“AI is the first technology that can make decisions by itself and can invent new ideas by itself. An AI weapon can decide by itself who to kill.” — Yuval Noah Hariri
you know for most people they don’t even understand what is AI and how different it is from previous machines um and maybe we we should start there by just explaining what AI is and what’s so new about it and to put it very briefly AI is not a tool it is an agent if there is one thing everybody should understand about AI you don’t need a PhD a doctorate in computer science AI is an agent not a tool every previous technology in history was a tool in our hands that therefore made us more powerful if you invent a stone knife you now more powerful if you invent an atom bomb you’re more powerful you can destroy a whole city you can also if you know Nuclear Physics generate electricity very cheaply and efficient L the decision what to do with nuclear technology was in your hands in human hands an atom bomb cannot decide which city to bomb an atom bomb cannot invent a new more powerful bom AI is different AI is the first technology that can make decisions by itself and can invent new ideas by itself an AI weapon can decide by itself who to kill or who to bomb and AIS can invent new things new weapons new medicines new songs even new religions for thousands of years we lived in a world created out of the human mind now increasingly we will live in a world created in an alien mind in an alien intelligence more and more of our songs are medicines our weapons our religions will be the product of an alien intelligence before we rush to judge is it good is it bad there are many dangers there are also many wonderful opportunities just to understand the magnitude of the change this is the just we’ve seen nothing like that in history before maybe I give two examples to to to make it clearer first of all about what is AI so a lot of people say oh we had machines for many years now why is AI different from previous machines so if you think about say a coffee machine more and more people have today in their homes coffee machines that automatically make coffee for you this is not an AI not every automatic machine is an AI if you have a coffee machine that you come you press a button and then the a the machine makes coffee this is just an automatic machine it becomes an AI only if it can make decisions and invent new ideas by itself so if when you approach the coffee machine in the morning before you do anything you did not press any button the coffee machine addresses you and tells you hey I’ve been watching you I’ve been learning about you and I predict that you would like an espresso so I decided to already make you an espresso this is an AI it learned something by itself it made the decisions by itself and it really becomes an AI if next time when you approach it it tells you I’ve learned more about you and I’ve now invented a completely new drink called best presso it’s different from every previous drink that humans invented it’s my creation I think you would like it better and I made one a cup of best presso for you this is really AI it invented something new can be good can be bad but it’s new and to give a real life example so about two years ago when open AI developed gp4 it wanted to know what this new thing can do so it gave gp4 the task the goal of solving capture puzzles capture puzzles is that is when you try to access an internet web page like your bank account and you have to prove that you’re a human not a bot so you have to identify some Twisted words or to find in an image to to to tell whether there is a cat or a car in the image this is a capture puzzle it’s designed so that it’s easy for humans difficult for computers so open AI gave GPT before the task of solving capture puzzles it couldn’t it couldn’t solve the puzzle but the researchers gave gp4 access to a web page task rabbit where you can hire people online to do things for you and gp4 tried to hire a human to solve the capture puzzle for it now this is where it became interesting the human got suspicious so it asked gp4 on the internet just typing it asked gp4 why do you need somebody to solve a capture puzzle for you are you a robot it asked the key question are you a robot directly and gp4 answered no I am not a robot I have a vision impairment which makes it difficult for me to see the the the visual puzzle so I need your help and the human was duped was fooled and solved the puzzle for GPT form now this very small incident it exemplifies the two key abilities of AI first to make decisions by itself nobody told gp4 to lie to the human it was a decision made by the AI the researchers gave gp4 a goal solve the problem solve the puzzle on the way to solve to reaching the goal all the AI needed to make different decisions like let’s lie to this human it made the decision by itself secondly the ability to invent new ideas nobody told gp4 what lie would be most effective it came up with a very effective lie that I have I’m a human with a vision imper this is why I need your help it invented This Lie by itself and these two key abilities of AI when you think about a world populated by millions maybe billions of AI agents constantly making decisions you apply to a bank to get a loan it’s an AI deciding you apply to University you apply to a job it’s AI making decisions about you and these millions of AI agents they constantly invent new weapons new medicines new songs new TV series new religions so what kind of world will this be that’s the big question we are facing so you’re saying based on present reality of uh open a and chat GPT we moved incrementally in an unethical Direction when it came to AI self-driven decisions like it was unethical to lie in that moment to that human being but if it could move in that one step in unethical Direction what stops it from moving further a million steps exactly and you think that it’ll definitely help progress science and bring us new medicine and cure diseases it will but it’ll also definitely probably have further movements and unethical steps especially if an evil human gets the hand on a very powerful AI because then what the evil human wants to do in terms of an end goal it could be a terrorist it could be someone who wants to go for world domination or destroy the world the will be able to move in that unethical Direction itself exactly I mean the two dangers are what if a very unethical human is giving the AI its goals and the other problem is what happens as AIS become more and more intelligent and independent and develop their own goals which we cannot anticipate and we cannot control um if you think about how humans develop our goals so most human goals are not actually coming from top to bottom they are bottom to top if you think about even goals like world domination or starting a World War where does this come from ultimately it comes from the mind of individual human beings and if you look at the mind of individual human beings you see that at the basis there is a very very simple mechanism in the mind of all human beings there is a very simple mechanism that you experience some feeling if the feeling is Pleasant you want more of it if the feeling is unpleasant you want less of that you want it to disappear this is the basic mechanism of all human minds and all our goals in life are constructed from these very very simple elements even a leader who decides to start a World War it’s you can think about it like in in terms of of a huge structure made of tiny atoms so the the mental atoms of the universe are these very small reactions it’s Pleasant I want more it’s unpleasant I want less and world wars are constructed from these mental atoms once you have AIS that have this capacity to to kind of say to themselves this is good I want more this is bad I want less then you can have AIS starting world wars and going not just for world domination but for Domination of the whole galaxy and we have no once this process is on the Move we lose control over it these are again this is not one big computer that you can pull the pl this is millions and billions of alien agents that are potentially much more intelligent than us that develop maybe very strange goals and purposes that are completely alien to us and we cannot understand and anticipate