What is artificial intelligence? What is the impact of 'artificial intelligence'? Are we going to solve world problems with this? Or is it going to destroy us?

In this article everything, you want to know about artificial intelligence. I spoke with Neil Jacobstein (Singularity University, photo above), Nicky Hekster (IBM Watson Netherlands), Nell Watson (futurist) and I gave a lecture on artificial intelligence at an event by eOffice and IBM (September 2016) and Focus in Arnhem (April 2017).

What is artificial intelligence?
Artificial intelligence is systems that are self-learning. There is still some confusion about the term because it is a broad concept. Another reason that artificial intelligence is a vague concept comes from two factors.

We associate it with (science fiction) films. But: artificial intelligence is already used on the computer or mobile phone you are currently using. Also, think of systems in your car (ABS) or how Google and Facebook work. That is all artificial intelligence at work.
We think of robots. But: the robot is the packaging or the form, artificial intelligence is about the content. What a robot does?
As I wrote, the term is very broad. In broad lines there are three forms:

Artificial intelligence: specialized in one thing. Like the chess computers. This is also called weak AI.
Artificial intelligence: good in several things. This is also called strong AI. An example is Watson from IBM. This artificial intelligence can also have multiple forms of thinking: problem-solving, abstracting, reflecting and learning from previous experiences.
Singularity. According to Nick Bostrom, this is a comprehensive form in which artificial intelligence transcends total human intelligence. Think also of creativity, wisdom and social skills.
Now you may think of the first form. It is also very useful for that. It would be inhuman to index all the websites on the internet. Fortunately, the artificial intelligence of Google is doing very well. You already use artificial intelligence, without knowing it.

Hod Lipson (Columbia University): Software is eating the world. Artificial Intelligence is eating software.

But if the growth continues to forms 2 and 3, then the consequences for the business community, society and for you are immense. A nice statement is from Hod Lipson, just like Neil Jacobstein speaker at the Singularity University Summit 2016: "Software is eating the world. Artificial Intelligence is eating software. "

Definition artificial intelligence
Many terms come along when it comes to artificial intelligence. What's up with that? According to Neil Jacobstein (chair of AI at Singularity University), 'deep learning' is part of 'machine learning' - which in turn is part of 'artificial intelligence', which in turn is part of computer science.

The essence is to make software smarter on the basis of algorithms. These algorithms are used for recognizing patterns (in data and data structures) and problem-solving.

Expectations artificial intelligence
Expectations about artificial intelligence are high. Technology thinker Kevin Kelly is convinced that artificial intelligence is everywhere. According to him, we started stopping electricity in things. Think of a staircase (> escalator) or whistle boiler (> kettle).

The next phase is that intelligence and intelligence will also be present in all devices and things.

That is also the opinion of Neil Jacobstein (Singularity University): "It becomes a part of everyday life."

Kevin Kelly: Artificial intelligence will soon be everywhere.

That offers huge opportunities. I gave a lecture at a conference of e-office. Nicky Hekster spoke for me. He is responsible for IBM Watson Health in the Netherlands. IBM has developed Watson, a computer. Or rather, artificial intelligence software.

He says that in 2011 Watson won the television game Jeopardy. This was a huge achievement because language is an elusive thing. Now IBM works a lot with Watson for healthcare. The care is ideal: lots of data (also in images), guidelines, studies, research, articles and performance data. Getting the most important conclusions from 800 million pages? Watson can do that in one second.

This example shows that many people are very enthusiastic about the possibilities of artificial intelligence. We are faced with limitations with our current cognitive abilities. What if someone can think with us? Someone who is infinitely smarter.

What about that singularity? The expectation of Ray Kurzweil, a well-known thinker in this area, is that in 2025 computers are more intelligent than the human brain and in 2050 more intelligent than all people together.

