical trials. If they're made of small peptide machines, they'll "look" like drugs. Clinical trials aren't made for this threat model.

[Alex Meiburg @Timeroot]

I think people picture "another weird drug" and move on with accepted fears. Would you take the pill if the AI said it was nanobots that are programmed only to kill cancer? Would you take it if Elon Musk (or your least-favorite tech bro) had run the AI to design them?

[Andrew Curran @AndrewCurran_]

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.

[Alex Meiburg @Timeroot]

What if the treatment was "stare into this laptop screen as it flashes a weird picture really fast. You'll fall asleep for something from 3 to 86 days -- even the AI says it can't predict how long -- but when you wake up the cancer's stopped growing completely." Would you accept?