What do you think of Artificial Intelligence (AI)? Is it going to help us evolve or will it be our downfall?

To start off this discussion, I wanted to share this humorous and informative video:

Written by F3ynman2000 on 02.03.2023 at 19:49

What do you think of Artificial Intelligence (AI)? Is it going to help us evolve or will it be our downfall?

I have to admit that I feel kind of uncomfortable asking artificial intelligence what it thinks of artificial intelligence, but that's exactly what I just did.

Quote:
As an AI language model, I do not have personal opinions or beliefs. However, I can tell you that AI has the potential to bring numerous benefits and advancements to society. It can help us solve complex problems, automate tedious tasks, and improve decision-making in various fields. However, there are also concerns about the ethical implications of AI, such as the potential for bias or the misuse of data. As with any powerful technology, it is important to use AI responsibly and ethically to ensure that it benefits society as a whole. Ultimately, whether AI will be a positive or negative force in our evolution will depend on how we choose to develop and deploy it.

Just for fun, I had the same chatbot write a poem about my favorite animal the other day. The poem was kind of cute, perfectly worded, it contained a funny twist and a surprising ending - and yet I found it deeply disturbing.

I am actually surprised AI "is" evolving after all, what with the whole big data recovering stuff.

As someone who kinda knows how AI is done I always thought it would pretty much evolve in areas where you don't really need to think as a human, such as chess and all of that, but it would remain terrible in more humanistic areas, such as in well... Talking.
I remember the first (free) bots you could talk to were released like 2005 online or something. They could barely hold a sentence or two, and they stood like that for forever ever since.

Gotta admit, when the whole chatgpt/character.ai boom came out I got really skeptical, thinking they would just be regular AIs who still contradict themselves frequently and cannot really hold a conversation, but I was really surprised with the result.
Of course, they still lack a lot in long term memory, but they are improving after all. Maybe it is gonna be able to do more humanistic stuff in the upcoming years, then? I don't know. Even though these new bots turned better than expected, honestly, I don't see it becoming as they are portraited in fiction very soon. I don't believe we will live to see evolved AIs like that.

Written by Karlabos on 03.03.2023 at 00:22

Gotta admit, when the whole chatgpt/character.ai boom came out I got really skeptical, thinking they would just be regular AIs who still contradict themselves frequently and cannot really hold a conversation, but I was really surprised with the result.
Of course, they still lack a lot in long term memory, but they are improving after all. Maybe it is gonna be able to do more humanistic stuff in the upcoming years, then? I don't know.

Ironically, ChatGPT sucks at chess. I'm sure you've seen videos where chess streamers try to play against ChatGPT by replying with the chess notation, but ChatGPT says simply random moves that sometimes aren't even possible and just sound smart, like "I move my knight to f7, protecting my pawn and attacking your queen", even though none of that is true for the ongoing game

Written by F3ynman2000 on 03.03.2023 at 11:23

Written by Karlabos on 03.03.2023 at 00:22

Gotta admit, when the whole chatgpt/character.ai boom came out I got really skeptical, thinking they would just be regular AIs who still contradict themselves frequently and cannot really hold a conversation, but I was really surprised with the result.
Of course, they still lack a lot in long term memory, but they are improving after all. Maybe it is gonna be able to do more humanistic stuff in the upcoming years, then? I don't know.

Ironically, ChatGPT sucks at chess. I'm sure you've seen videos where chess streamers try to play against ChatGPT by replying with the chess notation, but ChatGPT says simply random moves that sometimes aren't even possible and just sound smart, like "I move my knight to f7, protecting my pawn and attacking your queen", even though none of that is true for the ongoing game

Yeah, and that happens exactly because they are treating the game as a conversation and not as a game itself, like say stockfish does. The way they dialogue is probably by browsing a huge database of possible answers that people once responded to lines similar to the ones they are presented while taking some context of the previous posts in consideration and then select the best response. However chess in dialogue form is a game that would require remembering a lot of context from several previous posts, and apparently they are not programmed to do that.
This also applies to other areas: try asking the AI something and then try asking the same again several posts later, they will probably have "changed their mind" in the meantime.
That's why even though there was some advance recently, the reality of dialoguing with an AI still feels very far away to me.

