As Philip is on a well-deserved break, our Researcher Derin Kocer took over today’s newsletter.
In his annual presentation, technology analyst Benedict Evans revealed that he refuses to talk about artificial general intelligence because – like anybody else – he “doesn’t know” what it can be.
Frankly, this is true for most AI talk. Although there is a general feeling that this novel technology has the potential to change our lives, the consensus fades when it comes to what that change may look like.
At the extremes, 'doomsters' think that one day somebody will ask a chatbot to handle their taxes and the AGI will conclude that destroying humanity is the best way to complete the task (the only two things certain in life: superintelligence and taxes). 'Boomsters' assume that any day now it’ll cure cancer for us. Nobody knows who is right.
Other than European policymakers. EU lawmakers have been awkwardly boasting about writing the world’s first comprehensive AI regulation framework. Debate on how to regulate this novel technology matters a great deal. It’s not a coincidence that American conglomerates are spending so much money to shape the debate or that European startups are hiring former ministers to lobby on their behalf.
However, regulation shouldn’t come at the expense of killing innovation. Europe should’ve been aware of this: compared to the US in market value, Europe’s tech sector is just a blip. No wonder why some worry about Europe turning into an open-air museum. Moving to America is part of the business plan for many successful European startups. Regulations aren’t the only reason, but they aren’t helping – that’s for sure.
The AI Act risks being a part of this negative story. The EU could have produced a rulebook for startups to follow in the use cases of artificial intelligence, but policymakers went further. If this Act becomes the law of the continent, AI companies that build foundational models will need to “adhere to transparency requirements, including technical documentation and compliance with EU copyright law; conduct model evaluations, assess and mitigate systemic risks, conduct adversarial testing, report to the Commission on serious incidents, ensure cybersecurity and report on energy efficiency.”
It is estimated that compliance may cost around €300,000 for a 50-person startup. It will inevitably mean more lawyers and fewer AI engineers. While agreeing that a regulatory framework is needed for this technology, DigitalEurope’s Director General Cecilia Bonefeld-Dahl asks the right question: “At what cost?”
Rule Britannia
If we play our hand well, all this can be good news for Britain.
The UK is already Europe’s leading technology hub; it has a few of the world’s best universities and research institutions and, according to our survey with Mishcon de Reya, British entrepreneurs see more opportunities in AI than risks. The foundations are there to lead Europe on this novel technology.
Putting more barriers in front of our startups’ access to talent is the only thing we shouldn’t do. However, the government doesn’t seem to agree.
As I wrote in the New Statesman this week, the newly proposed immigration rules – which will increase the salary threshold for workers’ visas to £38,000 (£4,000 higher than the national average) – will disproportionately impact startups and research institutions. Most growing firms don’t have the financial resources to afford high salaries and, currently, a postdoc researcher starts at £36,000 at the University of Oxford.
Before changing immigration rules for good, the government should do three things to make it work for our inventors and innovators:
The salary threshold for recent graduates of British universities should be lowered;
Equity ownership should count towards the salary threshold;
Academic institutions and research facilities should be kept outside the new salary thresholds.
If you have any other policy suggestions to make the new immigration policy work, get in touch!
You can read the whole newsletter here, and sign up for the newsletter here.