“It takes 25 years to build a research institute, but only one month to destroy one”

Last night, we were very privileged to hear a talk from Hervé Bourlard, an EPFL Professor who is also Director of the Idiap, a research institute based in Martigny in Valais.

About Idiap

That it took Hervé more than an hour to present his four intro slides on Idiap underlines clearly the level of interest in its cutting-edge research.

The research focuses on five areas, where Hervé proudly told us their work is world-leading. These five areas are namely:

  • Perceptual and cognitive systems (tackling areas like speech and natural language processing)
  • Social/human behaviour (analysing people’s social media interactions)
  • Information interfaces and presentation
  • Biometric person identification (e.g. face detection, tracking and recognition)
  • Machine learning (statistical- and neural-based)

Hervé also highlighted three areas of emerging (and complementary) focus:

  • Uncertainty and optimal decision-making (which Hervé described as techniques to overcome the limitations of machine learning, which he said the institute was putting to use on initiatives like virtual crash tests for auto makers)
  • Robot learning and interaction
  • Computational bio-imaging (imaging human parts like veins which can be used to model the workings of organs and project the impact of, say, medication)

The perils of academic research

Hervé was extremely open in discussing the many areas of controversy surrounding academic research and happy to share the lessons of a lifetime in the field

  • Time – there are no shortcuts to success. As Idiap approaches its 25th anniversary in 2016, Hervé stressed how it has taken all of this time to achieve its current status and results. He also highlighted how, notwithstanding public perception, it takes time to get breakthroughs when dealing with high complex subject matters and cited the example of the US government’s multi-billion dollar 10 year programme to map and uncover the working of the human brain (the 10 years were up in 1999!)
  • Brain drain – Hervé confirmed that the stories we read in the media – like this about Uber –are not isolated cases; that the threat of Silicon Valley poaching the best people is very real and prevalent. As he said, he never fears losing the best researchers to other institutions, only to “Facebook and Google”, which has happened several times so far in the life of Idiap
  • IP – Hervé also opined on the thorny subject of IP rights. Generally an advocate of openness, Hervé has been moved in recent years to write tighter contracts and seek additional patents as the institute tries to generate more revenue from the breakthroughs it makes. Somewhat reluctantly, Hervé also agreed with the premise of work from economists like Mariana Mazzucato that private companies appropriate unfair rents on many public goods, including academic research.
  • Money – Hervé’s chief learning and his most important advice to others is to “follow the money”. Obtaining funding is hard, but it is self-evidently essential to workings of any institute and lab and it must remain top of mind at all times.

Europe’s big challenge

Hervé waxed cautious on the Europe’s prospects for innovation and future prosperity. While, he argues, academic research is better in the Europe than the rest of the world, Europe has a poor record of commercial exploitation.

When Philips was too slow to launch flash memory, he said, it ceded the hardware industry to the US (CPU) and Japan (memory). Europe has failed to build many world-beating software companies. And the digital platforms like Facebook and Google are all US. One area, incidentally, where he thought Europe still has a distinct advantage over the US is in gaming and, in light of this comment, it is interesting to note yesterday’s agreed takeover of King Digital by Activision

The weaknesses of the EU, such as fragmented markets, are well rehearsed and not worth revisiting here. And many attempts by the European Commission have been disastrous, normally when Europe attempts to build the US version of anything (e.g., Galileo)

But Hervé still sees a big role for supranational bodies like the EU to foster innovation, working in partnership with academia, member states and corporates. And, Hervé is very bullish about speech processing…

Speech processing – a beacon of hope?

Speech processing is an area where Europe can lead, argues Hervé. Hervé gives several reasons:

  • A strong research community and infrastructure in this area
  • A cooperative SME sector prepared to work together with academia (sharing data for instance)
  • A multi-lingual environment– the Americans only care about English, he said, whereas in the EU alone there are more than 50 languages

Hervé’s goal, for which he is seeking EUR20m of funding from the EU, is to create a “multilingual digital single market” in Europe. This would be a completely open system, built on open architecture, free for anyone to exploit – all in the pursuit of raising Europe’s innovation metabolic rate

To access the work that Hervé is currently undertaking with the European commission to define a roadmap for conversation technologies, please visit: http://rockit-project.eu/

To access the slides from the presentation, use this link: https://www.dropbox.com/sh/dp8xhddpv6tnoan/AABDks9OuxUe6pNcb0XXyKoBa?dl=0

Advertisements