World Summit A.I. Recap (1/2)
Last week I visited the World Summit AI in Amsterdam to catch up with the latest trends and developments in the field of Artificial Intelligence. Here is my recap of the event as well as some of the most memorable moments in my opinion.
We’re not there yet.
One important takeaway for me was that we still have a long way to go before A.I. gets really “intelligent”.
According to Gary Marcus, professor of Psychology and Neural Science at the New York University, A.I. is on the brink of getting stuck. His presentation was focused on disclaiming the following statement by Google Brain Co-Founder Andrew Ng:
“If a human can perform a mental task in less than a second, it’s likely a computer aided by A.I. can take over that task” – Andrew Ng
Gary Marcus argued that although a computer is good at certain tasks, it is not good in performing a lot of the more simpler ones, saying:
“if a typical person can do a mental task in a less than one second of thought, as long we can gather an enormous amount of directly relevant data, we have a fighting chance – so long as the test data isn’t too terribly different from the training data and the domain doesn’t change too much over time “- Gary Marcus
This claim was also reinforced by Laurens van der Maaten, Research Scientist at Facebook and Dmitry Kislyuk from Pinterest. Although their advances in the field of computer vision are impressive, really understanding what happens on an image is still impossible. Dmitry Kislyuk gave a good example:
”if a user takes a photo of some strawberries our models have no problem of identifying them as such, but what does the user want to do with them? Does he want recipes, or maybe a cool sweater with strawberries on them?”
The same principle applies to NLP (Natural Language Processing). Although the demo given by the British company Artificial Solutions was nice to see, it also demonstrated that computers don’t really understand the user yet. It basically tries to find previous similar conversations held by humans and responds accordingly.
Should we worry about Skynet?
An important topic during this summit was the dangers of A.I. For some speakers, the Skynet scenario is dooming, for others the danger lies in (the lack of) ethics. During the panel discussion certain aspects were discussed, for example the dangers of using autonomous weapons and cyber warfare. For most panelist, the community should be responsible for coming up with legislation/guidelines, because in most cases governments are lacking the knowledge necessary to keep up with the pace that the A.I. field is moving in.
A presentation about the project Hangzhou City Brain showed us a scary insight of the future. All available data within a city is used to track their inhabitants, combining multiple data sources like CCTV, social-media, weather and other. Impressive in terms of technology but scary on a personal level. An example they gave was that they use deep learning to identify a person and can recognize this person in other locations of the city on different days wearing different clothes.
Another use case where A.I. was used for the safety of a city was discussed in the PoliceLab session on day two. In this presentation, the National Police and the University of Amsterdam talked about their collaboration. And that the combination of the two worlds leads to more effective research.
The different opinions about the current state of A.I. and the ethics discussion were the two most interesting aspects for me at this World Summit. The presentations from Pinterest about their Visual Search, Facebook about image recognition, Netflix about their recommendations and E-bay about their developments were of course very cool to see!
Looking forward to next year!