A wide range of critics are taking aim at the San Francisco state lawmaker’s legislation, which creates a liability regime for advanced models
A bill making its way through the California State Legislature that’s intended to prevent artificial-intelligence systems from causing catastrophic effects is drawing widespread flak.
“We’re all for ‘Hey, we should make sure the risks of AI are properly dealt with,’” said James Landay, a co-founder and co-director of Stanford’s Human-centered Artificial Intelligence institute. “But that bill seems very flawed.” As it stands now, SB 1047 would apply to AI models that cost more than $100 million to train and require computing power above a threshold set by the bill. It would also apply to AI systems created by fine-tuning those models by training them on additional data using computing power that exceeds a similarly defined threshold.
The bill would allow the state attorney general to sue developers whose models cause catastrophic harm — defined as causing $500 million worth of damage or resulting in mass casualties — if they didn’t follow the bill’s safety requirements. It would also subject developers to prosecution for perjury if they falsely attest to their compliance with those requirements.
A big part of the concern is that such companies and organizations could be held liable for potentially malicious actions of those who use the models they develop in ways the developers didn’t intend. That’s going to encourage AI developers to keep tight control over their models and not make them available through open-source arrangements, through which others can freely download, study, modify and release new versions of the models, critics say.
'Unconscionable': HIV community in legal fight with Bay Area drugmaker State Supreme Court takes up case in which thousands claim harm from Gilead’s negligence “This threshold would quickly expand to cover many more models than just the largest, most cutting-edge ones being developed by tech giants,” Andreessen Horowitz partner Anjney Midha said on a recent podcast from the firm. “This would really hurt startups and burden small developers.”
Critics additionally charge that the bill is focused on the wrong things. In the wake of OpenAI’s release of ChatGPT in 2022, there were a lot of concerns raised about the potential for such sophisticated AI systems to cause catastrophic, potentially even existential harm to humanity. But much of those concerns have died down as people have become more familiar with such models and both their capabilities and shortcomings.
It doesn’t really matter all that much about when developers create systems that could cause catastrophic harms, Weil said.
Ai Artificial Intelligence Ai Safety Sb 1047
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
San Francisco Giants Pitcher Blasts Critics, Makes Bold PredictionAre the San Francisco Giants poised to make a big run down the stretch?
Read more »
Empty offices haunt downtown San Jose, Oakland and San FranciscoEmpty offices haunt the downtown districts of the Bay Area’s three largest cities.
Read more »
Missing girl found dead at San Jose recycling center after father found dead in San FranciscoSan Jose Police Department is investigating the death of a 3-year-old girl who was reported missing and then found dead at a recycling center, not long after her father was found dead in another part of the Bay Area.
Read more »
'Aggressive' Padres Reportedly Eyeing Reunion with San Francisco Giants StarThe San Diego Padres are considering a reunion with this San Francisco Giants ace.
Read more »
San Francisco mayor suggests paying welfare recipients extra $100 a week to stay soberToday's Video Headlines: 7/31/24
Read more »
San Francisco man convicted of chasing down, shooting at USPS workerA San Francisco man was sentenced to 11 years after pleading guilty to firing a firearm against a USPS employee while they were delivering mail.
Read more »