Sam Altman, OpenAI’s CEO, testified on the dangers of AI and need for regulation before the US Senate this week.
Altman’s company is the creator of GPT, the underlying technology of the world’s most popular AI tool, ChatGPT. Altman testifies before Congress at a time when AI technology is disrupting creative and business industries. AI tools are manipulating artists’ voices and music, being used for cheating, and spreading misinformation.
AI Should Be Handled with Care
It’s hard for me to say AI is dangerous without qualifiers. AI is a tool. It doesn’t think, can’t think, and can’t reason. Statistical models and probability are the underpinnings of AI technology extracted from human information and knowledge.
Altman’s Senate hearing happens one week after Geoff Hinton, the godfather of AI, quits his job at Google. And, one week after Steve Wozniak goes on international news calling for regulation.
The possibilities of abuse are high, but Sam Altman came armed with a plan for regulating AI.
Altman’s Plan
Altman proposes the US government create a new agency chartered with the following tasks:
- The agency should license large AI models empowered with the ability to revoke those licenses if companies run afoul of the law and policy.
- Create safety standards for AI models and protocols used for determining their danger to society. Models would have to pass safety tests demonstrating they can’t self-replicate or exfiltrate their environments.
- Require independent audits or their performance based upon a set of benchmarks.
One comment coming out of Altman’s plan is he isn’t calling for any sort of AI transparency to demonstrate how the models work. Nor is he stating models shouldn’t use copyrighted or protected intellectual property in their training methods.
The Future
It’s hard to say whether or not regulation is in AI’s future. Many federal legislators are expressing regret they didn’t regulate social media companies after witnessing the harm they’ve posed on system. They say the won’t make the same mistake this time around.
In addition to regulation, we should be talking more about the economic impacts AI and other models will have on society. GPT has demonstrated how entire swaths of the working public can be automated using this technology. The US has a very weak social safety net, and universal basic income talks go nowhere. The reality is many industries have not adapted to technology’s constant push, and AI technology is a game changing disrupter.
Many smart people are clamoring for government to take a look at this problem.
In the meantime, the Center for Equitable AI and Machine Learning Systems (CEAMLS) and other organizations are communicating the dangers and benefits AI can bring to society. Lot of great work in bias and communications systems are being done. CEAMLS was featured on CBS News Mornings and you can watch some of the great work being done while Congress gets its act together.