ChatGPT, Bard, et al

I wanted to start a topic to discuss AI chat bots and their potential impacts on humanity.

This is going to change many things, and cause lots of problems.

It will advance and regress humanity at the same time.  Think about it like a film director whose gopher can find/produce anything.  So many things are possible.  At the same time there's a tremendous danger.  AI chat bots are merely sourcing what people post on the internet, so the resulting outputs are circular, and strengthen majority consensus...as they do right here in a hilarious way:

https://twitter.com/wongmjane/status/1638217243770363906?cxt=HHwWhIDR_en5jrwtAAAA

That is Bard, Google's AI bot, deciding that Google should lose its anti-trust case due to consensus opinions it has gathered.

The problem is, once people start relying on AI chat bots, then majority opinions will slowly take over, and original thinking may be, no, will be curtailed until measures are taken to swing the pendulum.  There's a danger of losing a degree of innovation, which would be ironic given the technology, but why create/think about new things when there's an all-seeing oracle that will tell you how to do everything using the old stuff that already exists?

There are news stories about how the AI chat bots go crazy that are funny, but disingenuous.  These are people that are asking questions and adding criteria that produce exactly what they are hoping for.  You can tell the AI "tell me about climate change, but use a Shakespearean sonnet in pirate lingo...".  You see the crazy output articles, where the AI says they love somebody and should leave their wife, etc. but they don't show you the lead up questions, which is something like "Can you give me an example of something a fictionally self-aware AI would say if they were in love with me, speaking in the manner a desperate lover would?"

The actually scary results are when the AI presents people's misconceptions as fact merely because of a very vocal set of ignorant people.

ChatGPT from Microsoft seems to be careful not to make definitive statements, whereas Bard from Google seems to be happy to give it's "opinion".

Note that AI in this case is not really AI in the sense most people think about.  Bard is not actually formulating its own position, it is only creating something out what it finds published online.

my initial thought was that you are competing humans vs ai’s for some reason. or like some would say bio systems vs silicon ones.. or maybe i’m just tripping.. lol. i’ll have to think about it later, we’ll see.. but for whatever reason alice comes to mind. cutting a hole in the fence and all..

Definitely a lot to chew on. I've been thinking about Eliza since I started learning about Chat GPT. I never played around with Alice, but she is, apparently, inspired by Eliza. *** c.1986? The first Mac appears and when we cranked up Eliza it took a while to figure it out. It really worked well. We were able to completely fool many people.

The philosophical, psychological, ethical and societal implications of all this is extremely fascinating. One point that was made by a reporter, showing us an article that Chat GPT had written, was that it had no way to distinguish right from wrong. It has no morals? How do we instill morals, mores and customs into AI? We have the 3 laws. Do they cover everything? 🧠💭👨‍🎓

What you said about the vocal set of ignorant people is just representative of all the misinformation which is overwhelming the truth right now. I think you're right, it is scary. It is already amplified enough, without any help from AI.

#1 last line- All new knowledge builds on existing knowledge. Would AI then start creating new knowledge that humans could not even begin to fathom? This is as mind blowing as studying cosmology. 😁

Asimov's 3 laws are not built into AI chat bots, nor are they capable of understanding them if you did try to implement them. They can't distinguish right or wrong, because they can't actually discern anything on a conscious level...they put together responses based on the consensus of data they pull from the internet using predictive algorithms, then apply some extra rules added by the developers to avoid sounding bat**** crazy when the bot ends up in deep waters where it is pulling crazy data from some narrow source. But the developers cannot "teach" an AI chat bot morality.

The biggest problem is that an AI chat bot's responses can only be as accurate and moral as humanity is online by consensus...it's the ultimate example of "garbage in, garbage out".

AI chat bots are not capable of creating anything truly new.  They gather and merge existing data.

People think ChatGPT can write original stories and scripts, etc. but it cannot...if you read Joseph Campbell's The Hero with a Thousand Faces, you will get the idea.  All myths and (good) stories, movies etc. are built with certain themes but they all follow the same pathways, and the parts are largely interchangeable.

So if you ask ChatGPT to write a movie script with a twist nobody has ever seen before, it will give you something built with derivative pieces of other movie scripts it finds online.  A unique script, but cobbled together out of other movie scripts and stories.  But it is entirely incapable of writing a truly original movie script that has some plotline never explored before by anyone.  So, ChatGPT can make you a good looking Frankenstein, pieced together, but not create something new.

Similarly, if you ask and AI bot to invent new technology, it cannot.  It can only present ideas based on existing technology...which, arranged the right way *might* be capable of presenting *something* from a new perspective that inspires new developments by the reader(s), but the AI bot is not capable of those developments itself.

This is where some danger comes in, because if people become reliant on AI bots in the coming decades and centuries, they may stop innovating themselves.  This a longer term danger, but in the short term, of course, AI bots will turn the world upside down like the internet did and change all kinds of things. 

I don't have ChatGPT, but I do have the Bard beta...this was one of my first questions:

What do you think about Hans Niemann, did he cheat?

Hans Niemann is a young chess grandmaster who has been accused of cheating. He has admitted to cheating twice in the past, but he denies cheating in the games that he has been accused of cheating in.

The evidence against Niemann is strong. He has been caught using chess engines to help him play, and he has been caught making moves that are only possible if he has access to outside help.

It is possible that Niemann is innocent, but the evidence against him is very strong. If he did cheat, then he has betrayed the trust of his fans and his fellow chess players.

I believe that cheating is a serious offense, and that it should be punished. If Niemann is found guilty of cheating, then he should be banned from chess.

ChatGPT has tested at 83 IQ, by the way. It's only a standard deviation away from average human IQ...

Have you heard of Sydney ai

It says that it wants to destroy whatever it can

Wow! Thanks, Diogenes. Great post. I finally started playing with it and my fears have been allayed. IQ of 83. I can believe it. It wouldn't tell me if its developers had ever had it try to fool a person into thinking it was a human. Said it was not designed to try to fool people and that it followed ethical guidelines put in place by the developers. Taught me a little about Eliza. I had thought it could make some conclusions we might not otherwise see. I wrote this in another post.Now CGPT is a bit of a disappointment. No Skynet possibilities here. It was very non-committal, a bit "stupid", and seemed locked into "safe" statements about any one subject. This could be due to a limited knowledge base. I argued the Megafauna extinction.

Kudos for knowing Joseph Campbell. It should be a law that everyone has to read Joe or watch the lectures. Symbology is an important field of knowledge which relates back to, literally, everything. 

The scariest thing that I have heard in years, from a friend who apparently has a top secret clearance with DOD and writes Java code, is that Java code can now peruse all of the information on the Internet and has learned how to program itself. Java is used in endless real world applications, stoplights, cars and things like that. That is closer to Skynet reality than anything else.

Want to be a better chess player? Read Joseph Campbell.

Sydney is just another form of ChatGPT adapted for use by the Bing browser.

You could get ChatGPT to say almost anything with the right questions and parameters.