Examples artificial intelligence
Or the way we do science. Neil Jacobstein: "Science is going to change. With artificial intelligence, the data is the basis and not the hypothesis. "

With so much more intelligence, we can:

solve deadly and now incurable diseases?
solve the food problem?
explore the space?
to become immortal?
It is not surprising that the symbiosis of human and artificial intelligence is seen as the next step in the development of us as humans. As Vernor Vinge (Professor of Mathematics and Science Fiction Author) says in relation to artificial intelligence: "We are on the verge of a gigantic change." (loosely translated)

Vernor Vinge (professor of mathematics and science fiction author): We are on the other side of the world.

For me personally, this is very striking. The impact of this development on our body and life is immense. During the IBM symposium, for example, I also spoke with Tom. Tom has brought in a chip with me. He is enthusiastic about what you can do with artificial intelligence in combination with implanted chips [link to the article about the chip in my body is at the bottom].

Neil Jacobstein expresses this beautifully: "With artificial intelligence, we can become an improved version of ourselves." In my presentations on biohacking, I always tell you something about artificial intelligence. It does not stand alone but in connection with other developments concerning biohacking, technology, and innovation.

Opportunities and innovation
How fast does the development of artificial intelligence and more specifically: singularity (type 3)? According to Hod Lipson at the Singularity University Summit (2016), things are going strong in the coming years.

Why?

Calculation power
develop

1 The basic principle is that your software must have sufficient power to do all calculations, which we also do in our heads. This is where Moore's Law comes into play. There are several definitions of this Act. In fact, the description of the Law has changed a few times in all these years.

Moore's Law states that the number of transistors in an integrated circuit doubles every two years due to technological advances. In short: more computer power in less space, less money, and less energy.

According to experts, it will take until 2025 before we reach the stage that so much processing power is affordable and available. An example that is so fast is that Google's artificial intelligence was able to beat one of the best GO players in the world [link at the bottom]. GO is exponentially more difficult in terms of play than butter, cheese and eggs or chess.

2 The next problem is how to build artificial intelligence. There are 3 ways:

Neural networks. Build computers like our brains work.
Evolutionary. Our brain has been developed after centuries of evolution. With software and computing power we can accelerate the evolution towards singularity through 'trial and error'.
Making artificial intelligence self-made. Well, this sounds very meta. But you can instruct artificial intelligence to further develop itself.
An example of type # 1 (neural networks) is how Google Photos now works. The program can cluster and categorize photos itself.

An example of type # 2 (evolution) is a study by Bongard and Zykov that was published in 'Science'. In it, they show an investigation of a robot that learns to walk by trial and error.

An example of type # 3 (meta artificial intelligence) is chess computers that started to play and learn from each other. That is 'machine learning'.

If we are there, then that has a number of advantages. An artificial brain that is faster and more reliable. In addition, we can implement upgrades and we can connect them. We can not do that with the human brain.

By the way, that you can make a collective brain with computers, that has many advantages. Think of cars. They know each other where they drive and can, therefore, anticipate each other much better.

Much better than we can as people.

Threats of artificial intelligence
What are threats of artificial intelligence? Or rather, do singularity and superintelligence mean the end of us as humans?

This can be done in 2 ways:

The superintelligence does not need us as a human being anymore
The first country or group that has superintelligence, destroy the rest
It may end badly for us. That is the warning that people like Stephen Hawking and Elon Musk give. Or take philosopher Sam Harris. He warns, for example, of an arms race between countries for the development of superintelligence, because the country that is the first to control artificial intelligence is the first to control the future.

In 1993 author Vernor Vinge already warned: "Within 30 years we have technological possibilities to develop superhuman intelligence. Then the age of man is over. "

Sam Harris: The first country with superintelligence is in charge of the future. That is a huge danger.

The sounds also come from the industry itself. For example, the authors of the Stanford Report (including Google, IBM, and Facebook) indicate that the risks and considerations of artificial intelligence in various domains can be far-reaching and different [link at the bottom].

According to them, it is important to develop an ethical standard for artificial intelligence. In this, the role of government and politics is also important. Incidentally, they also indicate that it is a field of tension to ensure that these kinds of agreements do not delay innovation and development.

