Search results

Or to double down on the idea that there is a kind of mass psychosis going on and the many many people making decisions to use more AI are all fooling themselves, perceiving it as useful when it isn't. Related Q: AI skeptics, how many of you have tried "deep research"? It does much better with...

Ime it means the author used AI and didn't care to edit it after.

Both meta and openAI have announced moves to generative short term video. It doesn't feel like the move if AGI is months away.

I mean that whether or not I find it useful is relevant, but not decisive. If it is an environmentally costly and morally questionable technology (@Umbran 's view) then me finding it useful may not make it worthwhile. That said I do think in practice it is hard to separate these concerns, and...

Another anecdote of possible interest.

It's somewhat inseparable though, right? It's hard to take an area of technological progress and say "we're going to see how it affects this useful thing but ignore everything else". NVIDIA spent decades optimizing GPUs for gamers, and one could have argued it was a waste of energy and...

Ime prompting with some variant of "are you sure about that?" identifies a significant number of hallucinations.

This is all obviously true, the costs and social benefits are distinct from whether I find it useful. But, you go on to finish this post with a large leap that has little supporting evidence. I get you don't see much benefit to it. But the transformer architecture has led to real, measurable...

Likewise, I want to respond to this specifically. I agree with your entire description of the limitations based on mechanism, but I don't agree that this causes the technology to lack value. It just means it should be used properly. If you interface with a search function and ask it for...

A few comments I want to respond to in parallel: I don't think the way other people are describing LLMs being used here matches my use of them and that explains some of the disconnect. LLMs can search, and search is crucial to using their capabilities for research. When I say I use them for...

Reminds me of this post

That's why I added the qualifier about checking the references. I'm speaking as a user here--I used scholar for years, I use LLMs now, and LLMs are better. You can choose to believe that I'm getting false information or not verifying things or tricking myself into thinking the results are...

I believe this and it's too bad. The self checks are great when I'm in a hurry. So the experience will get a bit worse because people can't stop themselves from stealing things or deliberately putting in the wrong codes.

The podcast idea I've used just a little but seems pretty cool. The ability to take a scientific paper and have it explained to me during a commute would be valuable. Umbran, if you don't mind me asking, which types of models were used in your thesis work? I'm curious because there is an...

I'm in the sciences so will not speak to the humanities. But the issue is more the quantity of information available than the density of the literature. LLMs are great at search, assuming you have the presence of mind to check the references.

It's quite helpful for learning new things and reading and summarizing literature. I use those functions as an academic. The key is being able to get outputs you can check. It's better than scholar was because it handles context better. My wife works in medicine and it's been integrated into...

I picked that as an edge case. But it's easy to imagine guiding the LLM more subtly. "I wonder if there is any cover I can use here..." for example. If it isn't running a fixed world, you have a lot of power to create things on the fly. Some players like that; it would work much better with...

So here's my 'prove me wrong'. I've seen lots of people saying "look, you can play a d&d game with a LLM as a DM"! I've never seen anyone play a long campaign with a LLM DM and enjoy it. In contrast, I have seen loads of people play cRPGs. This, despite the fact that proof is as simple as...