Like most popular social media sites, you usually won’t see very valuable discussion in the comments, at least in my experience. It’s mostly for people to post news, research, and so on, and follow the big names or organizations in their field.

Most of the valuable information is diffused via posts but I do put a bit of time and effort into trying to filter out all the crap posts like memes, the faux inspirational stuff, self-aggrandizing nonsense, etc.

Fascinating that people with stutters can be helped by practicing speaking with speech jammers.

It makes me think about how ADHD medication will make people without ADHD more distractible while it’ll help focus people with it.

John Wayne Gacy is really unhappy with this feature.

I’ve often looked out of a 20th story window at pedestrians and thought, “They have no idea someone is looking at them right now,” and then I always wonder how many times that has happened to me.

That must happen to people all the time who visit that church.

Old-school AI systems from way back in the day called Expert Systems were just a crapload of IF statements. There’s never been a concrete agreed-upon definition of AI because there’s never been an agreed-upon definition of the word Intelligence.

I wish I could upvote this twice

The free version gets things wrong a bunch. It’s impressive how good GPT-4 is. Human brains are still a million times better in almost every way (they cost a few dollars of energy to operate per day, for example) but it’s really hard to believe how capable the state of the art of LLMs is until you’ve tried it.

You’re right about one thing though. Humans are able to know things, and to know when we don’t know things. Current LLMs (transformer-based architecture) simply can’t do that yet.

Look, I found your original point interesting, but if there was a major upset in the microwave industry, then that would belong in the technology section of a news site too.

I think the fundamental question is, as the Fediverse gets more popular, then how will servers get paid for? Here are some possibilities I see for how Fediverse hosting could work at scale:

• Surviving off donations alone: Possible but in my estimation unlikely, and it could veer into the territory of big donors having a controlling stake or exerting their interests.
• Instances limiting number of users to what they can afford: This would require the network of instances process to really work well.
• Big instances selling advertisements: Without oversight or moral commitment, this could easily go towards creepy personal data collection.
• Crowdsourcing the costs: This would require transparency and fundraising or some other model
• Hosts financing the operation in other ways: This could also easily get into creepy data collection practices or other dark patterns.

I hope we come up with some process or plan for avoiding the pitfalls and forging an honest and community-integrating way forward.

I haven’t heard of cognitive schema assimilation. That sounds interesting. It sounds like it might fall prey to challenges we’ve had with symbolic AI in the past though.

The threat is a new sustainable community that’s sheltered from advertising that people could leave Factbook/Instagram/whatever and go to.

So then why was Meta trying to get Threads to be on the Fediverse? Of course they’re aware of any potential threats, no matter how small.

When one side is suspiciously quiet or supportive about legislation that’s against their publicly stated goals, it’s because they secretly want it too.

I feel like I remember them being there since January of this year, which is when I started playing with ChatGPT, but I could be mistaken.

Recent papers have shown that LLMs build internal world models but about a topic as niche and complicated as cancer treatment, a chatbot based on GPT-3.5 be woefully ill-equipped to do any kind of proper reasoning.

It’s possible to build intelligent AI

What does intelligent AI that we can currently build look like?

I mean, on the ChatGPT site there’s literally a disclaimer along the bottom saying it’s able to say things that aren’t true…

IQ is mostly a pretty arbitrary and pointless metric because things like attitude, process, and creativity matter a lot more for getting results, but it can still help to diagnose learning disabilities and it has a solid statistical underpinning. The only thing it strongly correlates with is chess ability.

It looks like it!

Looks like the only completed Gen 3 nuclear reactors are in Asia, at Kashiwazaki (Japan), Kori (South Korea), Yangjiang, Fangchenggang, Tianwan (China), and Kudankulam (India).

Edit: I missed the Gen III+ part of that Wikipedia page. The other currently operation or under construction Gen 3+ reactors are in Sanmen, Shidao Bay, Taishan (China), Novovoronezh II, Leningrad II, Kursk (Russia), Akkuyu (Turkey), Rooppur (Bangladesh).