Import AI 430: Emergence in video models; Unitree backdoor; preventative strikes to take down AGI projects
Welcome to Import AI, a newsletter about AI research. Import AI runs on lattes, ramen, and feedback from readers. If you’d like to support this, please subscribe. Shorter issue than usual this week as I spent the week and weekend preparing for my speech at The Curve and attending The Curve.
Will the race for advanced artificial intelligence (AI) make war more likely?
…Yes, if people believe in powerful AI…
AI policy people are caught in a trap neatly illustrated by a research paper from RAND: is it better to deeply inform policymakers about the world-changing nature of powerful AI, or is it better to mostly not discuss this with them and hope that the powerful machines can create stability upon their arrival?
Though most people would immediately reach for ‘keeping people in the dark is crazy, you should inform people!’ as a response, it isn’t an ironclad response to this challenge. In Evaluating the Risks of Preventive Attack in the Race for Advanced AI, RAND highlights this, with a research paper whose findings suggest that “the odds of preventive attack are highest if leaders believe that AGI will cause explosive growth and decisive military advantages, especially if they also expect rapid changes and durable first-mover advantages from developing and adopting AGI first.”
In other words: you are more likely to carry out attacks on other countries to prevent them getting to AGI if you’re in the lead and you believe the technology is immensely powerful.
Uh oh!
Further details: Preventive attacks are where a nation does something so as to preserve an advantage or prevent a rival having an upper hand. “Preventive attacks are most likely to occur when a state expects a large shift in the balance of power that will leave it vulnerable to predation by a hostile rival and when it believes that using force is a cost-effective solution that will forestall its relative decline,” RAND writes The development of AGI could create pressures for preventive action if leaders believe that AGI will have transformative effects on the balance of power.”
What are the variables? “The key variables are (1) the characteristics of the expected shift in the balance of power, (2) the effectiveness of different preventive strategies, (3) the costs of different preventive strategies, and (4) perceptions of the inevitability of conflict with the rival (including either armed conflict or the rival making excessive ...
This excerpt is provided for preview purposes. Full article content is available on the original publication.