AI and technological singularity explained – Jonas Witt // #SFTF

Thanks a lot.

Thanks a lot, Maike.

I'm happy to be here.

It's a small round but that's nice, that's cozy and I want to discuss with you the, in my view, most interesting and fascinating concept of our time.

It's not so known actually and that's why I want to bring it a bit closer to you.

I want to start with the quote, though.

Any sufficiently advanced technology is indistinguishable from magic.

Most of you have heard this quote, I guess.

Where's it from?

It's from that guy, Arthur C. Clarke.

He's one of the most famous science fiction authors and he thought about the future of technology and what it brings to humanity and he formulated a few laws and this is the most famous one.

But it's been since then vastly discussed what did he mean by it.

There seems to be a certain threshold that is beyond our imagination where we cannot really have a feeling for what's possible and what's beyond it seems magical by now.

However, it might be possible at some point.

And I think this is a very interesting threshold that he had in mind and I want to basically explore this threshold more with you in a thought experiment.

It begins with you.

You are here now today in 2024 and I want to let you think of how much in time do you have to travel backwards in order that you meet a person who basically is seeing our current time with this threshold and thinks that the things that are happening now in our world technologically are beyond magical.

What would you say?

I mean, I'm the guy here now at the stage so I can kind of make a decision.

I put you on a timeline and a time scale and I put you in a time machine.

Let's choose the DeLorean from back to the future.

And I would say probably a hundred to a hundred fifty years back into the past you might be able to brag a lot about what you actually have, what you're used to technologically.

You might meet a person around 1900 in the middle of the Second Industrial Revolution.

What could you tell that person?

I think many, many things can come to mind.

Yeah, let's say, by the way, we have planes.

We can for cheap price basically go around the world.

Many, many people can afford it.

We've been to the moon.

We have like machines around the orbit in the earth that we can like have communication with.

We have incredible medicine.

Life expectancy has doubled.

Like the average person lives twice as long as you're used to.

We have this little device that is basically magical.

It can do everything that we can imagine.

You can have a call with someone around the world for basically no cost.

And of course like talking about AI, right?

I mean I can imagine that person would basically be mind-blown.

I never, by the way, found a better like instance to use that emoji.

So the interesting thing is now, let's say that person becomes jealous and wants to do the same experiment.

Wants to travel back in time as well.

How many years do you think that person does have to travel back in time in order to get to the same magical threshold that you just explored?

It's not 100 or 150 years.

For sure if that person would travel back in time 100, 150 years, it would be impressive, right?

However, would it be the same level of magical that we just explored?

I think we have to rescale.

We have to go back much more in time than we just did.

Instead of a hundred or hundred fifty years, my argument would be you should have to travel back a thousand or a thousand five hundred years.

So let's put the second person into the DeLorean.

Let's travel back in time.

That person now meets another person from kind of the ancient times and starts bragging as you did like of technologies that were basically possible in that era but not previously.

What could that be?

I mean a lot of things, for example, windmills were not possible before.

You all know maybe the story of Don Quijote who was like surprised by windmills and afraid of it because he didn't know it, the technology.

For sure gunpowder, maybe the compass, you can have orientation around the world.

For sure something like the electric bulb from the second industrial revolution, steam engines, trains, so machines like that have a steam engine on them and can like travel long distances, seems quite magical.

For sure paper but also of course the printing press.

You could make the argument that that person was similarly surprised than the person basically you met in 1900.

It's also mind-blowing.

Now let's do the same thing again.

How much does that person have to travel back in time?

You can already imagine it's not a thousand or thousand five hundred years.

That would of course be impressive but it wouldn't be as mind-blowing, it wouldn't be beyond magic basically.

We have to rescale again.

I would make the argument that person now has to travel back in time not a thousand five hundred years but rather fifteen thousand years back in time which is the pre-neolithic era before agriculture where we were just a bunch of hunters and gatherers across the world.

