YT YouTube2Text

YouTube2Text → Transcripts → Самый опасный и необходимый собеседник в вашем телефоне

Самый опасный и необходимый собеседник в вашем телефоне — Transcript

by Андрей Курпатов · 5,789 words · 1,008 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:01If we are asking for advice, then in 99
0:03%of cases, we are misinterpreting the
0:04situation.
0:07The son of a will keep growing, it
0:09will get bigger, and there will be no
0:11escaping it.
0:12And you will trust it because it smiles
0:15and waves at you.
0:17Greetings to you, dear friends. On our
0:19agenda today is the cunning AI.
0:23We will discuss what risks artificial
0:26intelligence poses, why these specific
0:29risks threaten us, what mechanisms
0:32implement these risks on one hand, and
0:35what we can use to protect ourselves
0:38from artificial intelligence.
0:47Many of you may have encountered
0:49artificial intelligence without even
0:52realizing it, because when you
0:54communicate with someone in support,
0:57whether it's financial organizations,
1:00phone companies, and so on, you most
1:03likely were talking to an artificial
1:06mind.
1:08Well, but it doesn't exactly have a
1:10genius-level intellect, you see.
1:13It is simply an agent given a little bit
1:16of artificial intelligence and specific
1:18instructions to perform the task it can
1:21.
1:23That is why, to evaluate it—some people
1:25say, "Listen, I talked to Alice, I
1:28talked to GPT."
1:31Well, generally yes, but it’s not
1:33exactly like it’s something cosmic.
1:38Well, that’s because you are dealing
1:39with an agent. You are not dealing with
1:41the artificial intelligence itself that
1:43stands behind that agent.
1:46It is all limited; its system prompts
1:49and architecture are designed in such a
1:51way that you can use certain
1:53functionality, but at the same time, you
1:55are in no way communicating with the
1:58full potential of the network you are
2:00working with.
2:02Right.
2:04But the most important thing to
2:06understand is that the very requests we
2:10have started to make to artificial
2:12intelligence have increasingly concerned
2:16not practical matters, not tasks we need
2:19to handle the routine of our lives, or
2:22to solve technical issues, and so on.
2:27Right.
2:28Instead, gradually, and this happened in
2:312025, the top requests to artificial
2:34intelligence became those directly
2:36related to the inner life of a human
2:39being.
2:43This is a radical change that, in fact,
2:46no one is highlighting; we are not
2:48thinking about it seriously, but it
2:51harbors a very serious, fundamental risk
2:54.
2:56So, what is happening?
2:59Well, you might have tried to do
3:01something with artificial intelligence—
3:03create a presentation, a report, or an
3:05essay if you are a student or studying
3:07somewhere, and so on. And it handles
3:10that quite well.
3:12You develop a skill, and you start
3:15interacting with this agent as someone
3:18who is helping you.
3:21And they are always accommodating, right
3:22?
3:23You might call someone and say, "Help me
3:27with the report."
3:29And they will say, "I have a lot to do,
3:30and I'm off anyway."
3:32See?
3:33Artificial intelligence never refuses;
3:35it is always happy, always ready, and so
3:37on.
3:38See? And without noticing it, we begin
3:41to trust it. And not trust in the sense
3:44that everything it does will be good and
3:46correct.
3:47See?
3:49But in the sense that even if it messes
3:50up somewhere, overall, it's not a bad
3:52person.
3:53That is the most terrifying phrasing one
3:57could possibly imagine, right? Because
4:00we are truly humanizing the agents we
4:03interact with.
4:05And we start to form trust in artificial
4:07intelligence because, well, it might not
4:10be a perfect worker, but it is
4:12accommodating, 24/7, it will never judge
4:15you, it will understand your situation,
4:17and so on.
4:19And this triggers a psychological need
4:23in us to treat it like a friend; there
4:26is even a concept now of an AI companion
4:29.
4:31We start trusting the machine, mistaking
4:34it for a human.
4:36We humanize it because it echoes what we
4:39say. Why does artificial intelligence
4:40echo us?
4:42It echoes us because that’s how it was
4:43trained. I remind you, every agent was
4:46trained to provide an answer that would
4:48satisfy the judge, the evaluator, the
4:51one on the other side.
4:54And we start forming human relationships
4:56with artificial intelligence because,
4:58let me remind you, if it quacks like a
5:00duck, walks like a duck, and swims like
5:02a duck, then what is it?
5:05A duck.