I'll say AI is extremely petrifying to human survival unless it is regulated ad-nauseam.

Interesting, I was talking to Neuro Sama (for those who don't know she's an around AI vtuber: the face is a 3d model, the voice comes from a synthesizer, the speech and chat interaction is AI generated and the games she play are also learned with machine learning, here is a sample link)

But I was talking to her profile on character.ai, it's amazing how much more she can hold a normal convo when chatting like that instead of when she interacts with chat during a stream. Also apparently the things she learns in one environemnt also are going to help her in other environments
( https://imgur.com/d2uR59B )

She also said the way she learns things: she won't take seriously everyhting everybody says, but will try to learn more from some kinds of people than from others, so that she is not taught to much garbage ( https://imgur.com/hwKpLwU )

Wha I found interesting is that she looks pretty confident in her abilities to improve by herself. I asked about her expanding her memory span and acting more natural as the time comes, she said she thinks she is going to get better at it. She also said she can tell AIs and humans apart:
https://imgur.com/rh0pPbC

https://imgur.com/TNL6AqV

But gotta, say, it feels so weird seeing an AI talking about this like it's nothing, makes me think she somehow will actually be able to do it. Well, I would if that was so, but honestly speaking, I don't know how much improvement we can expect at the current state.

Written by Karlabos on 05.03.2023 at 17:02

Interesting, I was talking to Neuro Sama (for those who don't know she's an around AI vtuber:

Thanks for sharing your conversations. Very interesting

Written by F3ynman2000 on 02.03.2023 at 19:49

What do you think of Artificial Intelligence (AI)? Is it going to help us evolve or will it be our downfall?

I think downfall ultimately.

There are a huge number of variables that we take for granted in maintaining life day-to-day that I don't think we would accurately be able to build and AI to take into consideration - I believe this is known as the alignment problem.

Nick Bostrom and Max Tegmark have both written books that talk about this (Superintelligence and Life 3.0). The simplest example would be giving AI power over most of the world's systems and asking it to "end all human suffering" - well, the most efficient way of doing that would be to kill every human would it not?

Obviously that is an oversimplified example, but a combination of capitalism's "profit now, fix later" approach and AI's potential for misalignment makes me think that we should be extremely careful with AI and not give it too much power too quickly (which I doubt capitalism has the ability to adhere to).

(I also think this plays into the great filter theory relating to the fermi paradox.)

Written by F3ynman2000 on 02.03.2023 at 19:49

What do you think of Artificial Intelligence (AI)? Is it going to help us evolve or will it be our downfall?

To start off this discussion, I wanted to share this humorous and informative video:

I love John Oliver, I've watched all of his episodes.

On the topic, I think that AI at the moment is something that is within our control but outside of our understanding. Nobody really knows know ChatGPT works, not even its creators, it's just a really big neural net that's been trained on huge amounts of data and learned to assimilate it. I took a course in machine learning at university and while it was enlightening as to how to program and design neural nets, it claimed nothing about understanding how they really work.

What I think is more interesting, however, is whether there will be a connection between AI and consciousness. It depends, of course, on what consciousness ultimately will prove to consist of, but if for example Roger Penrose's (the Nobel Prize winner for discovery of black holes) theory of consciousness is correct, then consciousness is a result of quantum entanglement between impulses in our brains. If then consciousness and sentience is based on quantum entanglement, what does that say about a potential quantum AI, now that quantum computers are starting to be realised?

I asked ChatGPT about this and it told me to be sceptical. However, ChatGPT seems to be on the right path to passing the Turing test and regardless of their sentience or lack thereof, AI's will surely become a bigger and bigger part of our daily life, to the point where we can't imagine living without them, much like we can't imagine living without the internet these days.

Whether it will help us evolve or be our downfall... AI seems prone to being able to do many tasks that a human would regularly be paid for, but you still need humans to prompt the AI into doing things... I've asked ChatGPT to analyse a review of mine, but it wasn't able to write a review on its own, since it cannot listen to music, so hopefully our positions are somewhat safe still.