Intelligente, ma non in modo umano | Nello Cristianini | TEDxLakeComo

Summary

The video script discusses the concept of intelligence and its various forms, highlighting the limitations of human understanding when attempting to communicate with or conceive of alien intelligence. It recounts Carl Sagan's efforts to send a message into space that could be understood by any form of intelligence, using universal concepts like hydrogen molecules and prime numbers. The speaker points out that Sagan's approach was limited by his difficulty in imagining something truly alien. The script then explores the idea that intelligence is not exclusive to humans or even living beings, using examples like a bee seeking nectar and a plant manipulating the bee for pollination. It also touches on the rise of machine intelligence, such as algorithms used by platforms like YouTube, which learn from statistical data and user behavior. The speaker emphasizes the need to understand, regulate, and coexist with these forms of intelligence, advocating for a shift in perspective to recognize that human intelligence is just one type among many. The summary concludes with a call for a Copernican shift in our understanding of intelligence, suggesting that recognizing the diversity of intelligence is crucial for ethical regulation and coexistence.

Takeaways

• 🌌 The concept of intelligence is explored through Carl Sagan's attempt to communicate with extraterrestrial life, highlighting the challenge of conceiving truly alien intelligence.
• 📡 Sagan's use of the hydrogen molecule and mathematical principles in crafting a message for the Pioneer 10 probe was based on the assumption that all intelligence evolves under the same physical laws.
• 🐱 The example of a cat understanding complex human concepts questions the effectiveness of Sagan's message and suggests that even human intelligence varies widely.
• 🤔 Sagan's difficulty in imagining alien intelligence reflects a common human tendency to seek familiarity, which can limit our understanding of diverse forms of intelligence.
• 🌼 Intelligence is not a human signature; it existed on Earth before humans and can be found in non-human entities, including machines.
• 💻 Machine intelligence, such as that used by YouTube, is based on learning from mistakes and adjusting, similar to how a bee learns to reach its goals.
• 📊 The development of intelligent machines relies on three main components: preexisting data, statistical relationships, and constant user observation.
• 🔢 Statistics and data have become crucial in creating machine intelligence, often replacing the need for deeper theories.
• 🌐 The internet serves as a vast source of natural data that can be utilized to understand and predict user behavior for machine learning purposes.
• 🤖 Algorithms impact our lives significantly, from content recommendation to decision-making, raising concerns about their neutrality, honesty, and potential for misuse.
• ⚖️ There is a growing global awareness and discussion about the role and regulation of artificial intelligence, with lawmakers beginning to address these issues.
• 🌱 The comparison to the environmental crisis suggests that understanding and regulating AI is crucial to avoid past mistakes, and there is still time to act responsibly.

Q & A

What was the special message Carl Sagan was asked to write in 1972?

The special message Carl Sagan was asked to write was intended to be sent to outer space on the 'Pioneer 10' probe, which was to leave the solar system. The message had to be easily understood by any form of alien intelligence.

What principles did Carl Sagan and his colleagues decide to use in drafting the message for the 'Pioneer 10' probe?

They decided to use the hydrogen molecule, the position of the stars, and a set of theoretical mathematical principles, assuming that any form of intelligence in the universe would have evolved within the same physical and mathematical laws as we know them today.

Why did Sagan later prepare a message using the principles of prime numbers for alien intelligence?

Sagan prepared a message using the principles of prime numbers because he believed that such a fundamental mathematical concept could be a universal language, or 'lingua franca,' that any intelligent being could understand.

What does the speaker imply about the effectiveness of Sagan's messages in communicating with alien intelligence?

The speaker implies that Sagan's messages might not have been effective because they did not receive any feedback, suggesting that the messages may not have been comprehensible to alien intelligence, or that the assumptions about shared understanding of physical laws and mathematics were incorrect.

What was the mistake Carl Sagan made according to the speaker?

The mistake Carl Sagan made, according to the speaker, was that he had difficulty imagining something truly alien and conceived a message that was too human-centric, essentially looking for another version of himself rather than a truly different form of intelligence.

