ИИ делает умных умнее, а глупых глупее | Что нас ждёт дальше?

0:00Marglalan Kolevich , I have a question .

0:02Artificial intelligence is entering

0:04business right now , becoming deeply

0:06integrated , but the question is : what

0:07will customers value in 5 to 7 years ?

0:10That is , once it’s all fully

0:11integrated ? What is it that we , as

0:13entrepreneurs , need to be working on

0:15right now ?

0:18You know , you say " customers , " but

0:20perhaps we could frame the question

0:22more broadly : what will we value in

0:25artificial intelligence ? Or , we could

0:29frame it even broader . Most people make

0:32the mistake of thinking that artificial

0:35intelligence will take us over . Movies

0:40have really scared us into thinking

0:42they’ll seize control and so on . But

0:46you know , the funny thing is , AI won't

0:49conquer us — we will hand it everything

0:52voluntarily and with joy . Why ? Because

0:56AI and robots will do everything

0:59cheaper , faster , easier , and with

1:02higher quality . If you have a robot

1:06housekeeper that does everything

1:07flawlessly , follows orders , and doesn't

1:10need reminders , what will happen ? What

1:12will you do ? You will trust it with

1:15more and more tasks . So , the better the

1:18AI gets , the more access to your

1:20computer you’ll grant it . You’ll

1:23give it access to your hard drive ,

1:25Google Drive , iCloud , your calendar ,

1:27and your email . You will offer that

1:30access yourself because it’s cheaper ,

1:32easier , better , and more efficient . And

1:35we will keep giving it that access as

1:37long as AI remains more efficient than

1:39humans . We’ll actually enjoy it . We

1:42will happily surrender our future and

1:45our fate to artificial intelligence .

1:49And when we’ve handed it all over ,

1:51one day we’ll wake up in horror ,

1:54realizing we can't even imagine life

1:57without it . And if we go back to your

2:02question about customers , they will

2:05always value what is faster , cheaper ,

2:07higher quality , and easier . Jeff Bezos

2:11put it well . He said : " Many people

2:14predict the future and ask me : ' On what

2:17basis do you make your plans ? ' " He says

2:20: " Actually , I don't base my plans on

2:22predictions . " " I base my plans on what

2:26will never change . " And what will

2:28always stay the same ? The fact that

2:30people will always want things cheaper ,

2:32faster , and of better quality .

2:36Accordingly , Amazon’s entire strategy

2:38is to provide exactly that : cheaper ,

2:39faster , and better . And customers , they

2:43absolutely don't care about the

2:45internal workings of our businesses .

2:50They just want to place an order and

2:51get it , that’s all . They don't really

2:55care about what we do , what our

2:57processes are , or how many factories we

2:59have . And , accordingly , if artificial

3:03intelligence helps us , the customers ,

3:05to deliver goods or services faster ,

3:07cheaper , better , and so on , that’s it

3:09, those companies will win . Accordingly

3:14, the only companies that will win are

3:16those that move most of their

3:17operations to artificial intelligence

3:19faster .

3:22Well , over time , it will be harder for

3:23us to do that , because artificial

3:24intelligence is now being integrated

3:26not only into business , but also into

3:27education . And our children are using

3:30it too , yes , which essentially takes

3:32away the opportunity for logical

3:34thinking . And further on , there will be

3:36fewer and fewer people every year who

3:38know how to think logically and ,

3:40accordingly , manage this system .

3:42Well , not exactly , actually . Look ,

3:45every generation thinks that children

3:48are getting dumber . Every generation .

3:51Our parents thought so , our parents '

3:54parents thought so . Why ? Because , yes ,

3:57for our parents ' tasks , we are dumb ,

4:00but for our tasks , our parents are dumb

4:02. The same will happen with our

4:04children . We talk about logic , this and

4:08that , our kids don't understand , are

4:10atrophying , and so on , but they

4:12understand things that we don't even

4:14get , like , what's that about . I mean , I

4:19just look , not many young guys , they

4:22understand such things , and for them ,

4:25it's natural . I mean , they didn't study

4:30it like we do — like I'm studying

4:32vibecoding , I'm really studying it ,

4:34taking topics , studying concepts , what

4:37it means , the history of the term , what

4:39it's for , how it helps . It's just

4:43natural for them , like breathing . And

4:46you see , many things that we —

4:47conditionally speaking , if you take

4:49intelligence and break it down into

4:51parameters and criteria — we have our

4:53own parameters and criteria that show

4:55us that a person is very smart .

