This is a really complicated subject and even though I have a lot to say, it's a drop in the bucket compared to what's actually going on.
ChatGTP and deep-leaning AI are just tools, they are not self-aware. But they are going to make the world incredibly more complicated soon, in both good and bad ways. We need other kinds of intelligences that are capable of collating and network the vast amount of human information out there, and then giving us summaries that we can use to implement solutions. ChatGTP is not God. You can't say "Solve this problem". You can only give it parameters and then it will search its databases for those answers, but not like a classic search engine does. It is generating the information de novo instead of simply copying and pasting results.
Humans have biases, information-volume limitations, and cognitive limitations that prevent us from recognizing a lot of patterns. AI can recognize those patterns and then bring them to our awareness. For example, the different disciplines of medicine don't really talk to one another. For years we've needed a group of high-level statisticians to dedicate their lives to collating all the medical stats from all the different fields in order to see novel root causes of disease, novel treatment potentials, and other info. We will soon have AI that can do that.
You can have a browsing AI and say to it, "Please read all of reddit and give me the best summaries of [X topic], but please leave out [XYZ] because I don't want to read those, and be sure to especially include posts that show a high level use of English."
You can vaguely know what you want to learn about but may not have the word for it and say, "What do you call that thing where we study molecules that relate to the brain and then try to turn it into medicine?" The AI will spit out: "Neurochemical medicine", and now you have the name of a topic you can research.
I already gave ChatGTP a list of all my personal interests, likes and dislikes, and the areas I'm already qualified in, and asked it to tell me what types of careers fit those definitions. It told me some things that were extremely helpful.
Wolfram Alpha is a science-based AI that is basically like a massive scientific calculator, and you need to know the underlying science in order to ask it to perform functions. Soon it will be linked to ChatGTP. So I can ask GTP to explain particle physics to me, and then it will go and ask Wolfram Alpha, in scientific terms that I don't understand, the best way to explain particle physics. Wolfram Alpha will then give ChatGTP the scientific rundown, and ChatGTP will spit that back at me in basic teaching language.
AIs are already starting to talk to AIs in order to meet human demands. In two months ChatGTP will start talking to Wolfram Alpha when you ask it about science. It's incredible.
Another example is... academia is starting to suffer from stagnation and it's because of sheer volume. There are more PhDs than ever in history, pumping out so many papers, that it's having a dilution effect on each individual paper. There is a lot of good info out there but no human beings can possible collate it all. A scholarly AI could do that. Most scholarly search engines are garbage by comparison what AI indexing could do.
The AIs themselves are not capable of good or evil because they are not aware of what they are saying. It's the human operator who determines that. We don't need to be afraid of SkyNets taking over the world. They can't generate content beyond what the humans input into their databases. ChatGTP 3.0 was fed 800 billion words of text from various volumes that the creators chose, which is not that comprehensive in the grand scheme. It can tell us a lot about mainstream knowledge, but if you ask it something obscure it will give you non-sense. AIs will be better at giving us generalities because general questions harness their entire network, whereas very specific questions drive them down single routes on their mind-maps and then they start spouting non-sense.
So I think AIs are going to become a tool that levels up humanity. AI cannot determine values, only human values can do that because we are self-aware. What AI can do is collate vast amounts of information and tell it to us in an unbiased manner, which is something humans cannot do. We can use its output to make our day to day lives easier. My friend recently needed to file a noise complaint with his landlord about another tenant. He asked ChatGTP to write a noise complaint letter that was polite yet assertive, and it generated one. He then gave it additional variables based on the output and the letter was modified accordingly. Apparently the letter was quite good. It would've taken him possibly hours to write it, so it saved him time. An AI can never decide that noise is bad and why anyone should complain, but if someone has decided (the value) that noise is bad and needs a complaint letter, it will write one.
Humanity has certain crises that we cannot solve so far, like global environmental collapse, which is coming soon. Humans think too locally, it's hard for us to grasp global problems. Government is too tied to stakeholders, greed, and geopolitics. We also fight among ourselves over our opinions. The AIs don't have this problem. They can network the data and give us novel viewpoints on our own human knowledge because they don't have a personal stake in what they are collating.
They can also organize humans (if we let them) into more efficient communication. For example... imagine two humans having a fight. The AI could recognize some of the causes of a conflict, such as words being chosen in communication, and suggest better ways that person A could talk to person B, based on an analysis of person B by the AI, so that person A no longer triggers them.
Or consider that there are already endless human groups and cooperatives in the world. If there is a global problem, an AI could say, "Group A in this part of the world should get together with groups B and C in this other part of the world, because I have read all of the research each group has done in the past 50 years and I think their combined efforts could really solve this problem." Then you ask, "Why do you say that, AI?" and then the AI gives you a 10 page treatise on why, instantly.
Yes, this can be used for manipulation, but also for good. Look at Facebook... it uses AI to make us addicted to it. Imagine if somebody created a social media platform whose AI could be asked to unaddict us to social media? "Hey AI, I've been using social media way too much lately but I'm having a hard time stepping away. Can you modify the algorithm for me to make social media less appealing?" Or the social media's algorithm is crowd-sourced and its functions are about the betterment of humanity? This would require AI to be under control of the collective, and not some corporate fat cat. Which is what this all ultimately boils down to... who is the human operator? Is it billionaires or it is it all of humanity?
My impression is, on the whole, AI could solve some of our more immediate crises and create better efficiency in our day to day lives. But what's going to happen is that all the good and bad things humanity does will just be upgraded to a whole new level by AI, but the net effect will be very little change. AI will be used to game the stock market so that the rich get richer. AI will be used to send out fleets of bots to convince people to vote for political opponents. One fleet will be sent to young-adult forums and be instructed to act like they are 18-year-old Trump supporters - hundreds of them - while another fleet will be told to infiltrate forums of the business community and act like concerned merchants who thinks a certain candidate is the right way to go. Once AI gets good at voice, it will do all the telemarketing calls, and will learn the most common objections and how to overcome them, getting better with each phone call. It could be used to find the most efficient ways to kill people in war, or it could be asked for a list of non-lethal ways to end a war given the current variables of that war. It could also be used to augment democracy because, let's face it, the old fogies running government can barely comprehend technology, so it would help if an AI could analyze entire networks in an instant and then tell the general public + government what is going on.
AI will also be used to create increasingly real-seeming artificial realities, especially once AI art gets better. AI can't tell us what is better, more valuable, more preferred, more beautiful, more humane or more cruel. It is incompatible of value preferences -- which is why human free will is still required to decide what the value of AI output is. AI displays what we ask it to display, and it displays those things by drawing on the human diaspora of information it has been fed. That's literally it. If you aren't aware of your own extreme biases, then you will use AI to create an echo chamber because of how you ask questions.
We have the means to solve material inequities on this planet right now. My main worry is that people with money and power are going to stop it from happening and will use AI to keep things just as they are. For AI to do maximum good, it has to be collectivized, and my hope is that there are enough non-corporate, non-shithead tech people out there who will create virtuous AIs that will trump everything else.
I think AI disruption is en par with the creation of the internet. It's going to affect literally everything. The possibilities are endless. And it's probably a good thing ultimately because once we created the internet we were bound to create endless amounts of data that nobody would ever be able to make sense of, except an AI.