How does the speaker define an intelligent agent?

The speaker defines an intelligent agent as an entity with goals, capable of learning from mistakes, adjusting, and thinking. This includes not just humans and other living beings but also machines that can understand, learn, and make decisions based on data and statistical relationships.

What are the three shortcuts taken to create today's intelligent machines?

The three shortcuts taken to create today's intelligent machines are: 1) the use of statistics to replace deep theories, 2) leveraging preexisting data found 'in nature' or on the web, and 3) constant user observation to build mathematical models of user preferences.

Why is it important to understand the nature of today's intelligent machines?

It is important to understand the nature of today's intelligent machines because they are at the heart of the global infrastructure, making decisions that impact our lives. Understanding them helps us draft regulations, solve real problems, and coexist effectively with these machines.

What are some of the concerns raised about the use of algorithms in today's society?

Some concerns raised about the use of algorithms include their potential to make suspicious and debatable decisions, issues of neutrality and honesty, the possibility of racist algorithms, the spread of fake news, and the problem of behavioral addiction to machine-recommended content.

Why is it crucial to regulate and understand the algorithms that power the web?

It is crucial to regulate and understand these algorithms because they have been delegated with important decisions that affect our lives. Proper regulation and understanding can prevent misuse, ensure fairness, and protect vulnerable individuals from harm.

What does the speaker suggest is the key to successfully coexisting with intelligent machines?

The speaker suggests that the key to successfully coexisting with intelligent machines is understanding that they possess a different type of intelligence than humans. Accepting and respecting this difference will allow us to adjust our expectations, regulations, and educational approaches accordingly.

How does the speaker relate the understanding of intelligence in machines to the historical realization that humans are not the center of the universe?

The speaker relates the understanding of intelligence in machines to the historical realization that humans are not the center of the universe by drawing a parallel between the two Copernican shifts in perspective. Just as we had to accept our place in the cosmos, we must also accept that our form of intelligence is not the only one and that machines can be intelligent in their own unique way.

Outlines

00:00

📡 The Quest for Universal Intelligence

This paragraph discusses the concept of intelligence and the challenges faced by Carl Sagan when tasked with crafting a message for the Pioneer 10 probe intended for extraterrestrial intelligence. Sagan's approach was to use universal elements such as the hydrogen molecule, star positions, and mathematical principles, assuming that any form of intelligence would evolve under the same physical laws. The paragraph also touches on the limitations of human imagination in conceiving truly alien intelligence and the anthropocentric bias in defining intelligence. It concludes with a broader perspective on intelligence, suggesting that it is not exclusive to humans and can be found in various forms, including in machines like YouTube's recommendation system.

05:04

🔍 The Rise of Statistical Intelligence

The second paragraph delves into the evolution of artificial intelligence, highlighting the shift from attempting to replicate human-like intelligence to embracing a statistical approach. It explains how modern AI systems, like those powering social media and e-commerce platforms, rely on vast amounts of data and statistical relationships to predict and respond to user behavior. The paragraph outlines three key components in creating these systems: pre-existing data, statistical analysis, and constant user monitoring. It also raises concerns about the potential downsides of AI, such as biased algorithms, the spread of fake news, and behavioral addiction, emphasizing the need for global dialogue and regulation to ensure ethical and beneficial AI development and use.

10:07

🌟 Embracing Diverse Intelligences

The final paragraph emphasizes the need for a paradigm shift in understanding intelligence, proposing that there are multiple forms of intelligence beyond human cognition. It calls for recognizing the unique form of intelligence that machines possess and adjusting our laws and educational systems to accommodate this diversity. The paragraph also draws parallels with the historical realization that humans are not the center of the universe or the ultimate species, suggesting that we are one form of intelligence among many. It concludes by encouraging a deeper understanding and coexistence with AI, learning from past mistakes, such as environmental degradation, to ensure a harmonious and beneficial relationship with these new forms of intelligence.

