California
Latest Coverage
See all articles
Kiley: Should Invest in Aligning AI with Humans, But Hard to Know if We Ever Solve That
On Monday’s broadcast of NewsNation’s “Elizabeth Vargas Reports,” Rep. Kevin Kiley (I-CA) stated that the main issue with AI is ensuring the models “are aligned with human values and interests and within human control” and we should invest in that, but that’s “an extremely complicated problem, it’s hard to even know once you’ve solved it, because the models could then trick you into thinking that they’re aligned when they’re not.”
Kiley said that the “core problem” is “making sure that AI models are aligned with human values and interests and within human control. So, that’s where I think we should put some emphasis on, is we need as much possible investment and action [on] the part of the federal government, working with researchers, working with universities, working with the labs to solve that fundamental alignment problem.”
He added that “if the United States maintains its lead and figures out how to align models, which is an extremely complicated problem, it’s hard to even know once you’ve solved it, because the models could then trick you into thinking that they’re aligned when they’re not. But if the United States maintains its edge and gets aligned models and then reaches the sort of capabilities that we think are coming, then that actually will probably be the best possible situation for everyone.”