4:57Uh-huh .

4:59Well , these parameters and criteria

5:01don't fit these kids because they are

5:03preparing for a different future . We

5:05were preparing for a different future .

5:07We as parents are always like generals

5:09preparing for the last war . And we are

5:12doing the same . We are preparing

5:15children for our ...

5:18for our past .

5:20We say things like : " Damn , I don't know

5:22English , I need to put my kid in

5:23English classes . " Damn , China is

5:25booming now , I need to put them in

5:26Chinese classes . Damn , I need to sign

5:29them up for programming , damn , they

5:30need to know math . Why ? Because these

5:33are the things that were missing for us

5:35, that we lacked . Accordingly , we are

5:37overloading the child , but the child

5:38has a different future . And maybe what

5:41they actually need is drawing , dancing ,

5:44art . We are not going through math or

5:47languages at all , because now with LLMs

5:49, you just wear a little device , talk ,

5:52and it translates into Chinese

5:53immediately . Why would you study

5:55Chinese if you can just put on a little

5:57device and talk to any Chinese person ?

6:00Accordingly , our requirements to know

6:02Chinese , English , or math and logic

6:04don't work ,

6:06they are not relevant . Now we are

6:08learning , we grew up , I even studied

6:11logic myself at law school . We studied

6:15linear logic , but the current world is

6:18moving toward non-linear logic . And

6:21that is something completely different .

6:24Both linear and non-linear logic are

6:26contradictory concepts . It’s like how

6:29I was lucky back in the day . I had a

6:32math teacher who graduated from , uh ,

6:34some Moscow mathematics institute . Well

6:38, he was a bit of an alcoholic , but

6:40very talented . All the kids loved him ,

6:43parents loved him , and so on . And he

6:45taught us math , not just math , but

6:48everything beyond math as well . And in

6:51this way , he broadened our horizons . I

6:54was amazed when he talked about

6:55non-Euclidean geometry , or

6:57Lobachevsky's geometry , so to speak . We

7:00know in ordinary geometry that two

7:02parallel lines do not intersect . But in

7:05Lobachevsky's or non-Euclidean geometry

7:07, they do intersect , and even twice ,

7:10two parallel lines . Furthermore , we

7:14know that in any square , all sides are

7:16equal , but in that geometry , they are

7:19not . We know that in our ordinary

7:23linear geometry , the shortest distance

7:26between two points is a straight line ,

7:28but in non-Euclidean geometry , the

7:30shortest distance is a curve . And how

7:33does that work ? Take a globe and look

7:36at how airplanes fly . Here is your city

7:39, here is your city . Just fly straight .

7:41Why are you flying like that over the

7:44Arctic ? You look , and airplanes are

7:46flying in some strange way .

7:47And in aviation , this is called an

7:49orthodrome . Orthodromy , when you map a

7:53route on a sphere , you always fly along

7:55a curve , because you can't fly straight

7:58, because along a curve , in short ,

8:01now we're going to end up in the flat

8:02earth theory . We're about to take a

8:04wrong turn here .

8:06My point is that this linear logic we

8:09live in doesn't work in this future , or

8:11it doesn't work the way we think .

8:15Therefore , our assessment of our

8:17children's intelligence through our

8:19parameters will always be flawed .

8:23Therefore , as usual , we have to grumble

8:26, complain about children , say a stupid

8:28generation is growing up , and where is

8:30the world heading ?

8:32Where is the world heading ? How will

8:34they feed us , how will they serve us a

8:36glass of tea ? Yes , yes , yes . Where is

8:39the world heading , with such an

8:40education we are all doomed , and so on .

8:43Well , nothing to worry about . Actually ,

8:47look , back in our day , our parents

8:49focused on things like memorizing the

8:51multiplication table , for example . That

8:54used to be considered really cool and

8:55very important . But if you tell someone

8:59that now , well , it’s just laughable .

