A first guide for policy-making on generative AI

She wanted to know if I had any suggestions and asked me what I thought all the new advancements meant for lawmakers. I spent a few days thinking, reading, and talking to the experts about this, and my response turned into this newsletter. So this is it!

Although GPT-4 is the flagship, it’s just one of many high-level generative AI releases in recent months: Google, Nvidia, Adobe and Baidu have all announced their own plans. In short, generative AI is what everyone is talking about. And while the technology isn’t new, its policy implications take months, if not years, to grasp.

GPT-4, released by OpenAI last week, is a large multimodal language model that uses deep learning to predict words in a sentence. It generates remarkably smooth text and can respond to images as well as word-based prompts. For paying customers, GPT-4 will now power ChatGPT, which has already been integrated into commercial applications.

The latest iteration caused a stir, and Bill Gates called it “groundbreaking” in a letter this week. However, OpenAI has also been criticized for its lack of transparency on how the model was trained and evaluated for bias.

Despite all the hype, generative AI carries significant risks. The models are trained on the toxic repository that is the internet, which means they often produce racist and sexist output. They also regularly make things up and utter them with convincing confidence. This could be a disinformation nightmare and could make scams more persuasive and prolific.

Generative AI tools are also potential threats to people’s security and privacy, and they take little notice of copyright laws. Companies using generative AI that have stolen the work of others are already being prosecuted.

Alex Engler, a governance fellow at the Brookings Institution, reflected on how policymakers should think about it and sees two main types of risk: harms from misuse and harms from commercial use. . Malicious uses of technology, like misinformation, automated hate speech and scams, “have a lot in common with content moderation,” Engler told me in an email, “and the best way to fight against these risks is probably the governance of the platform”. (If you want to know more about this, I recommend listening to this week’s Sunday Show from Tech Policy Press, where Justin Hendrix, editor and speaker on technology, media and democracy, talks with a panel of experts on whether generative AI systems should be regulated in the same way as search and recommendation algorithms. Hint: Article 230.)

Policy discussions on generative AI have so far focused on this second category: the risks associated with the commercial use of the technology, such as coding or advertising. So far, the US government has taken small but notable steps, mostly through the Federal Trade Commission (FTC). The FTC issued a statement warning companies last month urging them not to make claims about technical capabilities they can’t substantiate, such as exaggerating what AI can do. This week on his corporate blog, he used even stronger language about the risks companies should consider when using generative AI.