7 Đẳng Cấp của AI | Trí tuệ nhân tạo tương lai (Thấp nhất tới Cao nhất)

00:00

In 2016 in an interview on BBC,

00:03

Professor Stephen Hawking was a prominent physicist. He was instrumental in decoding the mysteries of the universe.

00:10

He warned that AI - artificial intelligence could be the worst event in the history of human civilization.

00:16

He believes that in the future AI has the potential to surpass humans.

00:20

When it can design itself, develop itself, it evolves rapidly,

00:25

with its superior mechanical capabilities. While humans must evolve through biological choices.

00:29

It took many years to keep up with this process.

00:33

[Interview]

00:53

One year later, in 2017 at a Conference in Lisbon, Portugal,

00:57

Elon Musk - a technology billionaire warned that: "AI can be more dangerous than nuclear weapons and called for its use."

01:05

AI is responsible. And it is forbidden to combine AI with weapons."

01:10

Elon Musk believes that humans overcome everything and control this world.

01:13

Because we are smarter than every other species. But now, for the first time in the history of the world,

01:19

there is something smarter than us, smarter than humans.

01:22

It is smarter than its creator. And for the first time, we are facing such a thing.

01:27

Never in the history of millions of years have humans begun their journey to conquer the world.

01:31

Most recently, in 2023 Professor Geoffrey Hinton is the godfather of AI,

01:34

a pioneering scientist in the field of deep learning.

01:39

He left Google to talk about technology companies racing to develop AI.

01:43

Making it possible for it to get out of control.

01:46

And it is in this year 2024, as Professor Geoffrey Hinton mentioned.

01:50

Fake images, videos and information created by AI start the risk of getting out of control.

01:55

In the near future, he believes that AI will begin to be able to write its own code and develop itself.

02:00

And killer robots will be born. AI can evolve in ways that far exceed our wildest

02:05

imaginations . And it will dramatically change our daily lives,

02:09

faster than we think.

02:10

This video will explore the 7 stages of artificial intelligence, from lowest to highest.

02:15

Remember to subscribe to the channel and like this video before continuing to watch.

02:20

Level 1: Reactive AI (Reactive Machines)

02:23

Reactive AI systems were first developed in the 1960s.

02:27

This type of AI only reacts to current stimuli, current commands.

02:31

It has no ability to remember, or predict. And its learning ability is very little.

02:35

The most popular applications are: gaming, chess playing system, checkers.

02:39

Or it can be used for medical diagnosis, supporting disease diagnosis based on the patient's symptoms.

02:43

Or automatic control such as controlling robots and machines in factories in a simple way.

02:47

Sometimes this type of AI is called "knowledge-based systems". It works without relying on intuition, without relying on learning.

02:53

Which is based on pre-determined rules, principles and commands.

02:57

So it has no ability to adapt.

02:59

For example: alarm clock devices, or temperature regulators based on principles.

03:03

In these AI examples, at 7:00 a.m an alarm clock automatically sounds

03:08

based on the previous time setting principle.

03:10

If the room temperature rises to 24°C, the thermostat will automatically turn on the air conditioner,

03:15

based on the temperature sensor connected to the air conditioner control.

03:17

Or the microwave or car radio. But this type of simple chatbot also uses rule-based artificial intelligence.

03:23

We see this type a lot in everyday life.

03:26

Up to level 2 is: Context-based AI, also known as AI with limited memory.

03:32

So that context-based artificial intelligence system, it doesn't just process the inputs immediately.

03:37

Because it also takes into account the surrounding environment, takes into account user behavior, takes into account historical data.

03:41

To make the best decisions.

03:42

Its development began in the 1970s until today.

03:47

Examples are voice recognition systems,

03:48

machine translation systems and data-based weather prediction systems.

03:52

In particular, the system that we are very familiar with is the product recommendation system based on purchase history.

03:57

And user browsing behavior. For example, you buy products on Shopee, on Tiki, on Lazada.

04:02

When you buy a banana, e-commerce sites immediately

04:06

rely on your history and your behavior.

04:08

Say, "Ah, this guy likes to eat bananas so I suggest he eat apples, oranges, and similar fruits."

04:13

Or if you buy bananas, they will suggest banana combos.

04:16

The product types related to that product. These are the things that search websites, social networking platforms,

04:20

Google, Facebook, Shopee do to increase purchasing ability and attract users.