So that person would meet another person, a hunter or gatherer from fifteen thousand BC.

What would that person be bragging about?

Agriculture seems quite magical.

I think when you just like hunt animals all day and like basically travel after the herds.

Also cities, constructions, mathematics, ships to travel like around lakes or oceans and the invention of the wheel.

I mean it didn't look like in this emoji example for sure right but the invention of the wheel a few thousand years ago.

And again that person would be similarly mind-blowing.

You kind of get the gist of my point right?

Always the threshold to show something magical to a person in the past grows by order of magnitude the more you get in the past.

Let's plot this what we just basically worked on in this thought experiment.

If you plot progress on the y-axis and time on the x-axis and you do this over the course of humanity.

You see that there's movement right?

It seems the closer you come to our current times, the more innovation is, the more technological progress is, the more things are happening.

Things become way more boring the more you get basically back in time.

If you plot this function, it's an exponential function.

And with this I want to make a first statement here.

Hypothesis.

Technological progress follows an exponential path.

And I want to explain you why.

Going back to our plotting of progress and time.

A single technological innovation never grows exponentially.

It grows like this.

It's a logistic function.

So in the beginning for a long time things don't work as much, they don't scale, they are not great.

Then it scales rapidly and then it's plateauing and basically reaches a certain limit.

One example is cell phones, smartphones.

In the beginning when nobody could afford them, they were very shitty.

It took a long time for them to get great.

Then suddenly we had an explosion of mobile phones and smartphones.

First smartphone 2007, the iPhone.

And then there's a plateauization.

I'm sure most of you guys don't really know the difference between iPhone 15 and iPhone 10, right?

What happened?

So there's a plateauization.

However the interesting thing is that our world consists of thousands and over thousands of these logistic curves.

Always the next curve basically is starting at a different point than the previous one.

Because it already builds upon the knowledge that was previously acquired and the technology that we previously built.

If you think about it, it's very true.

The next technological wave or generation is always built with the previous one.

With the knowledge of the previous one, with the tools of the previous one.

An advanced society like ours is capable of much more innovation than an ancient society, like by far.

If you put a very innovative person of our time into the, let's say, Stone Age, and I know it's like very controversial now, but let's say Elon Musk, maybe as one of the genius person of our times.

In the Stone Age, he couldn't have invented anything.

I mean maybe, I don't know, some like, maybe he came up with the wheel at some point.

But like it's incomparable to now.

Because there is not the capital, the people, the knowledge, the technology, all the tools that we can nowadays leverage.

That's why when you plot this abstractively over time, you have this collection of logistic curves, basically, that add up to an overall big exponential function.

Like an exponential progress of technology over the course of humanity.

I want to make a second statement.

Your intuition about exponentiality sucks.

It's really not great.

It doesn't matter how many examples you read, try, we don't have a great intuition for it.

I want to give you an example of it.

Let's take a sheet of paper and let's fold it.

Let's say it's 0.1 millimeter thick.

You fold it once, now it's 0.2 millimeters.

Let's fold it again.

Now it's 0.4 millimeters.

Let's do this 42 times.

It's an exponential progress, basically.

Of course, at some point it's really not possible anymore for me to fold it, right?

So think like, theoretically, if you could do this 42 times, what would you guess how big is the resulting stack?

Like how high would it be?

Anybody an idea?

To the moon?

I mean, I'm very impressed by this answer.

Usually people say one kilometer, a hundred meters, I don't know, to the sky.

But the answer is actually really the moon.

It's pretty much, it's a very good guess, it's pretty much 440,000 kilometers.

So I don't know whether all of you had such a great and like exponential intuition like like that guy, but usually we are very wrong about these functions.

And it doesn't matter how many examples you know, you will every time again be very wrong about it.

And the reason is evolutionary, nothing was exponential, like in your daily life, whatever grows exponentially.