9:02Memorizing the multiplication table

9:04when you have a calculator at hand , and

9:06so on . There’s really no need for

9:09that at all . There’s no need to know

9:10languages . There’s no need to know

9:12how to use a calculator . There’s no

9:14need to know a whole bunch of other

9:15things either .

9:17Margulan Kalievich , isn't this the very

9:20moment that forces the brain to move ,

9:22to actually start working at least a

9:24little bit ? I mean , those skills should

9:28be replaced by some other ones . But

9:31nowadays , when you look around , people

9:33usually have a phone in their hands ,

9:35not a book . I mean , like

9:37Yes . Well , look , the danger of our

9:40current time is actually that there

9:43will be a sharp stratification of

9:46people based on intelligence ,

9:49not based on wealth , not based on money

9:51, but on intelligence . People will be

9:53divided into two camps . One camp of

9:55thinking people and another camp of

9:57non-thinking people , and unfortunately ,

9:59the non-thinkers are the majority . And

10:02the gap between the thinkers and the

10:03non-thinkers will be catastrophic . It

10:07is growing . Why ? Because artificial

10:09intelligence is like a magnifying glass

10:11. It makes the dull even duller , and

10:15the smart even smarter .

10:16It makes them even smarter ,

10:18because it has built-in algorithms . The

10:20dumber the questions you ask , the

10:22dumber the answers it starts to give

10:23you . It starts answering more simply

10:26because it has a mechanism to adapt to

10:28the user .

10:29Consequently , it starts explaining

10:31everything to you even more simply , and

10:33in this way , you become even duller .

10:36Because you have no need to ask

10:37different questions . It gives you

10:39ready-made , templated answers . But when

10:42a person is smart , they start asking

10:43critical questions , deeper questions ,

10:45questions on a different level , and so

10:47on . The intelligence starts to adapt

10:49and answer more intelligently , and it

10:51starts building a conversation with you

10:53in a completely different way . And thus

10:56, even at the level of a simple ChatGPT

10:59— I'm not even talking about coding ,

11:01web coding , agents , and so on — even at

11:04this level , stratification occurs . And ,

11:08as I said , I believe the era of

11:10intelligence is upon us . What does

11:13artificial intelligence actually

11:15provide ? It poses a challenge to

11:18natural intelligence . And the task of

11:23who will control the future depends

11:25solely on who turns their brain on

11:28faster . I mean , artificial intelligence

11:32will outpace us in all routine

11:34operations , in all massive operations

11:37where speed , accuracy , data volume ,

11:39data processing , and so on are needed .

11:43But artificial intelligence , I think ,

11:45won't be able to build non-linear

11:48connections and dependencies for a very

11:50long time , because you could , for

11:52example , say : " We're sitting here at

11:55Lake Balkhash , we have a Russian olive

11:57tree here , right , its scientific name

11:59is Elaeagnus commutata . " Silverberry .

12:02You softened it , like you did , Margulan

12:04Koliche . I , I said it more softly , but

12:06the scientific name is actually

12:08silverberry , yes . And we can ask , what

12:12do this person here and a guy who went

12:15to a casino and blew a few thousand

12:18dollars have in common ? Yes , will

12:22artificial intelligence understand the

12:24difference ? It won't . It will say : "

12:26What does the tree , Lake Balkhash , and

12:27some dude who blew money in a casino

12:29have to do with each other ? " But we ,

12:31humans , immediately understand , " loh "

12:34narrow-leafed or silverberry and " loh " [

12:36sucker / dupe ] , a golden boy who blew his

12:39money — it's essentially the same thing

12:41. Well , we make these connections ,

12:44notice .

12:45The tree is innocent in all this ,

12:46right ? We aren't insulting the tree ,

12:49but for us , these are identical

12:51concepts , and artificial intelligence

12:52cannot link that yet . It's too

12:55non-linear . One is from botany , and the

12:58other is from folklore , from slang , you

13:00see ? And we , as humans , are the ones

13:04who can — there are layers , different

13:07layers of data or knowledge — we can

13:09build non-linear dependencies and

13:11connections , and non-linear means they

13:14come from chaos theory , they aren't

13:16subject to algorithmization , and

13:18artificial intelligence is just

13:20algorithms , understand ? And they don't

13:23yield to algorithmization . This is at

13:26the level that in English is called

13:28serendipity , or at the level of a gut

13:30feeling , intuition . And at the level of

13:35Chinese philosophy , it's called

13:37knowledge without words , when a person

13:40knows , but cannot explain . Uh-huh .