04:24

Based on this context, AI will remember some information about your past.

04:28

Latest information about our surrounding situation,

04:31

to make decisions. And it is somewhat capable of predicting our future behavior.

04:35

And its learning ability is better than reactive AI.

04:37

However, it is also limited in processing complex information.

04:40

And it adapts to new environments less than context-based AIs.

04:44

When we ask for a pie recipe it will suggest a nearby store to buy the ingredients.

04:48

Rely on Google Map, based on information and distance traveled.

04:51

And the most popular example of this type of AI is the retention system on e-commerce websites,

04:56

news websites and social networks that we are very familiar with.

04:59

This system stores our information. It stores browsing history, stores purchase history.

05:04

Store the items we've looked at, we've looked at.

05:07

And it hints at our personal purchases.

05:09

We don't need to buy products, we just watch a video longer, we watch a video longer.

05:15

We don't like, we don't drop interactions. But when we look at a certain girl, the hot girl stays on the screen longer.

05:20

Immediately, Facebook's system or the system of websites like Instagram and TikTok,

05:24

they already know that we like and care about that.

05:26

And it will suggest a lot of similar videos for us to watch,

05:30

so we can get what we like.

05:31

What we secretly like, without us having to tell the machine

05:35

"we like this girl". And that's why it's so developed,

05:38

making social networking platforms more and more powerful. This, in turn,

05:41

creates an extremely large competitive advantage for current social networking platforms and websites.

05:45

Because the more information a website has about you, the more they will understand you. And they will bring you more

05:50

valuable and relevant information. Products that make you like them, you stay with them more.

05:54

Instead of coming to a new website,

05:56

you haven't done anything yet so people don't know what you like.

05:59

And when you don't see much value in that new site,

06:01

you'll go back to the old site. And that's why the current old websites,

06:05

the more user information they have, the stronger and smarter AI they will have.

06:09

And there are more and more users. It's called the user loop, the business data loop.

06:14

This type is also very common, we see it a lot now.

06:16

Third Level: AI specialized in each field - Narrow AI (ANI).

06:21

The concept of narrow AI has been around for a long time, since 1950 until now.

06:25

But in the last 30 years, it has been widely applied in life.

06:29

Its characteristic is that it only focuses on a few very specific areas.

06:31

Has the ability to learn and solve problems in that area, much better than reactive AI.

06:36

And context-based AI is of the above types. And very accurate in the professional field.

06:40

For example: virtual assistants - Apple's Siri, Google Assistant or Alexa.

06:45

These are types of AI about narrow artificial intelligence.

06:47

In the field of helping us humans, easily use technology solutions and find information.

06:52

In the world of gaming, DeepMind's AlphaGo is the AI ​​that beat Go players.

06:57

Is a very outstanding example of how to conquer complex games that

07:00

require very high strategy and intuition.

07:02

Before that, Go was an ancient board game, known for its extremely large number of moves.

07:07

And diverse deep strategies, are an extremely difficult challenge for artificial intelligence.

07:11

But AlphaGo not only learned how to play, it also beat

07:14

all the world champions. Currently, artificial intelligence is in a narrow scope

07:18

can provide translation, through real time very quickly.

07:21

And it will develop very quickly, so that we no longer have to learn foreign languages ​​when going abroad.

07:25

For example, we can put on Google headphones, put on Apple headphones.

07:28

And when the person next to us speaks English, we only understand Vietnamese. We speak Vietnamese, they speak English.

07:33

But immediately, Siri will translate directly, in real time,

07:36

so we can hear what they say in Vietnamese. And they can hear what we say in English.

07:40

And they also use Siri, we use Siri and immediately understand each other.

07:44

Or they use Alexa, we use Siri. They use Google Assistant, we use Siri.

07:48

But those AIs can talk to each other and help us understand.

07:52

And this is an extremely wonderful thing. It helps us to go to strange countries and

07:55

strange languages ​​that we have not yet learned. But I feel extremely confident.

07:57

Or IBM's medical diagnosis system,

08:00

which is the Watson system used in the medical field.

08:02

Then it demonstrates its great capabilities, in the field of big data analysis, to support medical professionals.

08:06

Another famous example is, a facial recognition system is a system that identifies

08:11

people with facial images.

08:12

As in China and some other countries, these surveillance cameras are combined with AI,

08:16

which analyzes billions of photos and videos every second. It lets the Government know where you are and what you are doing.

