In other words, the instructions we use to tell AI what to do do not result in the AI doing what we tell it to do? Or is it the humans who don’t understand?
The Illusion of Understanding: MIT Unmasks the Myth of AI’s Formal Specifications
As autonomous systems and artificial intelligence become increasingly common in daily life, new methods are emerging to help humans check that these systems are behaving as expected. One method, called formal specifications, uses mathematical formulas that can be translated into natural-language expressions. Some researchers claim that this method can be used to spell out decisions an AI will make in a way that is interpretable to humans.
MIT Lincoln Laboratory researchers wanted to check such claims of interpretability. Their findings point to the opposite: Formal specifications do not seem to be interpretable by humans. In the team’s study, participants were asked to check whether an AI agent’s plan would succeed in a virtual game. Presented with the formal specification of the plan, the participants were correct less than half of the time.
(Related)
Either the law will govern AI, or AI will govern the law
… Part of what makes AI so challenging to regulate is that the systems reach far beyond their technical components and specific products. The impact of AI, when seen as a knowledge structure, can be better understood as a philosophical force. Generative AI and machine learning, algorithms, and other subsets of AI, do not operate absent the context through which they are developed and implemented. They are informed and learn by digesting collective narratives and can reflect existing hierarchies based on preexisting historical, philosophical, political and socioeconomic structures. Acknowledging this allows us to visualize how AI may perpetuate inequities and antidemocratic values that constitutional democracies have sought to correct for generations.
… The U.S. Constitution is inspired by a philosophy of how to guarantee rights and constrain power. It separates and decentralizes power, and installs checks and balances, to avoid power abuses. AI must be viewed in much the same way. Both the Constitution and AI are highly philosophical. Putting them side-by-side allows us to understand how they might be in tension with each other on a philosophical level. If we look at AI as only a technology, we will miss how AI can transform into a governing philosophy that attempts to rival the governing philosophy of a constitutional democracy.
No comments:
Post a Comment