Description du poste
The discourse in Europe around AI has finally been evolving. The past few months have seen high-quality scenarios and proposals by interdisciplinary elites as Europeans are realising that the often touted platitude of Europe having “missed the train on AI” cannot be solved by waiting for the next one to pass. In fact, there is no next train to Narnia, and we are now getting ready to chase after it. But the train is not going to Narnia, and we, along with everyone else, actually got shoved in the hopper car with the gravel and the GPUs, hoping the driver knows not to speed into the chicane coming up.
I believe the timelines and risk perception of the people who drafted and advised on these proposals, well-meaning as they are, still lag behind and have not incorporated all the information we currently have at our disposal. It looks likely that we are heading for an intelligence explosion over the next year, and that this gap is driving proposals that are fundamentally incapable of addressing the real problem. If this is the case, one can imagine that there will warning shots, and that the political mood in Brussels will swing towards safety. It is vital that, when this happens, there are proposals to be put on the table at short notice that say something other than "we are not in the frontier, so we cannot do anything about it".
It seems wise to remind ourselves what our (broad) goals with European policy should be:
• Survive• Retain sovereigntyThese objectives are ordered subjectively, but I trust they will not be too controversial. If most people prioritise the second over the first, it is almost never because they have different preferences, but simply because they do not believe in significant x-risk from AI.
The key trade-off for Europe is trying to stay close to the frontier to retain its sovereignty and leverage to influence and mediate it, but at the same time not take actions that would accelerate the frontier (and, similarly, not prevent itself from taking a