Zoomposium with Prof. Dr. Konrad Kording: “How do machines think?”

Information about the person

In another episode of our “Zoomposium series” on the topic of “AI research”, my colleague Axel Stöcker from the “Blog der großen Fragen” and I had the opportunity to interview the well-known and renowned German physicist and neuroscientist Prof. Dr. Konrad Kording.

I got to know Konrad through a correspondence on a social media platform. I was already familiar with his name in connection with one of our last Zoomposium interviews with Patrick Krauß, as he also mentioned Konrad’s often-cited article “Could a Neuroscientist Understand a Microprocessor?” (2017) was mentioned.

In this well-known article, he developed an information technology experiment together with Eric Jonas. This involved running old Atari games such as “Donkey Kong” and “Space Invaders” on a very simple, old MOS 6502 processor (3510 transistors), which was used in the Apple I or Commodore 64, for example. The dynamic processes of this “very simple, artificial organism” were then analyzed using neuroscientific methods in order to obtain information about what is actually “happening” — in the truest sense of the word — in its “logic gate network”.

One result was:

“A real understanding of even these comparatively simple human-made systems therefore remains elusive, and sometimes they even surprise us with genuinely surprising properties [70]. The brain is clearly much more complicated, and our difficulties in understanding deep learning may suggest that the brain is hard to understand when it applies something like gradient descent to a cost function.” (https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1005268, emphasis added)

This exciting topic on “How do machines think?” was one more reason for me to ask Konrad for an interview on the possible “cognitive processes on machines”, which he kindly complied with.

More about this: https://philosophies.de/index.php/2024/01/25/wie-denken-maschinen/

Watch the full interview in German (automated translation is available via Youtube subtitles):