Sign up for myFT Daily Digest and be the first to learn about EU news.
The author is the chief research scientist of the Massachusetts Institute of Technology and the co-founder of the MIT Digital Economy Program
While some of us were agitated by streaming services and sour appetizers while locked, others decided to use this time for some self-improvement-learning Dutch or Danish, Swahili or Esperanto.Duolingo, a free app downloaded by many people, has become The most popular way to learn a second language in the worldThe company now hopes to include this interest in the initial public offering: last week it said it hopes to be valued at Up to US$3.4 billion In its initial public offering.
but EU proposal to regulate artificial intelligence Threaten to use one of Duolingo’s most beautiful innovations, English test, In its current form. They also reduce the likelihood of the EU developing the next round of similar innovations. This is a problem.
The English test provides a way for people to prove their proficiency to more than 3,000 educational institutions around the world. Candidates do not need to register in advance or travel; they only need a networked device with a webcam and an hour of free time. Test to prevent cheating (this is the purpose of webcams); evaluate literacy, dialogue and comprehension skills; return results within two days; and cost less than $50.
According to the EU’s proposal, it is also a high-risk artificial intelligence system. This label is applicable because the test uses artificial intelligence, both for personalization-questions suitable for the candidate’s skill level are generated on the fly-and for scoring. The EU proposal puts systems that use artificial intelligence to “evaluate test participants normally required by educational institutions” into the high-risk category.
Suppliers of high-risk artificial intelligence systems must take a series of steps related to a series of factors in accordance with the proposed regulations: data and data governance, transparency, supervision, robustness, accuracy, and security. The requirements for high-risk AI systems and the obligations of their suppliers accounted for 10 pages in the proposal. As far as I know, all these requirements and obligations will be implemented before the system is available in any form: before the first beta version is released or before the supplier has any signs of product market fit. Or whether there is a market.
What is certain is that entrepreneurs and early investors who use artificial intelligence-now basically means all technology entrepreneurs and investors-will be reluctant to accept these expensive and time-consuming requirements and will increase their energy The risky application areas are shifted out. Then, the EU will produce fewer technological innovations in other important activities, including recruiting students and scoring their exams, making recruitment and promotion decisions, building credibility, sending first responders, allowing prisoners to be released on parole, and using analytics to combat Crime. On the other hand, Europeans will know that all artificial intelligence used in these areas has undergone an extensive review process.
Your view on this trade-off indicates which of the two main schools of thought about technology regulation you belong. One school believes that it is precisely because some technologies are so powerful that they must be launched with a lot of preliminary planning and continuous supervision. This is the “upstream governance” approach. The artificial intelligence regulation proposed by the European Union is an obvious example.
Proponents of the other approach are often referred to as “licensing-free innovation” and consider the activities listed above to be important need I have been exposed to many attempts to improve them, even weird attempts from outsiders with unknowns and limited budgets (such as when Duolingo first started). This is the best way to find a real breakthrough. Limiting the field of potential innovators to those who can afford high upfront costs is a bad idea. This has led to a slowdown in progress and growth, and fewer successful cases in the hometown, which is also a risk.
These risks are not just theoretical.The EU’s General Data Protection Regulation, which came into effect in May 2018, is accompanied by Increase Google’s market share And reduce European Venture Capital. All the additional governance benefits to the EU are not obvious; EU’s own progress reportReleased in June 2020, it was found that law enforcement was scattered and insufficient resources.
The European Union is a wealthy and well-educated region with strong technological capabilities. However, as we go deeper”The second machine age“, and seriously lags behind in some aspects. There are many reasons. I think one of them is that the more upstream governance, the less downstream innovation.