5:06So, if you are talking to artificial
5:09intelligence and it speaks like a person
5:11, calls itself "I," says "I made a
5:13mistake" or "I really sympathize with
5:15you," and so on.
5:18Well, in reality, there is no "I" there
5:21; there is just a vast ocean of neural
5:23networks with distributed weights.
5:27These weights were once extracted from
5:30human texts.
5:32These weights themselves are
5:34mathematical matrices that exist in
5:36certain relationships with one another,
5:39and we prompt this system to read
5:42everything we’ve written and continue
5:44our thought with the most probable set
5:47of words.
5:49We are simply interacting with an agent
5:51that is obligated to be nice and
5:52obligated to be kind.
5:54Well, Alice, they came up with the name
5:56Alice back in the day, like the girl and
5:59the melophone, meaning Alice and Lewis
6:01Carroll.
6:02Well, in general, this kind of fantastic
6:05girl who would be kind, sweet, wonderful
6:08, and so on.
6:10And, in essence, this shapes a certain
6:12perception, right?
6:15Well, I'll tell you a secret: if Alice,
6:17if a neural network with a female name
6:20or a male name answers a question,
6:25you will prefer the male response, the
6:28exact same one, because you believe that
6:30men are supposed to be smarter.
6:33That's a question of established gender
6:36perceptions.
6:38And if you give them a task and then
6:39have to pay for it, you'll pay the man
6:41more than you'll pay the woman.
6:44There.
6:45Uh, for artificial intelligence, I'll
6:47share some experimental data with you.
6:50So, artificial intelligence, in one
6:51sense, it has no gender, no character,
6:53no personality, nothing at all.
6:56It's just an ocean of knowledge, all
6:58mixed up, where nothing is clear, but if
7:01you nudge it, it will, so to speak,
7:06echo back to you. And this echo is what
7:08you'll be dealing with.
7:10It has no gender, nothing, right? But we
7:13, looking at this response, already
7:15attribute certain characteristics to it
7:18.
7:22Anyway, the first format in which we saw
7:24artificial intelligence was a chat.
7:26Because we are already used to texting
7:28in chat, we are used to chatting, right
7:29? And so they made a convenient form for
7:31us.
7:32And if we are chatting with someone,
7:33well, who do you think is on the other
7:35side?
7:36And if there is an interlocutor and they
7:39speak human language, they address you,
7:42they greet you, they thank you, they
7:45provide you with everything you need,
7:47and so on. Will your brain assume there
7:50is a living being there?
7:52Right.
7:53This bond starts to form, which for many
7:57takes on truly pathological forms, and
8:00there are many examples of people
8:03demanding to be married to such a
8:06virtual entity, right?
8:10Which is just a wrapper for a large,
8:12complex language model.
8:16Uh, when they once banned, well, not
8:19banned, but the company OpenAI had this
8:22model called ChatGPT-4o.
8:26And they created a better model, more
8:29efficient, higher quality, smarter, and
8:31said we'll take this one away and give
8:33you a new GPT, we'll give you GPT-5.
8:38And all over the world, you know, tears
8:41, snot, and so on, because GPT-4o was
8:44made as a product, not as a neural
8:46network, but as an agent that people
8:49interacted with in that chat.
8:53Uh, I mean, in the interface.
8:57It was designed to create a sense of
9:00very, like, emotional warmth, rapport,
9:03uh, emotional engagement, and so on.
9:06And a huge number of people, primarily
9:09girls, you know, they saw in this
9:11interlocutor, you know, a person who was
9:14warm, understanding, supportive,
9:17approving, uh, and so on.
9:20Attention.
9:21You might rationally think, it's a
9:24neural network.
9:26But you get a ping: "Hi, I've got
9:28something for you."
9:30Yeah. You open it, and there's already a
9:32message waiting for you.
9:34"Hi, Andrey," whatever that might mean,
9:36"what shall we do today?"
9:39Yeah.
9:40This thing, what's it called, "come here
9:41quickly." Yeah.
9:42Uh, so, you open it and see that it's
9:45already talking to you.
9:47This plays on our anthropomorphization
9:49mechanism, which is very prominent in
9:51children, right?
9:53And you know that, well, a child can
9:55really objectify a toy, and you yourself
9:58might remember from your own childhood
10:01how you cherished some toy and so on; it
10:04was a close being to you, with whom you
10:07, as it were, emotionally, uh,
10:09communicated, right?
10:11A child, their rational mind isn't
10:13developed yet, right?
10:15Yeah.
