An interesting conference last evening in Geneva on applied AI, which confirmed my perception that we are no longer in the times of spectacular news, like winning different games against human players, but that we are talking now of challenges of practical implementations. A couple of insights:
Insight #1) The conference was organized by HEPIA, which stands for Engineering College in Geneva. (Note: you should not mix this with Geneva collège, which is the same like lycée in France or high school in the US). There was an impressive lineup of professors and researchers from this school talking about various aspects and challenges of implementing AI. So, an important insight is that you no longer need PhD to work on AI solutions. It is equally possible to do an AI bachelor’s or master’s degree from a university of applied sciences like HEPIA. This fact will have a huge impact on reducing costs and go-to-market time for most of the practical applications of AI.
This (and not not this) leads to my insight #2) Scientific work needs to become more agile and crossdisciplinary. There was a nice presentation of Prof. Nabil Abdennadher of their energy project, with the conclusion that the progress of AI will depend on its social acceptance. Which implies that AI research needs to be closely interconnected with social science. And in many cases not only social science, but many other domains. Another element comes to awareness with a recent study at Stanford : Why ideas get harder to find? which points out that the productivity of scientific work falls in all domains of research. In case of semiconductors the decline is 6.8% per year. Well, all this is a big problem and scientists have to find a solution for it. One solution track is : do agile research and get inspired with ideas of one of the leading thinkers in this field Alain Aspuru Guzik who works on highly automated, self-driving labs.
In the end, the opening speech by Pierre Modet was a bit disappointing. Geneva is falling behind other major cities in the adoption of AI, and this situation calls for more attention and support from the political circles, but I didn’t hear anything concrete in this sense. We are the global center of private wealth, commodity trading, trade finance, major international organizations and global governance, and there is no organized way of gathering use cases from these sectors and channeling them to be implemented with AI. And all this despite the fact that Switzerland is quickly becoming a startup nation. So, insight #3) Development of an AI ecosystem requires conscious and comprehensive efforts. We need the breakthrough innovations, but we also need the agile infrastructure and programs to convert these innovations into business and social value. And falling behind in this domain means loss of job opportunities for the young generations in the years to come and negative impact on economic growth.
In the end, a very nice demonstration by Théophile Allard from Neural Concept of their AI solution for the automation of engineering design, which helped break the world record in bicycle speed.
Nov 1, 2019