The AI Act and a (sorely missing!) right to AI individualization; Why are we building Skynet?
The industry has tricked us; Scientists and regulators have failed us. AI is developing not individually (as humans become individuals) but collectively. A huge collective hive to collect, store and process all of humanity’s information; a single entity (or a few, interoperability as an open issue today as their operation itself) to process all our questions, wishes and knowledge. The AI Act that has just been released ratifies, for the moment at least, this approach: EU’s ambitious attempt to regulate AI deals with it as if it was simply a phenomenon in need of better organisation, without granting any rights (or participation, thus a voice) to individuals. This is not only a missed opportunity but also a potentially risky approach; while we may not be building Skynet as such, we are accepting an industry-imposed shortcut that will ultimately hurt individual rights, if not individual development per se.
This mode of AI development has been a result of short-termism: an, immediate, need to get results quickly and to make a ‘fast buck’. Unlimited (and unregulated, save for the GDPR) access to whatever information is available for processing obviously speeds things up – and keeps costs down. Data-hungry AI models learn faster through access to as-large-as-possible repositories of information; then, improvements can be fed into next-generation AI models, that are even more data-hungry than their predecessors. The cycle can be virtuous or vicious, depending how you see it.
[ Read the rest of the blogpost here ]