ptdn

It may sound absurd, but that doesn't make it so. There are good reasons to believe that at least in many cases (some of them being complex decisions), the conscious thought process is either ineffective or counterproductive.

Obligatory reference to Peter Watts' Blindsight, as well.

The last century has involved the fastest reduction in severe poverty ever seen in human history, and it's not stopping anytime soon. We're allowed to also do other things.

The context windows of LLMs are now significantly larger than 2048 tokens, and there are clever ways to autopopulate context window to remind it of things.

Iran is not considered part of the Arabic world and is a completely different branch of Islam.

Yes, of course. Existing bodies of freshwater can be put to use. Counterpoint: https://en.wikipedia.org/wiki/Aral_Sea

Is this a weird behavior? I've always been this way.

superpower by 2020!

The mods for modern games involve making those games better for a specific audience. I've played a bunch of Stardew Valley and Minecraft mods. High-end Minecraft graphics and massive content packs make for an interesting (if buggy) experience.

On the other hand, it's completely different from the Use Map Settings games I grew up playing in Brood War, which were radically different games using Starcraft assets. DotA was literally the same thing in WC3. The game became a game engine. There's nothing like that now. (Garry's Mod?)

The golden age of moddable games ended a long time ago, and it's a major shame. Mods for Starcraft, Warcraft III, Half Life, etc. spawned thousands of sophisticated game modes that turned into dozens of independent games.

Social media delenda est. Reputation systems, politico-cultural tribalism, and human cognitive vulnerability w/r/t viral information have turned the entire cultural landscape into anarcho-1984.

Don't look at me. I was screaming about COVID in January 2020. The lib moral panic changing directions on a dime without a change in intensity was disorienting.

I am compelled to point out that in one of the info pages, the site includes screenshots of a conversation with ChatGPT where the author claims to trick AI detection by generating text with a lower temperature. But asking ChatGPT, through the LLM interface, to lower the temperature doesn't lower the temperature. There's no mechanism for it to do so. It may have some (nevertheless real) placebo effect, because the LLM thinks it should behave different and assigns some vague "meaning" to "temperature" -- but this isn't a technical change to the model operation.

It does mention a hope for long-term durable immune response, but the point about being able to pre-emptively roll out a universal vaccine is probably more important. This would lower the total rate of infections by a lot and allow people to reliably get a useful vaccine without waiting to see which strains are dominant that year. (Also, an immune response doesn't mean complete immunity. mRNA Covid vaccines don't prevent you from getting Covid. It lowers the severity and spread.)

I don't know that ChatGPT would be very successful at it, but I could imagine a multimodal LLM being extremely useful. There are tens of thousands of untranslated tablets.

Yes, of course. The only people with good debugging skills are the people who have spent a lot of time debugging their own code (or the code of others). However, in an LLM-dominated environment, it may be plausible for someone to develop strong debugging skills while having only mediocre programming skills. This would be similar to the "boot camp web developer" archetype who has reasonable skills only in a narrow domain.

Full transparency: I think I'm one of those bad programmers who is a good debugger, but I've also been a full-time Linux nerd since Ubuntu 8.04, so I'm very comfortable reading error messages.

I suspect that a coordination between a human programmer and an LLM doesn't require strong programming skills, but it does require strong debugging fundamentals. A month ago I had ChatGPT write a function in Racket just given a text description. Take two lists of symbols of any arbitrary length (but only if both lists are the same size) and construct a new list which selects one at random from the other two lists at the same location. There was some other logic in there, too, based on the way I'd done the structs.

ChatGPT wrote the function perfectly on the first shot, but then I realized it was only working most of the time -- turned out ChatGPT had done a really obvious off-by-one error in the loop, and it was breaking on (1/n) attempts where n is the size of the list.

It's exactly the same as how ChatGPT usually knows what formulas and approaches to take when solving graduate-level mathematics, and its reasoning about the problem is pretty good, but it can't get the right answer because it can't add integers reliably.