Thoughts on AI?

Out of curiosity, what are the thoughts of everybody on the current AI revolution? is it good? is it bad? and what should one do to survive in the changing world?
Totally not stressed about the future >w<

Basically I want Z-K IRL so AI the right step.

If we go to far it becomes Total Annihilation though. I for one welcome our CORE masters and wish for the destruction of the ARM rebels if my toaster asks.

It is silently taking over the world ;)

Current AI has issues explaining how to replace the batteries in a remote without patronizing and wasting 6 paragraphs on the ethics of buying disposable batteries. If they put that into a robot, it'll be the end of the world for whoever is wasting money funding that.

LLMs have ruined drivel for me

Previously it was like "does this person even have any point with their bloviating logorrhea"

Now it's "did they just prompt chatgpt to write this?"

It is very hard to judge. While LLMs and generative computer vision models are crazy useful tools, those people calling for halting their development, etc., are needlessly paranoid. Those models have no agency of their own, the models themselves are harmless (however, spreading fake news and such will be already is much easier with them and that is a big threat, but that won't be slowed by halting their development). All of it is just a bunch of statistics applied to a huge pile of data.

Now, their strength depends on how many task can be "solved" with pure statistics; I bet a lot of people thought this approach can't get us that far in language generation, but it did. ChatGPT and others certainly have some interesting emergent capabilites - will it be enough to correct scientific articles in, let's say, 2-5 years, or will we need something stronger than transformers? If it understands how code works and can write unseen code to deal with new tasks, why couldn't it grasp some physics concepts?
Time will tell. I personally think something a bit stronger will be needed (pure text and model capacity can only get us so far) but we might be on the right track.

What I thought was interesting was that reinforcement learning (AlphaGO - destroying that GO master, AlphaStar - beating the strongest StarCraft player, ...). But after having a (very short, potentially wrong) look at it, it seemed to me that it is just this old Agent/Graph search paradigm enhanced by some neural networks - it feels different from how humans think; we will need imho a different technique to create a model that can be as general as humans are and that can get good in games/driving/ThisTypeOfTask only after a few trials. Still, it found its use in protein identification, perhaps there are other niches where it can be as effective as well.

I am overall a bit skeptical as I have kind of seen an association between the level of enthusiasm about AI/LLM and the level of knowledge in various fields (lot of very very enthusiastic people were the ones that knew fewer things). Then after X months same people were blaming "model changes" for "bad performance" (which my guess is was just them realizing it is not as great),

That being said there is true potential in the LLM-s, but for now we don't have a clear grasp on how hard it is to use that potential. I am using LLM-s for some projects and I am baffled both by how well they work sometimes and how bad they work in couple of cases. Reproducibility and stability seem to be complex issues at this point.

One aspect mentioned rarely is that some of these models require huge computing resources. I ran couple of models on desktop and quality was much lower than what you would get via an API in the cloud. There is still lot of work until someone makes some "device" that works with comparable efficiency to a human brain (ok, there are various useless things attached to the brain, but overall it does not consume much energy)

As long as AI is developed by human programming, AI will be as flawed as us. So i think its best we start AI developing AI developing AI developing AI developing AI for many generations and have it design their own optimal processing units. That way there will be a moment AI will truly surpass us at any aspect and they might even be worthy of ruling us so we can finally have peace in the world. God bless AI!!

I found chatgdp quite underwhelming as a writing tool. Sure, you can use it to produce generic texts like applications, maybe you can use it to explain tabellas in scientific literature etc. But working with it has a structural problem.
Task: Recreate the post of Steel_Blue above so that it says exactly the same thing, but without copying the original post into it and telling it to reformulate it. First, even if you did the reformulation method, you very likely get some gibber that is maybe correct, but not exactly what Steel_Blue said. And even if it was really good, then YOU HAVE ALREADY DONE the work.
If you start from zero, it takes 10 times longer to explain the bot what you want to express than just writing it yourself.
As long as you cant plug a cable in your brain, have the bot read your thoughts and write them down - hope your organized in your head tho - communicating with it is absolutely inefficient.

Someone else came to the same conclusion as me:

I'm curious about whether anyone poasting in this thread has actually tried GPT-4

quote: I have seen people using ChatGPT to ask a question because they had no clue what keywords to put in google to get what they were looking for

To be fair, a large piece of the problem here is that all the search engines have become crap through concerted efforts of SEO companies and also of search engine companies