Sense and PSYOPS - Destroying the Discourse on AI Safety

Lies, LLMs and Laughingstocks

One twist in the discussion on AI safety is whether we can even have a conversation, as a nation or a species, if so many arguments run to extremes, or are being made in bad faith. The Guardian recently did a piece on Five Ways AI Might Destroy the World, reviewed here on YouTube.

Some arguments clearly seem better thought out than others, but the real concern is what happens when people take the most-dramatic positions, which suck oxygen away from real concerns and real strategies.

You can see this in self-appointed pundits insisting a superintelligent artificial intelligence inevitably means instant death at some point for our entire species.

Or, at the opposite pole, someone insisting AI is a hoax.

The are three potential drivers of this problem. One is simply bad analysis, or someone trying to shoehorn this debate into a particular worldview or ideology without thinking it through.

Another is the tendency of search algorithms to reward controversial positions with more clicks/engagement/eyeballs. The crazier and more exciting or upsetting you are, the more you’re apt to get, if not a following, then at least some outrage.

The third are people operating as foreign agents or with other questionable allegiances. If we’re having a debate over AI, and a hostile foreign power such as Russia or China wants to disrupt it, then dispatching bribed or blackmailed assets to drive “the Narrative” in strange directions, or merely off the rails, makes a great deal of sense.

Vladimir Putin has famously said of artificial intelligence that “Whoever becomes the leader in this sphere will become the ruler of the world.” China has famously created an AI based strategy for the intelligentization of warfare, which Beijing has been emphasizing for years.

If you were hostile to America and the larger, advanced democratic world, getting us to stop research in AI would make a great deal of sense, as would sabotaging our efforts in maintaining safety… if you ignored the existential risks.

Given much more powerful voices stepping into the debate, from President Biden to the G7 to NATO to the EU to leading AI developers, it will be much harder for willfully destructive actors to have an impact.

But because they can have an impact - and we all need to remain humble and flexible in a field where no one has all the answers - we may find it useful to leaven the debate with the best questions we can find, the best insights we can muster…

And to amplify the voices we find helpful, rather than divisive or deranged.