oulipo

I mean Lean has two things: a full programming language, and proof checker. The programming language is used to create either code, or proofs. But the "created proofs" are using a very simple subset (a bit like an "assembly language"), and only need somewhat simple rules to be checked

So as long as the "somewhat simple" core that checks to proof is correct, it doesn't matter if the "programming language" has bugs: either in the end the "proof assembly code" generated by the program is right and is validated by the core, or it's not, and you need to revise your proof, or to check for either a bug in the programming language (it didn't generate the proof you were expecting), or in very rare case detect a bug in the simple "core"

It can possibly have bugs. That said, the "core" of Lean is a somewhat "simple" core that does simple things to check the proof, so it's quite rare to find meaningful bugs

The main issue would rather that the formalization itself has a bug, eg. those proofs that you're trying to formalize are not "1 + 1 = 2" but theorems which take already multiple pages just to enunciate. And in these enunciations (that are made by humans) there can be errors.

In general, once the proof has been formalized, it's supposed to be the proof of the correct theorem (that's what the Lean proof-checker checks)

There's been many bugs found in Lean already, and there are probably still a few. So that's not the main issue.

But the question is whether the AI will itself be able to decide on which problem to advance, and do it in an ethical and aligned way with humanity. And that if it does, its significance will be understood by humans.

Otherwise we're just creating a black box, easily manipulated by their owners, who would be ruling over human lives. That's not what people want

Très cool ! Si tu es sur Paris et que tu veux lancer un petit workshop pour apprendre aux autres à le faire je suis partant haha

It's not "tax those guys", it's "tax those corporations making profits to redistribute them fairly"

Exactly. That's why we NEED THE STATE to intervene and ask for wealth redistribution, because employers won't do it willingly

No, because these machines don't have a "will to live". So if they look to "become more efficient", it's only because we've told them (through their learning dataset possibly), that we, humans, need them to be more efficient.

"Pointing you" is very different from "finding questions on their own", or useful math. Pointing you presupposes a human is there to value their question.

Which is precisely my point. That's why humans will always be needed

Maths is made of two parts: questions, and solutions.

Machines could, presumably, become good enough at hammering at solutions. But they won't be very good at finding questions. Because questions are infinite. And to a machine, all these questions are the same.

Questions are "good questions" because they are good questions *to* humans. And that's what makes human mathematicians (and scientists) so important

Not tech savvy enough to push a button?

Not having to pay extra employees to maintain the fridges I guess

That's not clear. Perhaps the savings made by having connected fridges over the year has offset the hacking event waste, so overall it could still be worth it. Now they need to apply security patches, and possibly only use the connection for monitoring (read-only) and have a kind of "2FA" / physical presence required when sending commands from the web

Or it should/could be connected, but in a read-only mode (eg monitor the data). And control could require someone on the location doing a physical button (pushing an "authorize command" button physically behind the fridge, or in a local control room, or something)

It's routine for apps like Messenger to pre-download attachments (eg to compute a resized preview of the image etc)

They use ad network to try to target individual user visiting a page (eg you know your target is a male, 60 years old, living in that area, with a few other checks you can almost uniquely identify your target)

Then the ad network delivers a targeted ad that exploits a vulnerability (eg a broken JPEG decoder that, when sent a bad image, allows to run arbitrary code on your machine), and from there you have access to the computer

"Most browsers are highly secure"... erm no

Really? Nothing happened in church? No generalized paedophilia? No kids that got completely abused?

You probably never read about the stories, that's why you only remember your anecdotal story

Sure, it's always easy to imagine the "positive scenario" (the kid really did something wrong and got lightly restrained).

Now when you authorize a restrictive / possibly violent behavior on people having authority, and you will default to "believe them" (because they're the literal "adult in the room") you also open it to abuse

See what happened with police in the US...

Yes, everyone understands that. But tradeoffs that are acceptable for life-or-death questions are usually assumed to involve some human decision at some point. That's precisely what he's saying

Basically his argument is "we will still need mathematicians, because for some large classes of problems (one could argue, the most interesting ones, the "applied" ones) we still need to have a subjective human decisions of whether the machine solution is aligned with our values or not"

I think you misunderstood his point... it's the fact that if there's no human supervision involved in the "pass stage 3 trials", you don't really need if it passed it for legit reason (in which case everyone would be happy to let patients use it), or whether it found a "loophole" (eg hacked the submission to pass it without having succeeded the test)

So you're basically saying the old "I have nothing to hide" argument which has been debunked over a million times...

In particular because "I have nothing to hide" means "I have nothing to hide until now, and because of the corporations and states that have access to my data are lead right now by people I trust not to misuse that data"

But if tomorrow the government change, and suddenly your searches for "gay bars near my location" becomes illegal, you'd rather this info not be in everyone's hands...

I agree. I'm using AI on a mid-complexity system (IoT devices, encryption, concurrency, etc) and the number of errors / bad decisions taken "at first" by the AI (or if I don't guide it precisely enough) is huge, and I can only imagine how feeble and unmaintainable an entire codebase done without review from a senior coder would be

It won't defend against Putin sending 50 drones against it. And yes, it is useful to worry about nuclear proliferation, and absolutely not "silly"

It's not a "rogue AI", it' s a company letting a statistical model do actions in the real world, sending (by themselves, through that statistical model) fake informations to a federal agency...

That's one of the reasons that ecologist push-back against "technologists" who say "just build nuclear reactors everywhere"...

we want a safe a decentralized infrastructure

They literally haven't abandoned society because they're PRECISELY still communicating with it, reporting about their experiences, discoveries, etc

I don't think it's a good thing. The good thing would be to have a good education, and local groups with whom to discuss.

Blindly trusting a model that's easy to manipulate is not a good thing

But you'd have to believe in the fact that BOTH:

- their evidence is true (eg it's not just a too small sample size, in very particular conditions, etc)

- and MORE IMPORTANTLY that they haven't forgotten OTHER EVIDENCE which points to a much better way to do it

usually when you "see the world through metrics" you start to be very efficient at optimizing them... but then the first minute your metrics start pointing a different direction to the one it's "supposed" to (which happens to 100% of metrics since "all models are wrong but some are useful") and you go with the full speed of your optimization on the WRONG target, you start getting into issues quickly...