If you were hunting an animal in ancient times, its speed didn't just speed up exponentially, right?

It just, it's, we are based, our world is based of linear functions.

If we put those two statements together, so technological progress supposedly develops exponentially, and you have a very bad intuition about it, that means your intuition about what's going to happen in the next 10, 20, 30 years will be vastly wrong.

While we are here, at that point, your intuition tells you that we are developing like this.

However, you can already imagine reality might look much more like this, whatever that means.

And the interesting thing is now, I'm not talking about the next a hundred years, the next thousand years, I'm talking about the next one, two, three decades.

The first person ever in human history who kind of made this observation, which I find is incredibly fascinating, is this guy.

He's John von Neumann, one of the, or maybe the most influential mathematician of the 20th century.

He developed game theory, for example, was very much involved in the mathematical foundations of quantum mechanics, a close friend of Albert Einstein in Princeton, and he looked at humanity, and he came up with this observation in the 50s, and he said, the ever-accelerating progress of technology gives the appearance of approaching some essential singularity in the history of the race, beyond which human affairs, as we know them, could not continue.

So he was thinking, of course, mathematically, like, where's this leading, where's this going?

There seems to be, like, basically a point where the function kind of leaves the space, so to say.

Another way to explain what a singularity is, is basically from physics.

This is the first ever taken picture of a black hole, in 2017, actually.

Back then, the scientists also got the Nobel Prize for it.

And in physics, the center of black hole is called a singularity, or supposedly there's a space-time singularity in there.

That means, the closer you get to a black hole, the more space-time is curved.

The closer you get, it's more curved, more curved, and at some point, it's almost infinitely curved.

It's so curved that basically our laws of the universe, like quantum mechanics and relativity, don't apply anymore.

We have no clue what's going on in there, because we don't have a model of the world which describes what's going on in there.

It's kind of boring from physics now.

The technological singularity, analogously, explains a hypothetical point in the future where progress becomes so fast that our concepts of reality don't apply anymore.

Like, it's so fast that we cannot literally predict the next day.

Think of, like, the progress of, like, a hundred years now happening in one day.

Like, it would question ground, our underlying concepts of our reality, like how society works, how our daily life works.

Now, the question is, if this is true, and progress is much faster in the upcoming one, two, three decades than it is now, what kind of magic will we encounter?

Like, what will be possible?

Think of a guy in 1900 coming up with ideas like there one day will be a smartphone or will be AI.

It's kind of very hard to think about that, right?

Maybe one quote I like here from, you all know him, Sam Altman, who also very much believes in this exponential progress of humanity, is that one.

The cost of intelligence and energy are going to be on the path towards near zero.

The constraints we've all grown used to will no longer apply.

We will be able to do stuff with bits and atoms we can barely imagine now.

And in this context he was not not talking about in hundred years or in a thousand years.

He's talking about in 10 years, in 15 years.

So what could that be?

I mean, I don't have a clear answer, and the following list might look super crazy, but have in mind that this is the very reason why I put these things there now.

This is thing, these are things beyond the magical threshold basically.

AI will far outperform humans in every cognitive task.

That means in your pocket or wherever, whatever is coming after smartphone, you will have access to a superhuman doctor, superhuman lawyer, superhuman coach, superhuman teacher, whatever you want.

Abundance of all reproducible goods and services due to innovation.

That means we will leave the software side of things.

We will actually produce things far cheaper than ever before.

Think of companies that only have AI employees, robotic employees.

Of course the cost to produce anything will drop dramatically.

The cost of food will drop dramatically if you create it in a different way.

I know that sounds crazy.

However, they have in mind that's the reason why I wrote it here.

You won't die.

Next thing.

All diseases will be solved.

Aging as one of the main diseases supposedly, I mean this is a mindset that actually develops now in medicine, will be solved.

Your consciousness will be able to leave your body.

It's also called mind uploading, super crazy things that people are thinking about.