13:44Uh-huh .

13:46And have you looked at how we can only

13:47teach artificial intelligence , at most ,

13:49what we can explain ? We can't explain

13:53or teach it something we can't explain

13:55ourselves . But the problem is — or

13:58rather , not a problem , but our

14:00advantage is — that we know many things

14:02we cannot explain . By the way , wisdom

14:06belongs to this category . Wisdom is

14:09also a non-algorithmizable thing .

14:13Accordingly , there's no need to expect

14:14wisdom from artificial intelligence

14:16either . And wisdom is often more

14:18effective than science , than knowledge .

14:20Than knowledge ,

14:22yes .

14:23More like experience , probably .

14:25Experience is a lesson learned from a

14:29mistake made .

14:32And in that regard , you know ,

14:33artificial intelligence learns from its

14:35mistakes better . By the way , in terms

14:38of gathering experience , artificial

14:40intelligence is cooler than a human .

14:43But another question is that it

14:44extracts knowledge or experience from

14:47its mistakes in the form of algorithms .

14:50And what I am talking about is

14:51precisely non-linear connections . We

14:54can extract from experience not

14:57algorithms , but non-linear connections ,

15:00non-linear sequences , using non-linear

15:02logic that we never even studied , but

15:05have an internal sense for .

15:08Uh-huh . Let me give you , sorry , a

15:11simple example . Did anyone teach us

15:14about justice ?

15:16They didn't . Right ? Is there some

15:18algorithm for justice ? There isn't . But

15:20each of us understands inside what

15:22justice is . And it's hard to explain ,

15:25because we are born good . Well ,

15:27initially

15:28how

15:29we we are initially born good and kind .

15:33No , the question isn't — justice isn't

15:35just about kindness , and very often

15:37justice is cruel . That's something else

15:41. Virtue is something else . Here's

15:43justice . Where does it come from in us ,

15:46for example ? But we have it inside .

15:48Well , because our conscience bothers us

15:49, right , or how ?

15:49Ah ,

15:50because conscience bothers a person .

15:51Because of that .

15:52Well , the concept of conscience .

15:54Parents told us in childhood ,

15:55everything from childhood . Yes .

15:56But no . Look , there are people who grew

15:59up as orphans , but that doesn't mean

16:02they don't have a sense of justice .

16:05Society teaches us from birth , do you

16:07see , how did that happen ?

16:10Evolutionarily , we grew up , and all the

16:12unfair individuals did not survive

16:14because they were cast out of the tribe

16:16and called inadequate people .

16:18Uh-huh .

16:20If you just take , take a swing and slap

16:22Andrey on the head . We'll say : "

16:25Yaroslav is inadequate , we'll stop

16:27talking to you because you're

16:28inadequate . " And why inadequate ? What

16:30is an inadequate person ? An inadequate

16:32person is unpredictable . And if they

16:35are unpredictable , you can't do

16:36anything with them because you don't

16:38know their next move . Accordingly , such

16:41individuals were isolated from society

16:42and perished because they couldn't feed

16:44themselves ,

16:45couldn't survive .

16:46And , accordingly , such people did not

16:48produce offspring .

16:50Uh-huh . And so , historically , this

16:54sense of justice , the sense of mutual

16:57exchange , conscience , and so on , is in

17:00our genes . It's not at the intelligence

17:03level ; it's at the level of motor

17:05skills and genes . We feel , for example ,

17:09if we did something wrong — we don't

17:12even know yet that we did something

17:14wrong , but we feel uncomfortable , like

17:16we messed up . We feel it ,

17:19we feel it . And we can't even grasp or

17:22define the parameters . By what

17:23parameters did I mess up ? But I feel

17:26that I did something not very good .

17:31These things cannot be algorithmicized ,

17:32and we won't teach artificial

17:34intelligence to do this . M.