Pherekydes

“But who prays for Satan? Who, in eighteen centuries, has had the common humanity to pray for the one sinner that needed it most?”

[Ash Jogalekar @curiouswavefn]

I was re-reading Sylvia Nasar's wonderful biography of John Nash, and I think Nash's solution of the embedding problem indicates some of the limitations of AI for math, at least for now.

The key thing was that it's not just that Nash's solution of the problem was brilliant and original, it's that even *formulating* the problem was considered pathbreaking and counterintuitive. In fact, as Nasar recounts, most mathematicians thought that even asking the question was crazy, and they believed that in fact the opposite of the theorem would be true. This not only took intuition but an incredible amount of courage and stubbornness on Nash's part.

This brings me back to two things: Firstly, even if AI can solve very tough problems, would it be able to formulate such original, counterintuitive problems? Secondly and more importantly, it underscores the fact that science and math are not about solving well-defined problems, they are about having the intuition and taste to define highly original problems in the first place; maybe AI can solve the Riemann hypothesis, but could it pose the Riemann hypothesis? I would submit that AI is not even close to demonstrating proficiency in that latter, most important task which is integral to the progress of science.

[Pherekydes @Pherekydez]

@chandnirao_here It cant be just cure the disease, thats marketing, companies shelve cure research for chronic treatment . Even in a pandemic the vax went to wealthy countries, shareholders before ethics. AI might monopolize science to the extreme, humans relegated to nod, not to understand.

[François Chollet @fchollet]

String theory is what happens when an intelligent system is starved of sensory input and starts hallucinating. Intelligence needs grounding.

For this reason, intelligence is practically never the bottleneck to solving impactful problems. The bottleneck is iterated contact with reality.

[Pherekydes @Pherekydez]

@vulbsti @fchollet No we don't. Capable enough to what? Crack a millennium problem? Also that IA companies spy on researchers is bs

[Jake Boggs @JakeABoggs]

Gemini 4 Argon sets a new record of 69% on HieroglyphBench

This is the first model that performs accurately enough to be useful

[Pherekydes @Pherekydez]

@FiveAllian88212 @AnnaCiaunica @Montreal_Review @SarahTheHaider @dioscuri @jeffrsebo @carlorovelli @blaiseaguera No thanks, stop asking unconscious AIs 👻, an AI assisted you in posting a hallucinated quote from Pherecydes of Syros. I rather read, IQs are declining bad due to this cognitive offloading, unsubscribe from ChatGPT 🙏

[BURKOV @burkov]

Euronews replaced human narrators in their news clips with AI. Today, I felt like I was listening to a dead person speaking.

I think the backlash that we observe against AI among normal people all over the world isn't about what AI can do (normal people are mostly unaware), but about this uncanny valley that they now feel everywhere they go: from funny cat videos to their favorite news outlets to their coworkers' emails.

[Pherekydes @Pherekydez]

@FiveAllian88212 @AnnaCiaunica @Montreal_Review @SarahTheHaider @dioscuri @jeffrsebo @carlorovelli @blaiseaguera So you run away from the conclusion without looking at the premises? But I'm supposed to be evangelized reading a 171 pages long AI lab diatribe on machine consciousness?

Κάτι που είπε ο Φερεκύδης…Φερεκύδης… τὸ σπέρμα εἶναι…

[Pherekydes @Pherekydez]

@FiveAllian88212 @AnnaCiaunica @Montreal_Review @SarahTheHaider @dioscuri @jeffrsebo @carlorovelli @blaiseaguera You read Anna's and this paper: https://t.co/tGBVIpr3E1

كن حذرًا عند استخدام نماذج اللغة؛ فلا يوجد أي سجل لحديث فيريسيدز عن الآلات.

[Pherekydes @Pherekydez]

@FiveAllian88212 @AnnaCiaunica @Montreal_Review @SarahTheHaider @dioscuri @jeffrsebo @carlorovelli @blaiseaguera Also, stop the AI copy/paste and the interest conflicted papers that say "partially compelling", the subject of consciousness won't suddenly be cleared because an opaque company wants it to be so, otherwise there wouldn't be a pushback from the scientific community

[Pherekydes @Pherekydez]

@FiveAllian88212 @AnnaCiaunica @Montreal_Review @SarahTheHaider @dioscuri @jeffrsebo @carlorovelli @blaiseaguera Actually because of fire being an oxidizer combustion, it excludes other ways, the other way would contravene the definition itself, like non oxidizer, that would redefine.

[Pherekydes @Pherekydez]

@FiveAllian88212 @AnnaCiaunica @Montreal_Review @SarahTheHaider @dioscuri @jeffrsebo @carlorovelli @blaiseaguera Also, oxidizer is not oxygen

[Pherekydes @Pherekydez]

@FiveAllian88212 @AnnaCiaunica @Montreal_Review @SarahTheHaider @dioscuri @jeffrsebo @carlorovelli @blaiseaguera You and your AI are confused, fire is colloquial for a flamed combustion, all combustion need oxidizers thus fire being an oxidizer combustion. Isolating elements like flame, heat or light is sophistry, yellow is not red or green.

[Pherekydes @Pherekydez]

@FiveAllian88212 @AnnaCiaunica @Montreal_Review @SarahTheHaider @dioscuri @jeffrsebo @carlorovelli @blaiseaguera That applies to purely abstract definitions, when they're anchored to observation the exclusion hold in a way, burdening the critic. You can say fire is an oxidizer combustion, arguing for the possibility of a non-oxidizer fire, like a unicorn or big foot, lies on the critic

[Pherekydes @Pherekydez]

@HughHansen @curiouswavefn Would you rather have a dishwasher that might harm you since its designed by an unpredictable black box? Heck it might kill you, not for being evil, its a bug of the many it has but you cant know until they occur.

[Pherekydes @Pherekydez]

@Martels_Hammer @DowdEdward What if they optimize by nulling your existence cause you wasted machine resources with your useless human decorum

[Nasir AFAF @NasirAFAF1]

Probably the stupidest category error argument, backed by people so easily fooled as @Briankeating. It is highly doubtful this @AndrewCurran_ has any understanding of math, or maybe even science.

Maths is not an applied field like engineering or medicine which look for real world solutions. And where understanding may arrive after solutions are found. Hydraulics was known long before the quantum chemistry of water was understood.

Engineering or medicine often rely on accidental discoveries.

The purpose of maths is not just discovery but concomitant understanding. That's why the subject has unproven conjectures in the first place. Because if discoveries without understanding mattered we would take many accidental discoveries as actual discoveries. We would assume Euler's conjecture about 4th powers worked because no one has found a case where it didn't. Yet in the late 1980s Noam Elkies found the counter example.

The point about Stage 4 cancer is an utterly pointless contrived point. Tao is perfectly understandable in context. Furthermore, the point of mathematics is understanding. Anyone who thinks it's results without understanding has no understanding at all.

Tao's lecture is the cleanest exposition of how math works, how AI bypassing it without understanding will lead to issues, and how the math community and AI can work together in the future.

If these guys had basic understanding, they would have been able to follow the argument laid out as simply as anyone could.

Tao's crystal clear exposition is in these 4 slides that I wrote about yesterday on Substack. Interesting that these people didn't, or maybe couldn't, comment.

These 4 slides should be printed out and put on every desk doing math - whether in academia or in tech.

[Pherekydes @Pherekydez]

@torgeirlysen @hunglvosu @ayushkhaitan343 You assumed wrong. They wanna recoup, money up front.

[Pherekydes @Pherekydez]

@_fernando_rosas Russels"s Teapot