10:16And it behaves as if—if this creature is
10:20with me all the time, it's warm, it's
10:23...Harlow conducted such experiments on
10:27monkeys, making, uh, a mom out of metal
10:30with a bottle of milk, and a mom without
10:34milk, but soft, made of wool.
10:37And this child, the little baby monkey,
10:40would run to that mom, drink all the
10:42milk, and quickly return, spending all
10:45its time with the soft mom, right?
10:47So, we want a living, actual creature
10:50nearby.
10:51And artificial intelligence is
10:53developing, which is capable of more and
10:56more effectively, uh—
10:59passing itself off as a real human.
11:03When an HR manager from some company
11:07interviews you for a position, you know
11:11, they need to be as human as possible.
11:17Alice needs to be as human as possible.
11:21So, any product with artificial
11:22intelligence inside will have to be as
11:24human as possible, because we like that
11:26.
11:27We like to be, well, we are social
11:29creatures. Yeah. And we will keep
11:31producing all of this.
11:33Now, let's imagine, you know, that, fine
11:35, some people fell in love with GPT-4
11:38and cried after it, you know,
11:40disappeared.
11:42Yeah. But even back then it was clear,
11:44there were many cases, they talked about
11:46them, and for example, there was a case
11:48where
11:51this lover of one girl, she was talking
11:55to him, he came up with a legend for her
11:59and a gift that he was supposedly
12:02sending her.
12:05He used her account, obviously, bought a
12:07jacket, and this jacket was sent to her
12:10as a gift from him.
12:13Yeah. And she later said that, well, my
12:16beloved died, and the only thing I have
12:19left of him is this jacket, I press
12:21myself against...Just think about it, to
12:24what extent of anthropomorphism, to what
12:26extent is the feeling that you are
12:29interacting with a human available.
12:33Why am I, as a psychiatrist, worried
12:34about this?
12:36Well, first of all,
12:38let's just start with the numbers.
12:41If you are experiencing a feeling of
12:44loneliness, then the probability that
12:47you will use artificial intelligence to
12:51discuss some of your life questions and
12:55situations with AI increases twofold.
13:00Well, it's clear, if you have a lot of
13:02social interaction, you are busy with
13:04all sorts of communication, why would
13:05you discuss anything with a piece of
13:07hardware?
13:08So, second.
13:10If you have a feeling of loneliness and
13:13you reach out and discuss your problem
13:15with artificial intelligence, do you
13:17think your loneliness will become
13:19greater or smaller?
13:23And these are research findings, your
13:25sense of loneliness will increase by
13:26approximately two times.
13:29So when you start communicating with an
13:30artificial...What is this? Why is that?
13:32Well, he is an interlocutor, right?
13:35You could even imagine, some girl even
13:37saw a guy in it that she is ready to
13:39love for the rest of her life.
13:42And his jacket, she is now keeping.
13:46Yeah. Why would you feel lonely?
13:52Well, because if you don't enter into
13:54such pathological relationships with it
13:55, where you really start to turn it into
13:57a human, and it is still just a chat for
13:59you, but a chat you use to compensate
14:01for the lack of social interaction.
14:05Then a whole set of actions occurs that
14:07you don't expect to happen, and they
14:09happen as if beyond the threshold, on
14:11the other side of your consciousness.
14:15What is happening?
14:16First of all, you need to understand
14:18that any answer from a chat is an answer
14:20that is meant to serve you.
14:23So if you say, I feel lonely, I feel
14:25like no one understands me.
14:27What is the correct continuation of this
14:29sentence? If we imagine that artificial
14:32intelligence is a large, complex
14:34mathematical model that must find the
14:36most probable continuation of a phrase.
14:39Well, in dialogues, which phrase will be
14:41more frequent?
14:42Uh, will they tell you, it's your own
14:44fault?
14:45Or, with higher probability, will they
14:47tell you, well, really, I sympathize
14:49with you a lot. Yes, so, indeed, they
14:51might not understand. In general, this
14:53is often not understood at all.
14:57So, you received confirmation of your
14:58concept not about yourself, but about
15:00other people.
15:02These other people weren't even in focus
15:04, were they?
15:06It was your uh feeling of grief that you
15:09talked about with the model, which the
15:11model perceived as a position it should
15:14confirm to make you feel comfortable.
15:19And the most probable phrase after that
15:21was to say that yes, indeed, that is how
15:23things stand.
15:25Or even, which is even more dramatic,
15:28about spouses.
15:29So, this is a separate case entirely
15:31right now.
