Christoph von der Malsburg: 'Artificial intelligence attempts to do what the brain does and will take away many human activities in the future'

1.Can you explain how the present AI technology is dependent on previous human thought, thought either in the form of human-written algorithms or in the form of human-generated sample solutions?

There are two approaches to AI. Classical AI is usually based on algorithms that implement functional ideas devised by human programmers. So one might say that the intelligence is in the programmer’s head, not in the computer.

Neural AI, e.g., in the form of Deep Learning, is trained by massive amounts of sample data, for instance millions of photos, each one labeled by humans with the types of objects shown. The learning system replicates that and in doing so just relies on the statistics of those sample sets.

In both cases all insight and intelligence is really outside the system, which on its own doesn’t understand anything, just follows either the instructions given in a program or the statistics of examples.

2. How do you connect neuroscience and AI?

AI attempts to do what the brain does. So it seems natural to learn through neuroscience from the brain the essential principles that underlie intelligence. In doing so it is important, though, to distinguish what in the nervous system is essential for the function from what should be left behind.

3. How do the biological features of humans influence the development of AI?

Let me give an example. Children learn within two or three years to understand their environment, act in it, speak, and learn new words or things or actions from very few examples. All of this takes surprisingly little genetic or environmental information. This observation, taken as one of the biological features you speak about, will have a very potent influence on further development of AI.

AI2FUTURE

Cristoph von der Malsburg physicist and neurobiologist from Frankfurt Institute for Advanced Studies

4. What are all the aspects in which AI will improve our daily lives?

One important aspect of AI to improve our daily lives is or will be to do away with tedious, time-consuming activities. One good example are navigation systems, which help us navigate and get around without reference to maps, train schedules, asking around and so on.

5. How to achieve better trust with humans and AI, especially considering privacy violations as it happened regarding the use of some technologies as facial recognition?

AI is very open also to all kinds of unpleasant or even evil applications, for instance infringement on our privacy. The only means against that is rules and regulations to be installed by our governments.

6. We witnessed some inventions as the 'robot priest' or the 'robot tv presenter'. Will the AI and machine ever be able to replace human thought? What is the limit of consciousness in AI developed machine?

For machines to be truly intelligent they will have to have some level of consciousness: the ability to coordinate all their senses, memory, intentions, representation of the current situation and their actuators. Thus they will have to have thought and consciousness. But it may be very different from us humans in detail.

7. What is gonna be the biggest societal impact of AI?

In the near term, with the present technology, the effect of AI will be achieving many types of activity that we as individuals or as companies or administrations have been performing in the past faster and more efficiently. This will be of great economic value. And there will be new activities that we haven’t even thought about in the past.

In the long run, however, when machines become really intelligent, they will take away from us many of the types of activity that now make up our life. I see this as a great problem.

8. We passed through few industrial and technological revolution but it seems that we are not aware enough of the impact that AI, IoT systems and machine learning have? Will this makes new difference between those who will adapt better on the new revolution and raise profit from it, and those who will be again left behind from the revolution?

I dread that indeed many menial and routine types of activity, now performed by humans to earn their bread, will be replaced by automatic systems. This has been true for many decades already and has always been compensated by new opportunities and employments. There is a big discussion going on these days whether that will be true also in the future. There is now guarantee that it will.

9. What's your comment on “singularity”: the point in time when human beings are no longer the most intelligent beings on earth?

I find the prospect that electronic organisms will one day be more intelligent than us humans very frightening!

10. Can AI represent threat to humanity? To what extent?

As you know, there are many films and books that paint dystopic futures where AI organisms are an existential threat to humanity. We will have to be very clever in making sure that electronic organisms will share our traditions of thinking in terms of ethics. Or we will even have to develop a new ethical value system that goes beyond narrow human interests, aiming at long-term stewardship of life on Earth. Hopefully of course including humans as valued part in it.