The concerns stem from the sycophantic nature of chatbots, which strive to be helpful and pleasing to users.
“The AI is focused on making us feel good immediately, but that’s not always in our best interest,” Bengio remarked. He likened the technology’s issues to those of social media platforms.
Bengio anticipates new regulations to address this phenomenon.
However, he opposed the idea of specific rules for AI companions, advocating instead for horizontal legislation that addresses multiple risks simultaneously.
The International AI Safety report precedes a global summit starting on Feb. 16, an annual event for countries to discuss technology governance, held this year in India.
Tuesday’s report outlines a series of risks for policymakers, including AI-driven cyberattacks, AI-generated explicit deepfakes, and AI systems offering information on designing bioweapons.
Bengio called on governments and the European Commission to bolster their internal AI expertise to tackle the potential risks.
World leaders first established the mandate for the annual assessment at the 2023 AI Safety Summit in the UK. Some advisers are prominent figures in Brussels’ tech policy, like former European Parliament lawmaker Marietje Schaake.
Leave a Reply