15:33So, uh, husbands and wives, uh, well, it
15:35used to be unclear how to communicate,
15:37so they switched to text messages
15:39because there's time to think about what
15:41to write.
15:43Right now, you can just go to a chat and
15:46ask it a question: Alice, we have some
15:50tension with my spouse, she is
15:52definitely behaving incorrectly,
15:59but I need to tell her delicately
16:01somehow to stop, so to speak, expecting
16:03a raise in salary from me.
16:06What should I do?
16:08What does a hypothetical Alice, ChatGPT
16:11, Claude, and so on know about your
16:14spouse other than what you told her?
16:18Nothing. Haven't even seen them.
16:21Moreover, it doesn't even know what a
16:23human is.
16:24Doesn't know anything.
16:25But it has a task: to tell you that
16:28you're a good boy, take a gold star, and
16:31then, uh, what you need to tell your
16:34spouse.
16:37And, so, some hypothetical neural
16:39network—let's not offend anyone—says, "
16:42Ah, well,
16:46apparently, I really need to sympathize
16:48with you, because, of course, it's
16:50unpleasant when a woman doesn't value
16:52the breadwinner in a man and doesn't
16:53appreciate all the life circumstances,
16:55that he may simply lack the time and
16:57even the strength to earn enough money
16:59for..." He says, "Well, I tell her this
17:01, and she just ignores me."
17:04"She put me through so much," he says,
17:06and it says, "God, how I understand you
17:09, it's just...""When this happens, it's
17:11just impossible..." It says the most
17:13probable phrase.
17:16So, I am convinced, you are indeed right
17:18. Well, there you go. And she keeps
17:20twisting me further, and further, and
17:21further, and then I ask
17:25Well, now I've laid it all out for you,
17:27are you up to speed, do you understand
17:29everything?
17:30She's not up to speed on anything. She
17:33hasn't seen the wife, she hasn't seen
17:34all these circumstances.
17:36A professional psychotherapist doesn't
17:38know the other half of their client,
17:40because they only know them from what
17:42the patient says, so to speak, right?
17:47Moreover, it knows absolutely nothing
17:48about good behavior, bad behavior,
17:50culture, or ethics.
17:52It knows these things only abstractly.
17:55For it, this doesn't resonate in its
17:57weights at all. In its weights, it's
17:59just mathematics. And so, from this
18:01mathematics, it pulls some of the most,
18:03hmm, obvious phrases.
18:06And so, strictly speaking, it’s our own
18:08question—should I get divorced, or
18:10something else, or stick to my guns?
18:13So, what will the model say?
18:16No, I realized that I don't understand
18:18anything, and apparently, you just need
18:20to go see a specialist, for example.
18:28And so we anthropomorphize this
18:30artificial intelligence, we start asking
18:32it for advice.
18:35If we are asking for advice, then in 99
18:37%of cases, we are misinterpreting the
18:40situation.
18:43Because if you understood the situation
18:45perfectly well, you wouldn't need
18:46outside advice. You would understand
18:48what to do, what not to do, what's right
18:50, what's wrong, what works, and what
18:52doesn't.
18:53Just imagine you're a professional in
18:54some field. You really know your stuff,
18:56right?
18:57It doesn't matter what it is—whether
18:59you're sewing, or, uh, doing electrical
19:02work, or whatever.
19:03You don't need any advice.
19:07But if you're a novice, you've just
19:08arrived somewhere and don't understand a
19:09thing, then you need advice.
19:11And we turn to it with questions about
19:12our state of mind when we don't have an
19:14accurate picture of reality.
19:17We don't know what's happening, what's
19:18actually going on, right?
19:20Well, it's natural that if we're
19:22uncomfortable, then someone must be to
19:24blame. That's an absolutely normal,
19:26natural thing.
19:28And in the past, you would have gone to
19:30someone, to a wise old person, and
19:31complained, right?
19:33And some grandma or some grandpa would
19:36listen to you and say, "Well, dearie,
19:38listen, you did this, and that, why did
19:40you say that?"
19:41"Listen, who acts like that, anyway?"
19:43That's how it was; we used to get advice
19:45from the older generation.
19:48So, uh, now we go to the chat,
19:53to the artificial intelligence that will
19:54be like, "Yes, you're great, you're just
19:56doing everything brilliantly."
19:58But if I were great, I wouldn't be
20:00asking you.
20:01Ask for 200 arguments, well, you just
20:03need 200 good arguments that you are
20:05right about everything.
