Graham Evans

For humans, you have to imagine physics/ medical/ material science breakthroughs which, I admit, may never emerge. However for patient digital intelligences, it just requires that a civilisation like Earth currently has, manages to survive and keep progressing engineering another 50-100 years. I'm sceptical our current golden age will continue that long but it doesn't require too much imagination to hypothesise alternative alien civilsation/ engineering pathways. There are plenty of other filter hypotheses I find more convincing.

Science fiction nerds are gonna hate me for saying this but the Great Filter is that interstellar travel is probably more or less impossible and there is no reason to come up with any other explanation for that.

Just learned about the concept of a “telescope ranch” in Texas. People pay to have their $10,000+ telescope rigs set up in the middle of TX to avoid light pollution. Every night the roof rolls back off the warehouses. Then you can remote in to your telescope and use it from anywhere in the world.

Hardly anyone in the replies and quote posts genuinely thinks this is a sensible thing for France to do, for various reasons. For a country with a smoking prevalence of 26%, it's sheer madness.

@DAMendelsohnNYC @Odd_jedi That is the face of the sea, and like the face of the wine it is opaque, liquid, dark coloured. This makes me wonder if 'wine dark sea' is not a genius translation after all.

@Odd_jedi Right: but as I said, “dark” isn’t in the Greek.

Another Homeric epithet that’s fun to think about is the one that Anglophones know as “the wine-dark sea”—which, like “wingèd words,” has by now entered the English language so forcefully that it’s hard to see what image H was actually thinking of. The Greek adjective “oinôps”…

Ok, so I learned that these are called phonesthemes :)

It is actually funny that we basically just had to find the right genus of squamate to milk. Like we were directionally correct about snake oil being a miracle drug.

@chasews @tomieinlove @pachabelcanon

@tomieinlove @pachabelcanon How well can you tell AI writing from human prose (Joyce, Poe, Austen)? I get…mixed scores 😅 aitastetest.net

No, AI writing is not “too well-written.” Here are the primary tells of AI prose (no Pangram needed): -Dreamlike, languid mood, and an inability to tighten the action -General “SSRI prose” with a certain desaturation, lacking force and vitality -Shorter paragraphs, often just 2 or 3 sentences -Little sense of sequence, causality, or a logical progression from one sentence to another -Conventional sentence structure, albeit an overuse of parallelism -Figurative language that’s disconnected from the human sensory experience -Overemphasis on reminiscence and the concept of memory -Trouble arranging objects in space, especially if they’re moving around -General lack of long descriptions of places, people, or things -Inability to do subtext, inability to embed a deeper story in a surface story -Relative humorlessness

@olorininir @anderssandberg That's not whar the OpenAI paper describes.

@grayhamandeggs @anderssandberg My point is that the team behind this *probably* 1. Tried lots of prompts 2. Did alot of pre-training and talking with the model 3. Fed in lots of synthetic data 4. Probably spat out very many solutions that needed to be checked This required lots of work and money, by humans

This is impressive: it is a problem I had actually heard of. It looks like the solution approach is surprising to mathematicians. It was a general reasoning model rather than a specialized one: bitter lesson time. I think the stochastic parrot is now nuked from orbit.

You're just wrong. The entire prompt is in the paper on page 3. It is a concise statement of the problem and what should be considered success criteria: cdn.openai.com/pdf/74c24085-1… The entire solution was outputed by the model. The researchers were brought on board after OpenAIs grading pipeline said that the solution was correct.

@anderssandberg Did the stochastic parrot solve the problem or did the billions in funding and team of human researchers reviewing and creating thousands of prompts solve the problem? More like academics credibility getting nuked from orbit by marketing copy.

@olorininir @anderssandberg "the fact is that the AI was able to do here what lots of excellent human researchers tried and failed to do" Noga Alon. Just above Thomas Bloom's comments.

@grayhamandeggs @anderssandberg Bloom says that humans were looking to prove the upper bound and did not spend serious time trying to find a counterexample, which is what the AI did. That proves my point exactly. Humans didnt even try. Also the reward was $300. Not billions.

Sure the money is symbolic but you have drawn the wrong conclusion from that quote and you are making big claims without any research in evidence. The Open AI team addressed the exact terms of the Erdos problem and, I guess, get the $500. There are 398 papers on Google Scholar addressing this, and if you poke through the titles, some are cleary interested in the whole problem. The fact that many explored the sub problem of finding larger n's within the bounds of the conjecture space is because that's more tractable. But humans were very interested in the problem of proof/ disproof.

Again Thomas Bloom: "AI solutions to several other, much easier, Erdős problems... prompted some people to wrongly assume that all of Erdős’ problems were inconsequential trivialities that had only remained unsolved because nobody had tried to prove them. This is very much not the case – many of Erdős’ problems have been intensively studied for decades"

@grayhamandeggs @anderssandberg And we know this how?

@bartek_wl @anderssandberg AI was the collaborator that made the breakthrough.

@anderssandberg It was not done by their model solely but with collaborative work with mathematicians. OAI is making tons of PR on stochastic parrots

@libertarian_ass @thestustustudio it should be slower - people call 911 because they feel in danger. that's just not a reasonable objection.

@thestustustudio Shot Spotter is nothing but security theater. In the only controlled study done with it, in Baltimore, police response was actually slower to Shot Spotter than to 911 calls. Money wasted.

🚨 After the Bullets, Cambridge Turns Off the Alarm Last night, Cambridge City Council voted to end its use of ShotSpotter, even after Police Commissioner Pauline Wells practically begged them not to shut it off. This comes after one of the city’s most brazen recent shootings, when a gunman fired 60 rounds on Memorial Drive. Wells told councilors that Cambridge police had taken three additional illegally possessed loaded firearms off the streets in just two weeks, including weapons with large-capacity magazines. In each case, she said the people carrying them were convicted felons who never should have had guns in their hands. She also warned that there have been “at least 11 times” when ShotSpotter detected gunfire in Cambridge and not a single 911 call came in. Despite Wells’s warning that “lucky is not a strategy,” the council voted to shut it off anyway.

I cannot believe the word "spruik" is a pure Australian slang term I fully thought it was like a proper word this whole time

I often see people quote a study about men leaving their wives when they get sick. However, that study was completely retracted for a coding error that nullified the result! In response, they released a super p-hacked result lol

Not an advert but a recommendation - I've purchased a few ereaders. Boox do Android based ereaders which means you can run your local library app, the Kindle app, read random epubs, mobis and pdfs you get your hands on, and all the other stuff you do on a random tablet (as long as refresh rate and colour aren't big issues). Their recent line up has been color ereaders (seem sub-optimal in their current form), and ereaders without front lights, which is just a pain. Go 10.3 Lumi is their first available model for the past year which has what all their old readers offered: bright B&W i-ink plus a front light. It's also a lot cheaper than what I've paid for my last 2 Boox models.

@CSI_Starbase ...could it be the raptor pumps can deliver arbitrarily more water pressure than they need, and so they are working out how much they can get away with stressing the system for maximum flow/ longest life span.

The fact that SpaceX is testing their new deluge system twice a day, everyday for the past 4 days is a very odd sign to me. I get the feeling they are still trying to verify a specific issue has been resolved and are not yet happy with what they are seeing.