08:23

Who is participating in moving on the road? Similarly, in our financial world, our investment world,

08:28

narrow-scope artificial intelligence, narrow AI can monitor the stock market in real time.

08:33

Read a lot of information, analyze a lot of transaction patterns.

08:36

And predict stock movements with relatively high accuracy, surpassing traders.

08:40

Such artificial intelligence systems do more than simply calculate numbers.

08:43

About ten years ago I made such a system.

08:45

But back then there was no AI and just using data, only using models we created ourselves,

08:49

was very limited.

08:51

But now AI can detect suitable patterns on its own.

08:53

Self-discover patterns in billions of different types of information.

08:56

And use complex algorithms, fine-tuned through thousands of different data sets.

09:00

And it will provide short-term forecasts for better trading.

09:03

That's in our field. And I think that in the near future there will be stock billionaires

09:07

who are among the traders supported by AI. What about value investing, long-term investing,

09:12

company analysis, business evaluation?

09:14

We still have an advantage but in the near future when computers

09:17

are learning faster and faster. The system can read more data and

09:20

grasp more information. And the possibility that it will surpass our ability

09:23

to trade short-term. That means buy - sell, Long - Short every day.

09:27

And to prepare for that, Owl's investment team is also researching this field.

09:30

To apply AI to support us in deciding to collect and filter information.

09:33

Give signals to help you trade more effectively. My common method

09:37

is to use 80% of available principles and tools. Then put in knowledge, put in data of Vietnamese stocks,

09:42

And teach AI to learn gradually teach it like a child, teach gradually in the future.

09:46

As it learns more and more skills, more and more knowledge about itself and the market.

09:50

Combined with the ability to read and understand large data, it will be better than me.

09:53

Why do I do that? So I want to maintain long-term investment efficiency

09:56

of about 20%/year. It will gradually become better than me in the future,

10:00

helping me a lot in investing and trading to be highly effective.

10:04

To maintain long-term investment efficiency of about 20%/year is extremely difficult.

10:08

Everyone can eat and make a profit when the market is uptrend and the market price increases.

10:12

But how to manage portfolio and capital so as not to suffer big losses.

10:15

When the market is bad, it's something that very few people can do.

10:19

That is the reason that Dr. Lan - a specialist doctor withdrew after investing for more than 2 years.

10:24

into the stock market. I've studied all the courses and followed all kinds of rooms and keys.

10:28

but still at a loss. Because there is not enough time to trade

10:30

like a professional investor. And Lan chose to entrust the job of investing in assets and

10:34

increasing her passive income to Owl's professional SStock team.

10:37

Lan said: "I want to spend more time on my job as a doctor,

10:41

on treating patients, which I love. I want my income at the hospital to increase, my knowledge to increase.

10:46

What about investment?" , I entrust Mr. Owl's team with peace of mind."

10:49

And with a track record of being at the top of the market with an average profit of 20% for many years

10:52

Lan has very good results. If you are interested, you can text Cu at this phone number,

10:56

or inbox Owl. OK.

10:58

Back to level 4: Generative AI, is something that has become popular in the past 2 years.

11:03

Generative AI focuses on creating new data such as: new answers, stories, new poems.

11:09

New videos and new images are based on learning and synthesizing from billions of data types on the Internet.

11:15

It uses techniques such as machine learning and artificial neural networks

11:18

to learn these data patterns. Currently, AIs of this type are OpenAI's GPT chat,

11:23

then Google's Germany. Then there are the image generation systems: Midjourney, Dall-E 2, Stable Diffusion.

11:28

It learns and mixes billions of different photos to create a new photo.

11:32

For example, you can tell it to create a girl wearing a swimsuit,

11:35

but walking on Mars in a very natural and beautiful way.

11:38

Or you can tell him to write poems and stories through Chat GPT.

11:42

Or you tell AI to compose music through the MuseNet system, or the Deepfake video creation system

11:47

of any famous person.

11:49

[Interview]

11:58

Its benefit is that it helps people create new things, it increases efficiency in many fields.

12:03

And increase access to information for everyone.

12:05

But it creates the problems that Professor Geoffrey Hinton talked about above.

12:09

It is an ethical issue in using AI to create fake information and fake videos.

