World to AI: Slow down you crazy child

Say you want to ban or slow down X – a product of your choice.

You know the one thing that’ll absolutely guarantee that X thrives and its tribe grows? A clarion call to slow it down.

When the likes of Elon Musk put their names in support of a letter that calls for a pause in AI research, it’s bound to spur a reaction from other notable names like Bill Gates.

The official letter itself has been decried by some of the supposed signatories and researchers cited in it. But this was another reminder to delve into the idea of global governance for AI – the metaphorical Moby Dick that researchers, lawmakers, and businesses will likely attempt to pin down in the foreseeable future.

Image credit: Timmy Loen

Italy currently has a ban on OpenAI’s ChatGPT due to privacy concerns. The move has reportedly inspired other European countries like France, Ireland, and Germany to consider following suit.

While Italy has singled out OpenAI owing to its popularity, the maker of ChatGPT has already announced that it’s looking to resolve data privacy concerns.

These concerns spill over to other generative AI platforms too.

While rules and regulations are taking shape in the UK and the US, China’s guidelines are already out.

Besides espousing a certain world view – like having socialist values – China’s guidelines also say that the AI-generated content should be accurate, not infringe on intellectual property rights, etc. While generative AI models will need government approval before being offered to the public, the service providers will be held accountable for any violations, it adds.

Call it governance or regulation, the idea of making AI safer to use and experiment with isn’t new – neither is the debate on what this could look like.

In fact in 2018, at a meeting of Partnership for AI, a nonprofit coalition whose founding members include Amazon, Facebook (Meta), Google, Apple, and Microsoft, “the assembled engineers and philosophers couldn’t agree on the top priorities for making AI safer.”

Ironically, this was the inaugural meeting in 2018 to was do the thing they ultimately couldn’t agree on the parameters for.

One good thing came out of it: a growing database of everything related to the harms or near-harms involving AI. It’s called the AI Incident Database and this is indexing “the collective history” of harms so as to “learn from experience so we can prevent or mitigate bad outcomes.”