Keywords

💡Intelligence

Intelligence in the context of the video refers to the ability to learn from experience, adapt to new situations, understand complex concepts, and use knowledge to manipulate one's environment. It is not exclusive to humans and can be found in various forms across different entities, including animals, plants, and machines. The video discusses the concept of intelligence in relation to alien life, human evolution, and artificial intelligence, emphasizing that intelligence can take different forms and is not solely a 'human signature'.

💡Alien Intelligence

Alien intelligence, as mentioned in the video, is the hypothetical form of intelligence that may exist in extraterrestrial life. Carl Sagan's attempt to communicate with such intelligence using the 'Pioneer 10' probe illustrates the challenge of understanding and communicating with something truly alien. The video suggests that our difficulty in conceiving alien intelligence stems from our inability to imagine something vastly different from our own human perspective.

💡Lingua Franca

The term 'lingua franca' in the video is used to describe a common language or system of communication that is mutually understandable by different parties. In the context of Sagan's message to outer space, the 'lingua franca' was intended to be the universal laws of physics and mathematics, which it was assumed any evolved intelligence would understand. However, the video points out that this assumption may be flawed due to the diversity of potential intelligences.

💡Machine Learning

Machine learning is a type of artificial intelligence that allows machines to learn from data and improve their performance over time without being explicitly programmed. The video discusses how machine learning has been used to create intelligent machines that can predict user behavior and preferences, such as those used by YouTube and other social media platforms. This is achieved through the analysis of vast amounts of data and the establishment of statistical relationships.

💡Prime Numbers

Prime numbers are a mathematical concept that refers to numbers that are only divisible by 1 and themselves. In the video, Sagan's use of prime numbers in his message to alien intelligence is an example of using a theoretical mathematical principle as a form of communication. The assumption was that any intelligent life would understand these principles, but the video questions the universality of such an assumption.

💡

💡Statistics

Statistics in the video are used to describe the method of analyzing data to infer patterns, trends, or relationships. It is a key component in creating intelligent machines that can predict user behavior. The video explains that after failing to replicate human-like intelligence in machines, statisticians found success in using statistical criteria to understand and predict user needs, which has become a cornerstone of modern AI applications.

💡Data

Data in the context of the video refers to the raw information or facts collected through various means, often from the web. It is a crucial ingredient in the creation of intelligent machines, as it provides the material for statistical analysis and machine learning. The video highlights the importance of data in training algorithms to understand and predict user behavior, which is then used to personalize content on platforms like YouTube and Facebook.

💡Algorithms

Algorithms are the set of rules or procedures used by computers to solve problems and perform tasks. In the video, algorithms are central to the functioning of intelligent machines, such as those that power social media platforms and search engines. They are responsible for making decisions about what content to show users, based on statistical analysis and user behavior. The video raises concerns about the neutrality and honesty of these algorithms and their impact on society.

💡Artificial Intelligence Regulation

Artificial intelligence regulation refers to the set of laws, policies, and guidelines that govern the development and use of AI technologies. The video discusses the need for such regulation to ensure that AI systems are used responsibly and ethically. It mentions that parliaments and global leaders are now openly discussing AI, indicating a growing recognition of the importance of regulating these technologies.

💡Behavioral Addiction

Behavioral addiction, as mentioned in the video, is a condition where a person becomes compulsively engaged in a particular behavior, such as the use of social media or online content. The video raises concerns about the role of intelligent machines and algorithms in potentially fostering such addictions by recommending content that keeps users engaged, which can have negative side effects on individuals and society.

💡Copernicus Revolution

The Copernicus Revolution, named after Nicolaus Copernicus, refers to the paradigm shift that occurred when it was realized that the Earth is not the center of the universe. In the video, this concept is used metaphorically to describe the need for a similar shift in our understanding of intelligence. It suggests that we must recognize that human intelligence is not the only form or standard, and that we must learn to coexist with and understand other forms of intelligence, such as those found in our machines.

Highlights

Carl Sagan was asked to write a special message for the 'Pioneer 10' probe to be sent into outer space.

The message was intended to be easily understood by any form of alien intelligence using principles of physics and mathematics.