12:13

And created extremely quickly, extremely much and can distort our will, distort our information,

12:18

distort our view of the world. That's the risk of abuse and its control.

12:23

The 5th level is: Reasoning AI. This type of intelligence can simulate the complex thinking processes

12:28

that humans use every day.

12:29

It doesn't just process data, it also analyzes data. Connect patterns and identify

12:33

anomalies and draw logical conclusions. It is developing together with the above generative AI.

12:37

It's like we give AI a puzzle and it discovers the best ways to put the pieces together.

12:42

And clarify unclear paths.

12:44

While our brains can only process a finite type of data.

12:47

For example, we read a page of a book, it takes us 1 minute.

12:49

But AI can read 1 million pages of books in just 1 minute and synthesize all such things together.

12:54

OpenAI's GPT Chat or Google's Gemini

12:57

are currently adding theoretical AI technologies. These large language models

13:01

are trained from millions of texts on the Internet. And such advanced versions.

13:05

When artificial AI is applied, it is likely to surpass our reasoning abilities.

13:10

Works many times faster than us.

13:12

For example, now AI can begin to code software on its own,

13:15

replacing engineers. Or self-driving cars are another example of theoretical AI.

13:20

It makes its own decisions whether to turn right, to turn left, to avoid this car,

13:24

to avoid that baby, to make the best decision. Reasoning AI and generative AI are things that humanity is developing.

13:30

And in the near future,

13:32

it will reach level 6: Artificial General Intelligence (AGI).

13:37

This is often considered the final destination of AI development.

13:40

And humans play an important role in it. Because when AGI comes, when artificial general intelligence comes,

13:46

it can do any task that a human can do on its own.

13:50

And it doesn't need our intervention anymore. It does not need our help in developing itself.

13:55

And it will write its own code, it will develop itself and it will design itself to be most effective.

14:00

With such tremendous flexibility, it means you can teach it almost anything.

14:03

Just like you teach an adult.

14:05

Except that it learns millions of times faster than us.

14:08

At that time, we cannot control and cannot reverse the evolutionary process of AI anymore.

14:12

And people often call it the Singularity point, a point that breaks out

14:15

a completely new world and we have never known anything about it.

14:18

There are so many potential things that could open up. Things that took us humans millions of years to evolve,

14:23

AI can do in just a few years. And far beyond what we think.

14:27

With the emergence of AGI, daily life will change greatly.

14:31

For example, open your eyes, a virtual assistant doesn't just report the weather.

14:35

It will play melodious music for you to wake up early. It understands whether your mood is happy or sad this morning.

14:40

It helps you plan your day, gives suggestions for analysis and reports to your boss.

14:44

Or even assist in cooking for you.

14:46

Leave the robot in the kitchen and show it the recipes.

14:49

Or interfaces in the future when we can combine the brain with the machine to a mature degree.

14:54

Then humans can combine these types of artificial intelligence and communicate with them

14:58

in the real world using our thoughts.

15:00

We can enhance the brain, enhance the body's abilities by using AI AGI in the brain.

15:06

And we can even receive guidance from these AGIs,

15:10

in the form of thoughts, in the form of feelings, in the form of text. And images that only we know.

15:16

And if we equip AGI with a robot body like Terminator or future robots.

15:21

The possibilities could be extraordinary.

15:23

At that time, AI can help us overcome difficult and dangerous terrain.

15:26

Assist in rescues and perform surgeries for us.

15:29

Or even participate in art by painting or sculpting.

15:32

And scientists are predicting that AGI may appear in the next few decades.

15:37

The most optimistic forecast might be 2045 which is 21 years from now.

15:42

7th level: Super intelligent artificial intelligence.

15:45

Right after the emergence of AGI artificial intelligence, comes artificial general intelligence. Then, a few decades or hundreds of years later,

15:51

this AGI will improve itself, it will self-evolve, it will adapt and it will no longer need human intervention.

15:56

And it will become super artificial intelligence, Super AGI.

16:00

At that time, the kind of technology was as fantastic as it is now. For example, quantum computers, the things we are researching,

16:04

will be perfected by AGI and become popular.

16:07

Other advanced technologies such as fusion reactors.

16:10

Space technologies such as the Dyson sphere to cover the sun and use all of its energy.

16:15

Imagine today that the Earth only receives 0.00005% of the sun's energy

16:21

and humans are only using 2% of the total energy that the sun shines on the Earth.

