pull down to refresh
1 sat \ 7 replies \ @kepford OP 9h \ parent \ on: Letter to a Young Person Worrying About AI AI
Chatbots do not think. You lost me their. They are probability algorithms. AGI is very loosely defined and its kinda pointless to discuss unless we are talking about the same thing.
We are supposed to be afraid of quantum anyway.
I don't disagree with your thought about knowledge and recognizing reality though. I'm just skeptical about the power of "AI". I do think there are dangers within the scope of that tech though. Even with your examples though... nuclear has been a net good for mankind IMO. It could be used for ill and has been but it hasn't killed us yet.
We can quibble about language but if I ask ChatGPT a question it usually gives me a well thought out answer. Whether or not they think is some philosophical question. The reality is they are quite capable at thinking well, just ask any high school student trying to complete an essay.
AGI is not pointless to discuss; see my original remarks. We are living through the era of intelligence explosion. AGI is one of the most important topics of our lives.
reply
The reason I'm not maybe as concerned as many is because the more I have learned about how LLM's are created, trained, and used effectively the magic of it all has faded away. Its an evolution in computing. I still think it is up to humans what they do with technology. We are the real danger. Not the tech.
reply
Humans are still in the driver seat but with trends the way they are that will not last long. Militaries will use AI to plan their strategies. Drones will be controlled entirely by AI. Making complex decisions that run our world will be done by AI, not humans. Humans as thinkers are quickly being eclipsed by superior machines. Look at the list of things humans do better and you will notice how quickly that list has been shrinking. The trend is clear.
As far as doubting their abilities, i.e. the magic of it all fading away, I would just suggest you consider reading Situational Awareness. Look at the data, extrapolate out a few years and the trend is clear. Just for some low hanging fruit on the subject, just see how quickly ChatGPT has advanced.
reply
well thought out answer.
I know it might seem like a quibble but AI has no intent. Just like the Internet it can be used to magnify good or ill.
reply
Whether or not it has intent is irrelevant. What is your point? That it doesn't "think" like humans? I agree with you there. It doesn't think like us. And that is terrifying.
reply
Why is it terrifying at all? Maybe you should start there?
See, I think making it sound terrifying and magical is actually the worse response we could have. Instead we should be explaining how it functions. Why it fails. Why it gets things wrong so much. What it actually is.
Based on what I have seen it could be a net positive. It also could be a net negative. I don't see it as a humanity ending risk.
reply
Why is it terrifying at all?
Seeing a computer able to outperform humans at all tasks on the horizon is terrifying. No longer being the apex intelligence on this planet is terrifying. Understanding that the number one use for AGI is for military purposes is terrifying. Understanding the alignment problem is terrifying. AGI is not going to be aligned with humanity because we as a society are ill prepared to create an AGI that is humanity aligned. And even if we did it is extremely difficult to prove that what we have created is in fact humanity aligned.
This is a very doom and gloom video that I don't completely agree with but it does discuss some of the main points of Situational Awareness. Will give you a bit of my perspective: https://www.youtube.com/watch?v=k_onqn68GHY
reply