Pittsburgh AI Safety Coalition
Pittsburgh AI Safety Coalition
Our Mission
The decisions humanity makes around AI will be the most consequential decisions in the history of our species. It's vital that we get them right. Because AI built anywhere will affect people everywhere, we will need unprecedented global coordination to meet this critical moment. Therefore, PAISC was founded to build a movement to push our representatives in Congress to work toward an international treaty governing powerful AI.
Understanding the Risks of AI
As artificial intelligence becomes more powerful, understanding its risks is more important than ever.
Progress in AI is proceeding at breakneck pace.
By one metric, capabilities are doubling every four to seven months. But for all this progress in capabilities, progress in safety and governance has lagged far behind.
In 2023, CEOs of the major AI labs, along with many other prominent figures in the field, signed a statement declaring that AI represents an existential risk on the level of nuclear war and pandemics.
The risks come in many forms, some of which include:
-
Power Concentration: A CEO, government, or bad actor with access to a sufficiently powerful AI could concentrate power and lock in authoritarianism worldwide.
-
Losing Control: Humanity could lose control of AI altogether, with a misaligned system leading to catastrophic outcomes for humanity.
-
Destabilized Information Ecosystem: AI could severely damage our already dysfunctional information ecosystem through deepfakes (photos and videos that are impossible to identify as fabricated) and targeted persuasion campaigns.
-
Unprepared Socioeconomic Systems: AI capabilities are advancing at a much faster pace than previous technologies and threaten current economic, educational, and social systems. Without enough time to adapt, AI could destabilize society.
Despite these risks, AI companies are forging ahead with further development.
Arms race dynamics have made it difficult for any single company to slow down. Even if one did, the argument goes, competitors would not, and we'd still face the same problem of development outpacing safety. And while AI development is currently concentrated in the United States and China, the chips used for training will become more widely available over time, meaning AI could eventually be developed anywhere. Because of this, the only way to regulate frontier AI sufficiently is through an international treaty governing its development and deployment.
Even though these risks are real and severe, AI does have the potential to benefit humanity profoundly. Systems like AlphaFold have already proven valuable in medical research, and more general systems could advance society in countless ways. But it doesn't make sense to take the big risks that come by rushing to those advances a little sooner.
The CEOs and investors making decisions about AI are a vanishingly small group, but their decisions will affect everyone. We must ensure that choices with consequences for all of humanity are made democratically.