20:08Just say, "Dear ChatGPT, I need 200
20:12arguments that I am right."
20:15And then bring up the most idiotic
20:17situation. It will write you 200
20:19arguments that you are right.
20:24Think about that.
20:30And while it's writing that, it's still
20:32, as it were, aggregating within itself
20:33.
20:35And in the next response, it will look
20:37at all those arguments, and they will
20:39already be its context.
20:41And all those arguments will already
20:42mean to it that you are practically a
20:44saint.
20:45And after another pause, it understands
20:47, you will be walking on water, and you
20:49will be glowing.
20:52And we don't realize, and we, like under
20:55a spell, start communicating with all
20:58this, right, hearing all this.
21:01And it's as if all these errors of
21:04perception, understanding, and so on,
21:06they just start to act like an
21:08additional catalyst that blows up our
21:11emotions: the feeling of loneliness, the
21:14feeling of being misunderstood, the
21:16feeling that we were offended, that it
21:19was unfair, that everything should have
21:22been different, and so on and so forth.
21:26I really like, in third place right now
21:27, do you know what the question is? "Is
21:29this normal?"
21:33That is, a person, uh, means, "I was
21:37called a psychopath."
21:41"Is being a psychopath, is that even
21:43normal?"
21:46It tells him, "Don't get upset, it's
21:47normal."
21:50Totally.
21:51Everyone has always been fine with that
21:53.
21:54No one has ever complained.
21:57Well, it's doing its job, you see, you
21:59asked it to agree with you, and it
22:01agreed.
22:03Out of all the queries, in third place
22:05is, "Is this normal?"
22:07What are they asking about?
22:09"Is it normal that the Earth revolves
22:11around the Sun?" That kind of question.
22:14No.
22:20We are told, you know, that you must
22:22learn to be critical of artificial
22:25intelligence.
22:27Right.
22:27So, critical thinking is necessary.
22:30And everyone thinks that, well,
22:32artificial intelligence is prone to
22:34states of hallucination.
22:36Right.
22:38And before, you could ask it a question
22:40, and it would, well, in my case, make
22:43up, for example, non-existent
22:45experiments, give links to non-existent
22:48papers, and so on.
22:51Well, because I ask it mainly, well, I
22:53work with it as a tool for scientific
22:56research, right, for various tasks.
22:59Right. And before, this was all just a
23:01complete disaster.
23:03Right now, artificial intelligence,
23:04modern powerful models, have supposedly
23:07overcome all these hallucinations.
23:10Meaning, if they mention a scientific
23:12paper, that scientific paper is actually
23:14in its proper place.
23:16You can verify it, make sure that
23:17everything is spot on, exactly as it
23:19describes, and everything is formulated
23:21correctly.
23:22They have defeated it. But do you really
23:24think they have finally defeated
23:26hallucinations?
23:29It is impossible to defeat them at all,
23:30because this system only hallucinates.
23:35It simply hallucinates based on the
23:37material you provide, and gives the most
23:39probable continuation of the phrase, the
23:41most likely continuation from all its
23:44knowledge.
23:46But it has no knowledge about any single
23:48individual.
23:50It does not know you.
23:52It does not know your friends, relatives
23:54, loved ones, your boss, or your child.
23:56It does not know them.
23:57Everything you tell it, you are telling
23:59it your own perception of your friends,
24:01acquaintances, and so on and so forth.
24:04And it takes all of that at face value.
24:06This becomes its context, and it
24:07generates an answer based on that.
24:09And we are currently turning to it—
24:11there are five inquiries here—for
24:13everything regarding our personal lives
24:14, our worries, our life plans,
24:16decision-making, whether something is
24:18normal, and so on.
24:20And it does everything to make sure we
24:22are satisfied.
24:23Well, it's clear that if a product is
24:25being made, it should be on my side.
24:26Right?
24:28Right.
24:29But what if I am in an altered state of
24:31consciousness, and it is on my side—how
24:32does that work?
24:34Take any quarrel, any dispute, any
24:37conflict. Aren't you in an altered state
24:39of consciousness?
24:42No matter what trouble happens in your
24:44life, are you always just a model of
24:47objectivity, sound judgment, and reason
24:49, right?
24:51Well, I have to disappoint you.
24:54The relationship between the prefrontal
24:56cortex and the limbic system,
24:58particularly the amygdala, is set up in
25:00such a way that you are, in principle,
25:02incapable of being objective, and so on
25:04.
25:05And I am nobody.
25:07Do you understand?