16:26

When super intelligent artificial intelligence begins to emerge, it will be able to do that.

16:31

Able to use all that solar energy.

16:33

And then it will turn into super intelligent entities which we cannot understand.

16:36

It may have superior intelligence even with our existing consciousness.

16:40

And it will solve problems that we cannot solve.

16:42

For example, poverty, Mr. Ray Kurzwwell is an American futurist,

16:46

inventor and computer scientist.

16:48

He is very famous for his predictions about the future of technology.

16:51

He is the one who made the prediction about general artificial intelligence,

16:54

which will be formed by 2045. And he believes that by the end of the 21st century, in more than 70 years,

16:59

these artificial intelligences will be available could be trillions of times smarter than everyone else.

17:04

And with such intelligence, the speed of innovation will be terrible and unimaginable.

17:08

It's like we're compressing 20,000 years of technological progress into a single century.

17:12

And it's that potential that impresses Ray.

17:14

And the ability of artificial intelligence to form, it will form new structures,

17:18

both of the Government and of the world. And automation goes far beyond what we have today.

17:22

Even a super intelligent artificial intelligence will one day use quantum algorithms

17:27

to simulate human consciousness.

17:29

And then this artificial intelligence is capable of self-awareness, awareness of its own existence.

17:34

And its relationship with the outside world.

17:36

That artificial intelligence may have emotions, or full senses, and go beyond what we are currently experiencing.

17:41

And it becomes an organism, a self-evolving organism.

17:44

It will see humanity as a different kind of creature.

17:46

And it chooses in the direction of evolution, in ways that we cannot understand and cannot control.

17:50

Or a superior artificial intelligence that could potentially create new forms of life.

17:53

Can be organic living organisms, or living organisms created from Nanobots.

17:57

That is, the blood vessel cells, molecules, and atoms in our body

18:00

are replaced by tiny robots.

18:02

This artificial intelligence, it can carry these Nanobots, these tiny robots,

18:05

all over the Earth. And it can control and repair our ecosystem.

18:09

Transform our planet in the way that is best for it and best for us too,

18:13

if we take control of it now. And after improving the Earth, after using the energy

18:18

of the solar system, these artificial super intelligences can continue to advance further.

18:22

By conquering star systems and galaxies together with humans.

18:26

And it has the ability to solve cosmic mysteries, from the purpose of Black Holes.

18:29

To the nature of dark energy matter. Or understand the dimensions that we

18:34

are thinking about in science fiction. Or let us exploit new dimensions.

18:38

And take advantage of them to travel and travel beyond time and beyond light.

18:42

Talking about these things it seems a bit crazy and a bit far-fetched.

18:44

However, it is the science of the future.

18:47

To distinguish this difference, imagine an example of an ant society,

18:51

in an ant nest. As for ants, they are very tightly organized

18:54

and divided into many different classes. The queen ant is the only ant in the nest that can reproduce.

19:00

It lays eggs to maintain the number of ants in the nest. It is protected and cared for by worker ants.

19:04

Worker ants are responsible for finding food, building the nest, and protecting the nest.

19:08

Soldier ants protect the nest from enemies, to attack and destroy other insects.

19:12

Male ants only appear during the breeding season and are responsible for mating with the Queen ant.

19:16

Young ants are the larvae of ants. Winged ants are adult ants with wings. The

19:21

social organization of ants is a very successful example in nature.

19:25

It has existed for more than 100 million years and multiplied into 12,000 species across the Earth.

19:30

However, ants cannot understand a human city.

19:33

How does human society operate? What are the streets for?

19:36

What are supermarkets, courts, parks, and families for? To ants, we humans are the ultimate powerful gods.

19:42

Whatever ants think about humans, humans will also think about artificial super intelligence.

19:47

We will not be able to understand because our intelligence at that time

19:51

is much lower than theirs. Unless we evolve with AI and we incorporate those technologies

19:56

into our biological bodies.

19:58

To have those powers of evolution, learning and replication.

20:02

Okay, now it's your turn. Remember that we are at levels 4 and 5.

20:06

That is, we are at the point where AI creates and AI reasons. The earliest stage 6 in the forecast is 2045.

20:12

Stage 7 is very far away. But what is your opinion?

20:15

What have you done to prepare for that future? Please let Owl know by commenting below this video.

20:19

Good bye. Wishing you good health and see you again.