Sagan consulted with colleagues to draft a message using the hydrogen molecule, star positions, and mathematical principles.

The assumption was that any intelligence in the universe would evolve within the same physical and mathematical laws as us.

Sagan later prepared a message using prime numbers to be sent via radio telescopes.

The cat analogy illustrates the potential lack of feedback or understanding from truly alien intelligence.

Sagan's approach to crafting the message was critiqued for its difficulty in imagining something truly alien.

Intelligence is not a 'signature' unique to humans; other forms of life and even machines can exhibit intelligence.

YouTube's recommendation system is an example of machine intelligence that learns from user interactions.

The development of smart machines involves using shortcuts such as statistics, preexisting data, and user observation.

Machines use statistical relationships and constant user observation to predict user actions and needs.

Intelligent machines like Instagram, Facebook, Amazon, and TikTok are part of our global infrastructure.

Algorithms can sometimes make questionable decisions, raising concerns about their neutrality and honesty.

The issue of behavioral addiction to machine-recommended content is a growing concern.

Global discussions on artificial intelligence and its regulation are taking place at high levels, including parliaments and Heads of State.

The speaker has been working on these issues for 28 years and is surprised by the current level of global attention to A.I.

The speaker suggests we need a Copernican shift in understanding to recognize that human intelligence is not the only form.

Coexisting with different forms of intelligence requires understanding and adjusting our actions accordingly.

Transcripts

Translator: Elena Montrasio Reviewer: Michele Gianella

What does being intelligent mean?

This question suddenly became relevant,

in 1972,

to the American astronomer Carl Sagan,

when he was asked to write a very special message.

It was a message to be sent to outer space

on the "Pioneer 10" probe that was to leave the solar system.

The message had to be easily understood by any form of alien intelligence.

Carl Sagan consulted his physics and mathematics colleagues,

and they all decided to draft a message

that would use the hydrogen molecule, the position of the stars,

and a set of theoretical mathematical principles.

Their idea was that any form of intelligence in the universe

would have to evolve

within the same physical, math laws as we know them today,

and should therefore be in a position to understand them.

This was the "lingua franca" they intended to use.

Some years later,

Sagan even prepared a message

to be sent to alien intelligence via radio telescopes

featuring a series of bits

using the principles of prime numbers.

Well, I showed his messages

to my cat -

who is an intelligent being, understands everything,

can deceive both the mice and me whenever it wants to -

along with the other message,

without getting any feedback whatsoever.

This leads us to believe

that perhaps the aliens received Sagan's message, too.

The question is: where did Sagan go wrong?

The answer is not so difficult:

he had difficulty imagining something truly alien

because it's not easy to conceive something so different from us.

Indeed, his message would have been incomprehensible

even on this planet, even to humans,

Leonardo da Vinci, 500 years ago,

couldn't know the hydrogen molecule.

And even today, on this planet,

many people are not interested in prime numbers.

What he was looking for was not intelligence,

but rather another version of himself.

And it's a mistake we always make, we've often made throughout history,

and he was no exception.

It is a lesson for us to learn, one of the limitations we have.

And so, how can we think of an intelligent agent?

Well, imagine an entity, in this case you can see a bee,

with targets to reach.

The bee wants sugar, in the form of nectar;

it has to make decisions, maybe mistakes at first -

but learns eventually.

But there's also a plant that wants to pollinate other plants,

and to do so it must manipulate the bee.

So it does the same.

Agents with goals,

capable of learning from mistakes, adjusting and thinking

are intelligent agents.

Intelligence isn't a "signature" of humans.

For sure,human beings are not the only intelligent ones:

there was intelligence on this planet before us.

Living beings are not the only ones, either.

Machines can be intelligent.

YouTube is able to understand what to suggest,

when and to which user

in order to get us to click.

If it makes mistakes, it learns.

How were we able to do this?

It is important to understand how we did this version of smart machines

because we have to live with them, draft regulations;

and all of this makes no sense

if we try to solve problems that do not exist,

and regulate machines that do not exist.