Others, such as Marc Andreessen (by Andreessen Horowitz) and Martijn Aslander in his podcast with me, are more positive [link to the podcast is at the bottom]. A man has developed artificial intelligence. It is therefore highly unlikely that artificial intelligence would want to destroy humanity.

Disadvantages (4x)
What are the possible disadvantages and risks of artificial intelligence? For me, these are the three most important ones: transparency (1), goal (2), free will (3) and privacy (# 4).

The disadvantages are further elaborated below.

Transparency

1 Transparency. An important disadvantage that is mentioned is that if systems learn independently, it is no longer traceable and transparent how they came to a decision. Our human abilities fail in this. A well-known example is the use of algorithms in the judiciary.

Proponents believe that computers can interpret and weigh up the facts better and without prejudice, followed by a better decision. They certainly expect this in simple lawsuits such as divorces and traffic fines. Opponents of this development believe that speaking of right is precisely something that a person must continue to do.

Especially if systems learn themselves, then if you are convicted by a computer you want to know how it came to his or her decision. Is it clear how the algorithm works, what data the system used and how this led to the statement?

Target
2 Goal. Another and current disadvantage is already present in digital media. Algorithms determine on social media such as Facebook and Twitter which message you will see. The purpose of these algorithms is that you stay on the platform for as long as possible and preferably also click through on an advertisement.

The risk of this is twofold. On the one hand, selecting news may be something human, for which you can think about feeling, emotional values, and ethics. What if we only leave that to robots? A second risk is that the algorithms offer their offer specifically per person. This gives me something different to see you. This is the familiar 'filter bubble'. The result is that people do not understand each other as well because they lack a common basis.

Free will

3 Free will. Robots, software and artificial intelligence already take over a lot of tasks from us. Take a city navigating. Do I make a conscious choice or let myself be completely guided by an app on my phone. The same applies to recommendations on for example Netflix or Spotify.

An advantage of not letting yourself be guided too much is that you are faced with surprises, mysteries, mistakes, and serendipity. Or is artificial intelligence going to program that in itself?

Privacy

4 Privacy. Imagine that we get very clever artificial intelligence assistants. These take over a lot of tasks from us, such as arranging our finances, making appointments with the GP or planning time with friends and family.

Although we already allow this in some areas of our lives, this goes a step further. The more responsibilities we release, the more dangerous it becomes when things go wrong. Our assistant always follows us. What if it is hacked? Is the quality of human relations less? Are we not easily manipulated, on a much larger scale and with much more impact than is currently the case?

Now I sometimes say jokingly during a lecture or presentation: 'Google knows me probably better than my girlfriend, my mother or myself'.

Criticism
There is also criticism of the so-called potency of artificial intelligence. Lorenzi Florida is the philosophy professor at the University of Oxford. In 2014 his book 'The Fourth Revolution' was published. This revolution revolves around the fact that offline and online lives are increasingly intermingling. The first three revolutions were that the earth revolves around the sun (Copernicus), natural selection (Darwin) and that we do many things unconsciously (Freud).

He is not as enthusiastic about the current applications of artificial intelligence. According to developments are fast because technology is cheaper, more data is available, more computing power, better algorithms, more devices with internet are connected and more and more people (increasingly) online.

But it will take a while before we have 'strong AI'. Luciano Floridi in an interview with de Volkskrant: "I am impressed by specific solutions for specific problems. With the board game Go for example. But the breakthroughs remain within the context of such a game. However, real life is not a set of agreed rules. "

According to him, it is necessary to think better about the possibilities and impossibilities of artificial intelligence. Thus Luciano Floridi: "No device, no matter how clever, relieve us of our own responsibility."

Applications
What are other interesting developments and applications concerning artificial intelligence?

Open source
Smart earplugs
Privacy
Turing test

1 What I find fascinating is what Google does in this area. They have developed TensorFlow [link at the bottom]. The beauty is that they have made it 'open source'. Everyone can, therefore, get started.

Is that a vision of the future? Artificial intelligence accessible to everyone free of charge?

NB A programmer working on artificial intelligence told that he had analyzed photos with Tensor Flow and Watson. Tensor Flow was better.