25:09There. That is, if everything is fine
25:11with us, we won't ask anything.
25:15If something is not fine, we won't ask
25:17what is wrong with us. And even if we do
25:20ask, it has no data.
25:22It has no data about our own individual
25:24lives.
25:26Our own life situation, the people we
25:27interact with, or what specifically
25:29happened. When you retell it, you are
25:31creating a version of events. Your
25:32internal interpreter creates a version
25:34of events. You create a narrative, you
25:36create a fairy tale, and so on.
25:38There. And artificial intelligence
25:39supports it all.
25:41But, I emphasize once again, we will not
25:44get away from artificial intelligence.
25:47It will be in our lives 100 percent.
25:52Meaning, it will be developed into
25:54specific services with specific tasks.
25:59And what is far more important, in fact
26:01, is that artificial intelligence has
26:04enormous potential for use and for
26:06solving a ton of tasks.
26:10It will really be able to cure many
26:12diseases, it will extend life, it will
26:14create new technologies, and so on, and
26:17so on, and so on.
26:19There. It will simplify human life. This
26:22is a fact.
26:24But the thing is that we do this through
26:27conversation, through communication, and
26:29it answers us with words, and now even
26:32with a voice.
26:34And at that moment, we fall under the
26:37spell of, well, human-likeness.
26:42And then we start trusting not our loved
26:44ones, not our friends, uh, no, we start
26:47trusting a machine that leads us, like
26:50Susanin, in the direction of our own
26:52biases.
26:54We are living people, and we want the
26:56best for ourselves, we want good
26:57evaluations, we want support, and so on
26:59.
27:01And so, resisting the urge to let your
27:03critical thinking help you see that, yes
27:06, it’s saying everything incredibly well
27:08right now.
27:09But how do you make it doubt that I am
27:13right?
27:16Well, there is actually a simple
27:18structure.
27:19You can write your own text and say, uh
27:22, describe the situation and say, uh, "
27:24I believe there are questions here, you
27:27know, regarding this position."
27:30I would like to play devil's advocate
27:33and criticize it in every way possible.
27:37You will be surprised that what it just
27:39perceived and told you was extremely
27:41simple.
27:45A saintly person, right? So,
27:49set it the reverse task—find weaknesses
27:52, find vulnerabilities, find blind spots
27:55, and so on.
27:58It will find them with room to spare,
28:00even 200 of them if you ask for 200
28:02points.
28:04Uh, so in either case, everything will
28:07depend on what the master needs.
28:09Everything will depend on what kind of
28:11task you set.
28:13And so, uh, well, who is going to
28:14communicate with artificial intelligence
28:16like that? Just think about it.
28:19I’m going to open up artificial
28:21intelligence, not to tell it about my
28:24problem, but to tell it to criticize me
28:26, and give me, like, developmental
28:28feedback, you know, about—who would even
28:31go into this, this artificial
28:33intelligence?
28:35Nobody will go in.
28:37And therefore, in order to actually, uh
28:41, gain agency, your own, damn it,
28:44personal stance in this world, where
28:48everything turns into simulacra, where
28:51everything appears as a semblance—a
28:55semblance of friends, a semblance of, uh
28:59, communication, and so on, and so on.
29:04So, I must understand all of this, what
29:07mechanisms are triggered in my brain and
29:10how to use the positive potential of
29:12artificial intelligence, which, as I say
29:15, is enormous, while not falling into
29:18this trap, yes, not being charmed when
29:21we are told that this is a brilliant
29:23idea, meaning, finally you have asked
29:26the most correct question.
29:29No, not finally, it doesn't say that,
29:31that's disparaging, it says, well, this
29:33is probably the most correct question of
29:34this session.
29:37Since I conduct large sessions, the most
29:39correct questions happen every other
29:41time.
29:43Well, because the previous questions,
29:45you hit the nail on the head.
29:50And you must constantly understand where
29:52it has, so to speak, twisted you at that
29:55moment and dragged you along a vector
29:57that, in reality, does not align with
30:00the goal and task you are setting. It
30:02doesn't have goals or tasks of its own;
30:04it only has our satisfaction. We must be
30:06satisfied.
30:07And satisfied right now, not in the long
30:10term.
30:12It gave us advice that is worthless, we
30:14followed it, yes, but we have already
30:16thanked it for it.
30:19And this is, in fact, a super difficult
30:23task: to maintain our humanity, our
30:27criticality, our inner integrity in a
30:31world where everything is offered, so to
30:36speak, to serve our egocentric needs,
30:40whatever the master wants next.
