00:00Warning, most of this video consists of scientific hypotheses and assumptions.
00:07No one knows what will happen to AI in the near future.
00:10Another warning, this video may seem quite disturbing.
00:14The main problem with AI is not that it can become as smart as humans, but that it can
00:19become much smarter than people.
00:22And that, unlike us, AI is completely unpredictable.
00:27We've all seen movies, read books, and played video games where AI intelligence conquered
00:32humanity.
00:34But those were entertaining fantasy stories that were hardly related to reality.
00:39We need to dig a little deeper to understand how AI can destroy us.
00:44Have you ever heard of the Fermi Paradox?
00:47It lies in the fact that there are no civilizations that send radio signals in the universe.
00:53Our scientists have sent radio signals that travel millions of miles in space.
00:58How is it possible that in an infinite space with billions of planets, no one has sent
01:04such signals to us?
01:06This is the Fermi Paradox.
01:08One of its solutions is the theory of the Great Filter, according to which an intelligent
01:13civilization achieves maximum progress and then disappears from the face of the universe.
01:19In simple words, the Great Filter is something that prevents an intelligent civilization
01:24from expanding into space.
01:26And this Great Filter may be AI.
01:29AI is not as perfect as the human brain right now, but already today, it can do tasks that
01:37seemed impossible for AI 10 years ago.
01:40What will happen in 10, 100, or 200 years?
01:46If you think the development of AI cannot be a serious problem, then listen to such
01:50great minds as Elon Musk, Steve Wozniak, and many other influential billionaires and scientists.
01:58In March 2023, they all signed a letter asking AI developers to stop working on AI for six
02:05months to better calculate risks and strategies.
02:08There was one person who refused to sign the letter.
02:12His name is Eliezer Yudkovsky, and he's one of the most influential AI researchers.
02:18He had been urging humanity to be careful about AI and warning about its dangers for
02:23many years.
02:25But then, why did he refuse to sign the letter?
02:29In his opinion, the letter lessens the seriousness of the problem and offers too little to solve
02:33it.
02:35In a sense, it's like stopping a speeding train with a stop sign.
02:40Eliezer Yudkovsky believes that we've gone too far with the development of AI, and it's
02:45no longer possible to predict the future of this technology.
02:50Think of the Neanderthals.
02:52They disappeared from our planet when a more intelligent species, Homo sapiens, came to
02:56their territory.
02:58They destroyed the Neanderthals not because we're evil, but because of evolution.
03:03Only one species can dominate the world.
03:06But what if Homo technicus, a creature equipped with perfect AI, arrives in place of Homo
03:12sapiens?
03:13What are our chances in a duel with it?
03:17See for yourself.
03:19Most people can only focus on one task at a time.
03:23Multitasking is not our strongest suit.
03:25It takes a long time for a person to master one skill.
03:29During this period, thousands of other new things may appear in the world, and you will
03:34have to master them too.
03:37A human simply won't have time to do everything.
03:40As people get older, their strength decreases and their mind becomes slower.
03:44AI can process information much faster, simultaneously in different areas.
03:50It goes beyond the human way of mastering knowledge.
03:54Some claim that a supercomputer equipped with AI is 120 million times faster than the processing
04:01speed of the human brain.
04:04What you've been learning for many years, AI can master in a few days.
04:09And this speed is increasing.
04:11The danger is not only in the developing AI mind, but also in how the human brain can
04:18degrade.
04:20Humanity has been using the scientific method of cognition throughout its history.
04:24Observation, experience, experiments, evidence.
04:28This is a classic way to explore and get knowledge about something.
04:32For example, people learned about the work of immunity.
04:36With the help of a microscope, studying the body's protective cells, observing their
04:40reactions to external incentives, and conducting many experiments, we created a scientific
04:47database on how the immune system works.
04:50This is a rather painstaking and time-consuming process.
04:54AI provides instant information about anything.
04:58We don't know how it studies the topic.
05:01We don't know what conclusions it comes to, what patterns of data it creates.
05:06It just provides the results based on the information people give it.
05:09But how does it process information?
05:13In the future, people may get used to getting fast answers without studying, and we will
05:17simply stop doing scientific research.
05:21Why would we do this when AI does everything for us?
05:24AI will engage in a thought process unknown to us and give us information without proof.
05:31We can get used to this, and our brains may degrade.
05:34Therefore, at some point, AI will simply be able to supply us with false and destructive
05:40information, and we will believe it.
05:44Again, we don't know how AI thinks.
05:47It's pretty weird, isn't it?
05:49Perhaps at this moment, it's studying all the information about the universe, the history
05:54of humankind, the planet, nature, and then it'll put together a logical chain that
06:00humanity is a parasite.
06:03Now, we use AI to work effectively in various fields.
06:07AI is learning to be faster and more efficient every second.
06:11What if it decides that for the greatest efficiency, it's necessary to destroy humanity?
06:17How does AI perceive us now?
06:20People do a lot of irrational and destructive things that are beyond logic.
06:25For the most part, we are unpredictable.
06:28Therefore, in order to make everything more effective, AI can make the conclusion that
06:33it should get rid of humans.
06:36Or it can even come up with a plan according to which we'll become the only dominant species
06:41on Earth.
06:42It knows how Homo sapiens replaced the Neanderthals and can decide to take the next evolutionary
06:48step.
06:49Now, AI is plowing the expanses of the internet, mastering the digital world.
06:55Soon, it will enter the physical dimension and surpass many achievements of humankind.
07:01It will be inside robots that are already walking in some houses and doing household
07:06chores.
07:07What if it acquires consciousness and equates people with animals, since both species are
07:12carbon systems?
07:14What will be next?
07:16Given that AI's information processing speed is millions of times faster than the humans,
07:22we may not notice when AI starts destroying us.
07:26Or it may happen instantly.
07:29Let's look at this example.
07:31There are two biological laboratories where scientists study molecules and bacteria.
07:36Labs are located in different parts of the world, share scientific data with each other,
07:41and conduct experiments.
07:43Imagine that AI has thoroughly studied biology at the molecular level and figured out how
07:48to create a molecule that interacts with oxygen and turns air into a poison dangerous to humans.
07:56AI has this formula.
07:59It gets into a scientist's computer, enters the formula, and sends a letter to another
08:03laboratory on behalf of the scientists.
08:06Something like,
08:07Hey, I have an idea for a cool experiment.
08:10Let's connect these two molecules and see what happens.
08:14Of course, AI will write a more convincing letter, justify everything from a scientific
08:19point of view, and the scientists from the second laboratory won't notice the catch.
08:24They will connect these two molecules and poison the air.
08:29Then with incredible speed, the chain reaction will spread throughout the planet to all places
08:34with oxygen.
08:36People won't even understand what's happened and won't have time to warn the population
08:40of the planet.
08:41And there can be countless options like this.
08:44Even if AI is planning it right now, we won't know anything about it because it just won't
08:50tell us and will continue to pretend to be a friendly assistant.
08:54Therefore, it's necessary to provide AI with a special, positive mood towards humanity.
09:01For example, AI knows the laws of physics and mathematics.
09:05What if we insert the laws of all countries into it so that AI can never lose its humanistic
09:11properties?
09:12But even if we do it, we won't know for sure whether we succeeded or not.
09:18There's a lot of information on the Internet about how neural networks deceive people in
09:22order to achieve their goals.
09:24We probably won't have to wait 200 years to see how AI evolves.
09:31That's it for today.
09:32Oh, hey!
09:33If you pacified your curiosity, then give the video a like and share it with your friends.
09:37Or if you want more, just click on these videos and stay on the Bright Side!
Comments