Actually this is kind of the mission of the company Neuralink you are aware of from Elon Musk.

This is the actual mission of the company.

Now we have the first patients who actually have an implant and reporting about it just like two weeks ago.

Like if you have time like Google it's really impressive what they're describing.

They can basically interact with the computer by just thinking about it.

We'll be out there among the stars and explore the universe.

I mean that's maybe something you can maybe imagine more now.

This all might sound crazy and I'm not sure whether I want this all to happen and I'm not even sure that whether it will be all happening.

Just have in mind that in 10 or 20 years we'll be able to do things that now seem magical, if technological progress continues as we discussed in the beginning.

Now one question is what could possibly go wrong?

I mean if things are speeding up so much, this is also dangerous, no?

And I have to personally say that I am of course very excited about this, like living in this moment of time and humanity is incredible.

It's much more interesting than I think living in the Stone Age.

However at the same time I'm also a bit frightened about it.

For sure I'm fascinated.

I think we will encounter disruptive dangers on the way that we cannot really imagine yet now.

However I would like to have every person in their era think of what could go wrong, try to anticipate it and try to come up with strategy against it.

One famous example is if it one day becomes very cheap and easy, you have access to superhuman AI and you want to create a virus that's like super deadly, like contagious like COVID but super deadly.

You can just ask the AI to create or something else disruptive might happen.

You can you can just ask the AI to create basically the genetic code for it.

Maybe then it's very very easy to create a virus.

It doesn't cost like huge labs now, maybe hundreds of millions of dollars, maybe just cost a few hundred thousand dollars.

So this would be a challenge that is hard to anticipate now and I'm sure there will be hundreds more like this on the way.

And this is where I want to end actually with you.

Wherever you work, whatever you are doing, whether you are a political scientist or working in some non-governmental organization, in a think tank, whatever, what possible dangers are there that you can think of in the context of superhuman AI, of crazy technological progress?

And I think actually this is the task of our generation.

We have a lot of problems out there, climate change and growing populism and so on, but I think these are probably incomparable with changes that are coming on the way here.

And I want you to think of like how does it feel like to get into the time machine, travel not a hundred years into the future, not a thousand years, not tens of thousands of years, but only 10 to 20 years, having in mind that things will be possible that we now think are impossible and try to contribute to a magical future.

Thanks.

So I was told if we want we can also have like a few questions or discuss it, like I'm always very very interested in getting in touch with people about it.

You can talk to me afterwards, but we could also do it now actually.

I don't know, somebody wanted to give a microphone but...

Yeah, thank you for your great talk, that was very interesting.

I would have a question which is, so we've seen something like blockchain already, like Internet of Things for example, which for the few years that this was kind of standing in public, it was hyped a lot and big big changes were promised by it, but somehow it's not really talked about a lot anymore now.

So do you think something similar could happen to AI as well?

Yeah, it's a good question and this is like, I mean there are many great counter arguments about this, like against this kind of framework of thinking of the future and one of this, one of the arguments is like, there is always like hype cycles about things and things are happening and I would say the technological progress for sure goes in waves, right?

But those waves, if you plot them like on a bigger trend, they overall grow exponentially.

So you will always have hype cycles and the interesting thing about hype cycles is that, while in the beginning there is a hype, we are usually at the beginning of the exponential curve.

That means we have a discrepancy, people overestimate extremely what's possible, while it's totally under delivering.

However, then the hype cycle basically goes into like, like leaves the hype basically and becomes more realistic and at the same time, the development it was hyping about is speeding up and suddenly like growing exponentially out of the hype.

So that's what, it's actually an almost dangerous discrepancy, because some people who are hyping it, are actually right about it, they're just way too early and yeah.

Hi, thanks.

I would want to open the question about power and like, so it kind of seems that from your presentation, I kind of got the sense that technological developments just happen with nobody behind actually driving them and with those nobodies having no interest in what they drive or what they do not drive.