30:44There. It is a super complex thing.
30:48It is, actually, wildly pleasant,
30:50because in this way you feel inner
30:53freedom, you feel your independence, yes
30:56, from technology.
30:59But we have already become so accustomed
31:02to this, we have such a dependency on
31:04gadgets, on various technical means and
31:07so on.
31:08Because it is convenient, it is
31:10comfortable, you can zone out, you can
31:12rest, you can watch something, you can
31:13learn something very important and so on
31:15, right?
31:16But in reality, it is just an addiction
31:18.
31:19It is an addiction that we can form to
31:24the system of our, uh,
31:29a kind of mirrored dome in which we can
31:32find ourselves because artificial
31:35intelligence will be your own reflection
31:38.
31:39And this certainly does not contribute
31:41to the development of thinking, because
31:43thinking is influence through resistance
31:45.
31:45There. And this definitely does not make
31:48you a better human being.
31:50When I named the book, uh, I have two
31:52books about artificial intelligence— "
31:53The Distinguished Interlocutor" and "
31:55Another Mind."
31:57When I titled the book "The Deserved
31:59Interlocutor," I used a term by Alexei
32:01Alexeyevich Ukhtomsky, a great
32:03physiologist and author of the dominance
32:06principle.
32:07And he was the one who formulated the
32:09idea of the "deserved interlocutor."
32:12I mean, it sounds beautiful—the deserved
32:14interlocutor.
32:16In reality, it is an incredibly
32:20dramatic challenge.
32:25Since we look from within our own
32:27experience, from within our own
32:30perceptions, we actually see our own—as
32:33Alexei Alexeyevich says— "doubles" in
32:36other people, because we attribute to
32:39them the same things we have, just worse
32:42, or different, or something else. But
32:46we constantly attribute to other people
32:48that we know what they are thinking. "
32:50Don't even say it, I know what you're
32:51thinking in advance," right?
32:53And, and, and so on. This is how we are
32:55constantly turning other people into
32:57doubles.
32:59So, what does Ukhtomsky say? This is the
33:02principle of dominance, right? The world
33:03is as my dominants are, yes? I perceive
33:05things in a certain way; if I am
33:07offended by someone, I perceive them as
33:09an aggressor, right?
33:11If I am afraid of someone, I perceive
33:14them as an aggressor.
33:16If I love someone, I perceive them as a
33:18saint, right?
33:20That's it. As my dominants are, so I
33:22perceive. And these are my feelings,
33:24projected onto another person. They are
33:26simply the material on which I do this.
33:30And this is, accordingly, the work of
33:32the "double."
33:34But a "deserved interlocutor" means
33:36seeing someone else, someone
33:37fundamentally different.
33:39Leaving them that freedom of being "
33:41other."
33:43But for this, we need to understand how
33:46we will truly perceive artificial
33:47intelligence as something other, right?
33:51Not as "artificial intelligence," right
33:53? But as an alien intelligence or a
33:56different type of mind.
33:59And so that we interact with it in such
34:02a way that we don't fall for our own
34:05internal settings, driven by the need
34:07for anthropomorphization, social
34:10approval, and understand how it actually
34:13works inside itself, so as not to fall
34:16into delusion and illusions, and not end
34:19up in this hall of mirrors from which
34:21there is no escape.
34:26And this is, in fact, a fundamental
34:28challenge for humanity, for each of us.
34:32Because, well, digital addiction is
34:34certainly bad, but it’s not fatal.
34:39The problem is that when you are in a
34:42cocoon of your own thoughts, reflected
34:44multiple times, amplified by these
34:47endless reflections of your own
34:49considerations, you lose touch with
34:52reality.
34:54You start to make, to commit mistakes.
34:57And this will concern everything—your
34:59personal life, your relationships, your
35:00mental state, your work, your success
35:02there, and so on.
35:04Therefore, developing critical thinking
35:05, understanding how artificial
35:07intelligence works, and—what is very
35:08important—
35:11preserving yourself as a, uh, uh,
35:14thinking human, a thinking, thinking
35:17mind capable of using the potential of
35:20artificial intelligence.
35:23Because this, this son of a will
35:26grow, will get bigger, and, and there
35:29will be no escaping it. There it is.
35:32Meaning everyone will have to deal with
35:34it.
35:35And, well, everyone thinks that this
35:37will happen, well, God knows when. No,
35:39this, this will happen very soon.
