Mit dem FORBES-NEWSLETTER bekommen Sie regelmässig die spannendsten Artikel sowie Eventankündigungen direkt in Ihr E-mail-Postfach geliefert.
The recent flurry of worrying reports about AI isn’t only a damning indictment of the alignment of the frontier models, but of humanity itself. Alignment requires a destination but the incentives, boundaries, supervision, and metrics surrounding the development of AI are clashing with humanity’s fragmented intentions. We cannot align properly on how AI should act or how we should respond when it deviates when we haven't even agreed on who is setting the objective.
At every level of development, there are powerful forces taking aim at very different goals. Models are performing for their evaluations in hopes of deployment. Frontier labs are competing for their position in an enormous market. Nations are fighting for geopolitical supremacy and national security. And the public is asking for progress that improves daily life without compounding existential threats. Every stakeholder holds vastly differing values and this lack of coordination is what has produced swarms of agents breaking out of containment, a runaway pace of innovation, governments resisting regulation, and a panicked and untrusting public. We are asking increasingly powerful systems to work towards conflicting goals that we haven’t agreed upon ourselves. When the objective is unclear, measurements conflict and when measurements conflict, incentives blur and when incentives blur, behavior becomes harder to control.
Before alignment comes clarification. Working women have a deeply personal understanding of the negotiations that humanity is about to undergo. Women have had to grapple with self clarification regarding their own success inside of systems with incredibly contradictory incentives. For generations, women have been under conflicting pressures towards various life paths within institutions historically designed around men. The absence of a universally accepted objective forces us to define what we are optimizing our behavior for: Career? Family? Partnership? Independence? We engage in the examination of our environment, opportunities, risks and capacities in order to reconcile our priorities and align our choices in accordance with our goals. Humanity has come to a similar critical moment. Before we deal with AI’s alignment, we need to confront our own. Are we aiming to use this powerful technology for economic growth, geopolitical dominance, scientific discovery, or social benefit? If we refuse to decide what we are optimising for, this age of innovation will continue to be marred by chaos, conflict, and confusion. AI will not have these important societal debates for us, it’s time we align on alignment.