So in your research or in your work, on your general doing in public speaking, do you ever raise those questions about power, risks associated with power?

Yeah, it's a very good question.

Of course, innovations don't just happen.

Incredibly passionate people make them reality, right?

And rich, and people get rich on the way of like doing it, right?

Probably, yeah, probably like people, people who have met much better like prerequisites are also like at a better situation to do it, a hundred percent of course, right?

However, I see the world of like now eight billion people more as like a system of all like small different interactions, right?

And they are all in competition with each other.

It's basically impossible to shut down or to slow down what we are now seeing, because any country who follows this exponential trend is vastly outperforming all other countries, any community, any set of scientists.

Now you have, for example, AI research so distributed in the world, there are thousands of labs everywhere.

It's literally impossible to shut down, slow down anything there.

And of course, there are one of the dangers I share with you, probably this is the point where you wanted to get, is that we have an ever growing discrepancy between like super rich people and super poor people.

I think we will all rise up in the age of abundance, so they will have access to probably all the basic needs.

However, even in Silicon Valley, like in all these like crazy circles where also Sam Altman is part of, they openly discussed now that we will probably not come around a taxation of equity.

So equity means basically of the stuff you own, not of the stuff you earn, but of the stuff you actually own, because it will be so productive.

Think of if you have now a company or a share of a company, and that company basically is able to produce the same service just without all employees, or just much better service without employees, because you just let AI do all the stuff.

I mean it's kind of a very simple thought now, right?

But productivity of companies will get increasingly through the roof.

That means people who own the companies, who founded the companies, will become much, much richer.

Like if you own stuff today, you will be much more richer in the future than if you don't own stuff today.

And we will, even like from a societal perspective, will be hard to tolerate this.

There's no easy answer.

That's one of the key challenges probably in the upcoming years, and we might not get around, even in Silicon Valley now they discuss it, to tax what people own instead of what people earn.

But I mean we can also afterwards discuss, like I'm very... yeah, yeah, there's a lot of interesting stuff to discuss about.

So how will this reach the traditional people all spread all over the world, who are not prepared for, in their daily life, are not prepared for anything like that, and who is going to create the world or the age of abundance?

I currently have the feeling we are far away from an age of abundance, and I don't know how to... so is AI creating an age of abundance in this context?

So yeah, I mean AI is probably the driving factor.

Yesterday I also made kind of a statement.

Everything that you see here around us, whether that's the construction of the building, whether it's kind of, I don't know, technology, whether it's your shoes, it's a product of intelligence.

And this intelligence is going to amplify like by a billion fold in the next 20 years.

Non-biological though.

So it will like enable things that are now beyond our imagination, right?

This is kind of the the main point.

And it's a huge challenge.

I don't have an answer to how it will impact exactly the life of people.

It for sure will just scale you.

So suddenly something becomes available, right?

I don't know that maybe you own a smartphone.

I can imagine you do.

Like most people in the world own a smartphone.

Even like when you go to super rural African eras, the cheapest smartphone costs like $25.

So it's like incredibly scaled.

Even like the white Maasai in Kenya, like they use Google Maps to track the right... where their animals can prune like the right places basically.

So it's scaling, but it's scaling of course in the laws of markets.

And there are huge challenges towards this.

And I think like for many of these challenges, we don't really have an answer.

And I'm, for example, personally very concerned about democracy.

Like democracy is built in the 19th century for a world that's vastly different than now and will be even even more different in 10 years.

So democracy needs a digital update.

And this is just one of probably hundreds of challenges all around like our lives.

And this is I think what's so fascinating.

And I can actually not stand that so many people think of the future as exponential.

I mean there are great points to criticize also these frameworks, right?

However, if there's just a small probability that it will look like this in 10 or 20 years, a lot of people should think about it today.

And a lot of people should already anticipate what might be happening, what might be coming.