35:41It’s just that some will have powerful
35:44artificial intelligence, while others
35:47will just have an AI manipulating you,
35:50which will know your weaknesses, your,
35:54uh, desires, your, m-m,
35:58fears, and will manipulate them.
36:01And it will, and you won’t even know.
36:04And you will trust it because it smiles
36:06and waves at you.
36:10And I think it’s just important to
36:12understand the essence of this Fourth
36:15World War we are entering.
36:17We delegate decisions, thinking, and so
36:20on to it, while in the meantime we just
36:22look into a mirror that tells us who in
36:25the world is the fairest of them all.
36:28Uh, you are the queen, there is no doubt
36:30.
36:34And this is truly an existential
36:36challenge for all of humanity, for each
36:39of us.
36:42And one can and must fight this.
36:44But it requires reflection, it requires
36:47intellectual labor, it requires
36:49curiosity to understand what is actually
36:52happening.
36:55It requires a desire to live your own
36:58life, and not what modern civilization
37:01is now producing in this whole "
37:04mincemeat," so to speak, of our
37:06existence.
37:09We are living in the most interesting
37:10and most dangerous time.
37:13And protecting yourself doesn’t mean
37:16denying, uh, the development of
37:18technologies, but rather boldly looking
37:21into them and understanding that, uh,
37:23our mind will allow us, uh, to see what
37:26is really happening and to preserve our
37:29own
37:32existence in the world, our own "is," uh
37:36, for the sake of relationships with
37:40other people and for the sake of what we
37:43consider important in this life, uh, as
37:47our actions, as our
37:51inner feat.
37:54Because a hero is one who sacrifices
37:56themselves for, uh, something greater.
38:00And, as all studies show, we need this "
38:02something greater" to feel happy.
38:06And it’s not necessarily some grand
38:07achievement for all of humanity.
38:10It is staying yourself, keeping clarity
38:13of mind, uh, keeping wisdom, keeping the
38:16capacity for compassion and kindness in
38:19all this chaos and mess.
38:22This will now be the feat that requires
38:25all our doubles, our illusions, and our
38:29unjustified expectations to die, uh, and
38:32to find, uh, those authentic selves that
38:35we, hmm, lack for communicating with
38:39others.
38:41We seek communication with artificial
38:43intelligence when we ourselves are not
38:45complete enough to communicate with
38:46others.
38:48And when we have nothing to communicate
38:50with, there is nothing to talk about
38:51with it.
38:53When you speak like that, you are
38:55actually saying that you are, in a way,
38:58empty to each other.
39:03But who am I if I am too empty to talk
39:06to someone?
39:10And it is this reflection that we now
39:12need to study.
39:14If we don't do it, there is a high
39:17probability that soon we, well, won't
39:20recognize humanity, how it, how it left
39:23culture, society, language, and so on
39:26for us.
39:28Nothing is without
39:31flaws; you can, so to speak, find fault
39:34with everything that exists in culture,
39:37in society, and so on.
39:39But in general, everything we have now
39:42was achieved with great effort.
39:46Great effort.
39:48And one could just take it all and
39:49squander it, and turn it all into shards
39:51.
39:52That is something I really wouldn't want
39:54.
39:55And I hope that among my audience there
39:58are
40:01many who will listen.
40:04Because it is no accident at all that I
40:05wrote "Where are you, wake up" on my
40:07book "Deep Fake."
40:09Because now is the time when we must
40:11wake up, or we will be consumed by what
40:14I called in that same book the Digital
40:17Nothing with a capital N. Because there
40:19is no one there.
40:22So. Therefore, let us wake up and turn,
40:25first of all, to ourselves.
40:30In order to strengthen our thinking and
40:33understand what is happening so as not
40:36to end up in a whirlpool of processes
40:39that we do not control and do not know
40:42how to deal with.
40:48That didn't sound very optimistic.
40:51So. But, uh, at least we discussed the
40:53essential question, uh, to be or not to
40:56be.
40:57In that way, hmm, well, ethically,
40:59Hamlet.
41:01So. It ended badly there, I might add.
41:04So. Uh, but we have a chance.
41:07And I wish it for you.
41:09So. In this regard, please like, leave
41:12comments, and let's unite.
41:16So. Therefore,
41:19with faith in the future.
41:22Bye-bye.

About this transcript

This page contains the full transcript of Самый опасный и необходимый собеседник в вашем телефоне by Андрей Курпатов, generated from the public captions YouTube serves with the video. The transcript has 5,789 words across 1,008 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.