European AI – Optimism Unravelling

VivaTech’s dreams are already coming apart.

  • Despite intense pressure from both European and US companies to rethink the AI Act, the EU is pushing ahead in a move that will control AI but also make life so difficult that start-ups and talent may well go elsewhere.
  • With at least 5 stands at VivaTech between them, of which 2 were dedicated to showcasing the ease and quality of life on offer in the Middle East, it is clear that the Middle East is serious about attracting talent.
  • Historically, this is something it has really struggled with, but there are signs that this is beginning to change in the UAE, particularly.
  • The AI Act has been in the works for some years, and parts of it are beginning to come into force and will continue to do so over the next few years.
  • The two main problems appear to be that it will place excessive burdens on small companies who will be unable to comply, as well as impose requirements on models that are almost impossible to meet.  
  • Model creators will be required to put restrictions on models that will prevent them from generating copyrighted content in a clear demonstration that the EU does not understand how these systems work.
  • There are two problems at the outset:
    • First, letter of the law: as copyright refers to making copies of content and then reproducing it, which is not how these systems work.
    • LLMs work by looking at the word or pixel that has just been produced and then calculating the statistically most likely word or pixel that comes next.
    • This is not really copying as defined by the law, and I suspect that unless there is something blatant like lifting of an entire passage from a published work, it will quickly get into a legal quagmire.
    • Second, control: where the EU demonstrates that it does not understand how these systems work and is asking for controls that are almost impossible to implement reliably.
    • LLMs are black boxes, and as such, their creators don’t really know how they work, making them impossible to properly control.
    • Instead, one has to place filters outside of the black box that monitors the output against a list and restricts it when it gets a hit.
    • These are pretty crude controls, and with a little bit of effort, not very difficult to get around.  
  • The net result is that, in its current form, the AI Act will make it harder for small companies to compete with those that are outside the EU, disincentivising start-ups from being present within its borders.
  • Borders are far more porous these days, and there is very little preventing these companies from packing up and moving elsewhere, as these sorts of services can be delivered from anywhere.
  • There is already substantial migration of capital from China, the UK and Europe, where the USA, Australia, Singapore and the Middle East have been the main beneficiaries.
  • It is quite likely that unless the EU rethinks some of the AI Act (that its own representatives will quietly admit is overbearing), talent will follow the capital and set up outside of the EU.
  • This would prevent the EU from catching up in AI (where it already badly lags) and allow the economic benefits to accrue elsewhere.
  • The vibe at the Vivatech trade show was supposed to light a fire under the tail of AI in Europe, but even a few weeks later, my cautious optimism already looks like it was overblown.
  • The most immediate beneficiary will be the Middle East, which sits on Europe’s doorstep and has plenty of space, money and energy to build the infrastructure for small private companies to use.
  • I suspect that we will also see an influx of talent into the USA, which has taken the opposite approach to AI and is already by far the global leader for AI outside of China.
  • There is still time for the EU to rethink the rules, but the signs are that it has no intention of doing so.

RICHARD WINDSOR

Richard is founder, owner of research company, Radio Free Mobile. He has 16 years of experience working in sell side equity research. During his 11 year tenure at Nomura Securities, he focused on the equity coverage of the Global Technology sector.