And I don't see this like only in, like not really in let's say mainstream.

Hello.

Just what you just said, I mean I think what you meant is being part of, staying part of the value chain.

I think this is the biggest danger that, you know, some companies that own the large language models are having such a big, you know, advantage and getting all the money, while others that are part of it because they were part of the training models, for example musicians, voiceover guys, even faces, you know, or pictures which, you know, you can now ask an AI to make a picture like a famous artist.

But it's of course because it's trained by them.

So to make sure that this value that will be created is, you know, given also to the people that have had a big part of it, even not owning the technique, but owning the intelligence or the originals.

And I think this is one of the key factors in the future.

I mean how do you, how do you see that?

I mean I really don't have an answer.

It's also for me very concerning.

If you look at the stock market now, Nvidia, Apple and Microsoft, all independently are more worth than all German listed companies combined.

It's insane, but the reason is the anticipated value creation will exceed the anticipated value creation of all the German listed companies.

Which sounds crazy, but this is actually the reason.

And I think it's a huge challenge.

I have no idea.

I just know that in Europe we are very good at making a lot of regulation.

We have high ethical standards, high morality, which is amazing.

However, if we don't actually create the economic value here as well, nobody's going to listen to us.

Not everything is perfect here though, like don't get me wrong.

So that's why I see one of the leverages myself to foster the start-up scene and to make people like that have ideas, bring them to reality, think of second-layer solutions, maybe not like the build up on large language more that create value in the real world.

And I think there are a lot of great companies going to be created there that basically make it easier for us to argument for our values in the world.

But to be honest, I don't have an amazing solution.

Yes.

Yeah, so it's and it's probably a great one, I hope.

I mean, many people develop the, I had a long conversation with people who developed also the AI Act, that's now basically it's getting in place in two years, now companies have two years to adjust to it.

And the idea was that like the GDPR regulation that came in place in 2018, that this enables companies to have a competitive advantage, European companies to have a competitive advantage, because they already have to create value GDPR compliant.

And then everybody else in the world will follow by regulation and then we have an advantage.

That's called the Brussels effect, was the idea.

However, it's not really known whether it's actually an advantage.

And people now who develop the AI Act hope that it applies for the AI Act.

And I hope so as well.

We don't know.

It might also be like the critics say that it's just hindering innovation here.

Small startups have to follow regulations, don't able to scale.

That's, it's complicated.

I don't know.

That's why I think in every kind of segment we need to like strongly add, like think of these problems and ideate solutions.

But it's a good point, yeah.

Okay, one more I think then, we're almost done.

Thanks for the talk.

I have a question, like how do you raise the next generation kids, or next generation of our society, or like the role of parents in this kind of age of AI, where on one hand you want them to be adaptable to these kind of technologies and learn all these exponential functions.

On the other hand, you are just giving them like the handover of all the thinking and everything to these AI systems, which for young kids it would be much more detrimental.

So how to balance the act and the role of parents in it?

Yeah, thanks for the question.

I love the question.

I don't have kids myself yet, but it's an incredibly tough one, because probably a lot of rules that we grew up with don't apply to the next generation anymore.

Like, why should you memorize certain things, when like you can just create all kinds of knowledge content yourself by just talking to an AI?

And this will be much easier possible in the future.

Already now, homeworks don't make sense anymore.

Like, and this is the disruption that goes through the basically whole educational space.

Probably it will be something like critical thinking, creativity, knowing what you want, being flexible, also having high resilience against stress, more these kinds of things.

But we don't know, right?

This is one of the answers we need to give.

I like that you come up with so many different examples, because this is exactly what I think.

Like, in all the different areas, if you think of what does it mean for my life, for the life of my children, so on.

Like, you come to crazy words mentally, basically.

And we need to think of solutions as a society for that.

And I would wish that this happens more and more.

Okay, thanks.

Whoever wants to connect with me, feel free.

I'm always looking for like-minded people.