Those existing today

have been made by us more or less in the same way,

by taking shortcuts.

The first shortcut is the use of statistics.

After many decades of trying to do what Carl Sagan did,

and instill in machines

a type of intelligence similar to ours - human, mathematical -

the kind of intelligence one typically finds in a scientist -

we found out, that method wasn't successful.

Today, if I Google "Cernobbio",

without even typing the whole sentence, based on statistical criteria alone,

it understands what I want to say, what information I might need,

and tries to predict my actions and my needs.

Statistics can often replace

the deepest theories.

This, in turn,

triggers a need for data, lots of data,

in which to find these statistical relationships.

This is the second shortcut we took:

instead of producing data ourselves, by paying,

we take it from where we can find it - "in nature", as we use to say.

And data already existing in nature is to be found on the web,

where much can be learned about both Lake Como

and "The Betrothed", and many other sources,

with different degrees of accuracy and reliability.

And then?

And then the machine has to understand what the user wants,

which is no easy task for a machine.

So we took a third shortcut:

we decided to constantly monitor the user,

assuming that each click signals his or her preferences,

and eventually built a mathematical model of each user.

Here they are:

preexisting data in nature, statistical relationships

and constant user observation.

These are the three ingredients we use in todays' recipes

to create the intelligent machines we use all the time:

it goes for Instagram, Facebook, Amazon and Tik Tok.

And also for ChatGPT.

We don't fully understand what they know,

precisely because they're made with a statistical approach.

But we've placed them at the heart

of the global infrastructure that makes the world go round:

the web.

We have delegated important decisions

to these machines,

which have an impact on all of our lives.

What can go wrong?

Well, for starters, we notice

that these algorithms sometimes make suspicious and debatable decisions,

which make us question their neutrality and their honesty.

Some talk about racist algorithms,

some want to use them to win electoral support.

Some notice that fake news circulates because the algorithm recommends them.

And then there's the issue of behavioral addiction:

there are people who can't stop

using content recommended by a machine.

So how can we trust them?

How can we leave a child in front of YouTube, TikTok,

or a vulnerable adult?

Are there any side effects?

How can we make sure nobody will get hurt?

Fortunately, for a few years now,

we have been looking into this issue,

and not only scientists, everybody.

And today, parliaments across the globe

are openly talking about artificial intelligence.

Two weeks ago, Heads of State met in Bletchley Park

to discuss A.I.

This is something I never dreamed of as a student.

I've been working on these issues for 28 years now,

and I never thought I'd see the world's prime ministers

discussing intelligent algorithms.

MEPs have come a long way, the law is almost ready.

Maybe, this time we won't make the mistake we made with the environment;

perhaps, we started off in time

and will have enough time.

What matters is, we cannot pull the plug,

not only because it is technically impossible, today,

to pull the plug and live without these devices.

The entire web works thanks to smart algorithms.

There's no point in having a library without a catalogue,

and we can't use YouTube or the web

without an algorithm finding content for us.

And besides, it wouldn't be fair.

There are many people who waited for decades

for their chance to access affordable education

and affordable health care:

now it's their time.

These methods are going to give so much to everyone.

It would be crazy to pull the plug.

But we have to live with them, regulate and understand them.

This calls for a Copernican step: we must overcome a limit.

Just like we realized

we are not the planet at the center of the universe -

and we had to struggle to understand it -

and we realized we are not the ultimate animal species,

the final destination of every evolution:

we are one of the many species.

And we have yet to understand

that we are not the point of reference of every intelligence.

We are one type of intelligence.

There can be other types of intelligence.

Our machines are already intelligent, just in their own way.

We must treat them differently from us:

theirs is another type of intelligence.

The day we come to terms with this, it will be easy to pass the right laws

and teach the right things at school.

It's an important step

because coexisting means understanding, so that we can adjust accordingly.

The mistakes we made with the environment

were made

because we didn't really understand the underlying mechanisms for a long time.

In this case, we know what we can do

and we still have time.

At the end of the day, this is what being intelligent means.

Thank you.

(Applause)