2 Yes, is the answer to the previous question. So it is not crazy to think that you will wear earplugs with intelligence [link at the bottom of the collaboration between IBM and Draghi]. Ideal if you are in a foreign country where you do not speak the language. Or what else could you do with earplugs that can learn themselves?

3 What about privacy? To let systems learn themselves, they need access to your data. What price do we pay for artificial intelligence in all our stuff, such as in our earplugs (see development # 2)? Is all data transparent to make our lives easier? What do we give that?

4 When do we no longer know whether we are talking to a computer or to a real person? That is the basis behind the Turing test. In the film 'Ex Machina' this theme is unraveled in a fascinating way. Now there are countless apps and applications that make the first contact with the customer on the basis of software.

Is that test still relevant? Or will the boundary between man and machine soon fade away? I do not have answers, but these developments show that there is a lot going on and that it will have a huge impact on our lives.

healthcare
At the Get into the Future Congress 2017 in Amsterdam, I spoke with Kris Verburgh (author of the book 'De Voedselzandloper') about the application of artificial intelligence for health care.

He told about IBM Watson. This artificial intelligence helps doctors to make diagnoses, by analyzing genetic material of tumor cells, by dissecting big data and by distilling scientific publications.

An example is Japan doctors did not know what to do with a tumor in a young patient. Watson controlled the data and concluded that it was a rare form of leukemia. It even seemed to be a form that doctors had never even heard of.

Other examples include research into eyes (1), the development of medicines (3), the assessment of images (3) and the analysis of notes (4).

1 Google Deepmind analyzes images of eyes to learn from that [link at the bottom]. Can we later discover certain syndromes with a photo of the eyes?

2 Atomwise is a company that uses 'machine learning' to do research at a molecular level with the aim of discovering new medication [link at the bottom].

3 I was at the AI ​Congress 2018 in London. During the two-day event, one morning was devoted to artificial intelligence in health care. Often it was about companies that had solutions for 'object recognition' based on 'computational neural networks'.

This can be applied in autonomous vehicles to scan the environment, but also in assessing and diagnosing radiological images.

4 Finn Catling from University College London told about a project to encode and analyze the qualitative notes of doctors in a text. According to him, the most medical value is often in the doctor's opinion. By analyzing that data with self-learning systems, he hopes

The flaws still came to light in a panel discussion. According to the participants, artificial intelligence (at the moment) is of added value in the healthcare sector in terms of administrative and standardized processes. The technology must be supportive to professionals who make decisions and the care remains for a large part based on human contact.

Development of artificial intelligence
At the World AI Summit 2017 in Amsterdam, I spoke with Gary Marcus. Gary Marcus is a professor at the University of New York at the Psychology department. He has also been involved for a time as the head of Uber's artificial intelligence department.

Gary Marcus gave a number of warnings about the future of artificial intelligence. "We do not just need mathematicians. But also psychologists, linguists, anthropologists, and sociologists. "According to him, the danger is that the development of artificial intelligence will come to a standstill. He calls this the 'AI winter'.

Gary Marcus (University of New York): We do not just need mathematicians.

Commercial companies that deal with artificial intelligence, such as Google, Facebook, and Amazon, have their own goals. It is now necessary to expand the current artificial intelligence with human capabilities and capacities. He, therefore, argues for a CERN for artificial intelligence. A public initiative and collaboration to bring artificial intelligence to the next level.

Impact artificial intelligence
The call that artificial intelligence should be extended with human capabilities was also the call of Marco Vernochi of Accenture. He still had a nice one-liner: we have to think from 'user interface' to 'artificial intelligence interface'. We are going to work more and more and communicate with smart algorithms. Man must remain central to this.

That does not alter the fact that the impact of this development has, in his view, an enormous impact. Both on society, on economies, on industries, organizations, work processes, work, jobs, and people.

Accenture has conducted research into the impact on companies. They expect productivity to increase by 40% and corporate profits by 38%. How they have investigated this and what assumptions they have made, did not come out very clearly.