🇪🇺 The EU has some good ideas about AI regulation and the hits just keep coming for Google
🗳️ An automated algo decides on American Congressional representation
Hello all! Putting The Machinocene Review on a hiatus while I get through some deadlines, hopefully it won’t be too long. Stay tuned.
The Good
The Verge got their hands on a draft of the EU’s AI regulations. I, for one, am pleasantly surprised. Specifically these areas:
Bans on AI for “indiscriminate surveillance” and for systems for social credit scores. Special authorization will be needed for “remote biometric identification systems” like facial recognition in public spaces
New oversight for “high-risk” AI systems, like self-driving cars, and any that have a high chance of affecting someone’s livelihood, like those used for job hiring
There will also be assessments for high-risk systems before they’re put into service, which might include human overseers and checking to ensure the algos are trained on “high quality” datasets that were tested for bias
The oversight and assessment could end up being too weak to matter, but it’s a decent starting place and a model for other countries. I would have liked to seen rules that could potentially make it easier to create algos for the EU that get exported everywhere (like how California fuel standards can push car companies to just make all cars acceptable in California, not just the specific ones going there). But perhaps the assessments could do this?
The bans can always be revisited once these systems are tested and improved upon. I still have hope that we can reduce the impact of human bias on marginalized groups with carefully designed systems. I just don’t think what we have now is it.
+The European Commision has its own Giphy account with some ~weird~ gifs.
The Bad
Listen, you guys, I tried not to overwhelm you with this Google story. I didn’t even write about the researchers who have quit, or the professor who turned down Google money, or the conference that returned previously accepted funds from the company. But the continued fallout of Google’s firing of one, and then the other, ethics leads just keeps going.
Now, Trillium Asset Management, an Alphabet shareholder, has filed a shareholder resolution asking the board of directors for a third-party review of whistleblower protections at Google. Of course, there is a good-for-business angle: “Whistleblowers protect investors, not management,” according to the chief advocacy officer at Trillium. But still, you don’t typically think of an investment firm being activist shareholders over whistleblowers, so this is a pretty telling indictment of how badly this whole thing has been handled and perceived.
+TIL: some companies have chief advocacy officers!!
More News
Excellent long (looong) read about how an algorithm has determined our Congressional representation for a hundred years.
One of the first men falsely arrested because of face rec tech is suing the police in a case that might end up defining how AI like this is used in the future.
This cat does not exist. This rental does not exist. This music video does not exist. Does anything exist?
Insurers are using automated tools to estimate car repair costs but the folks actually responsible for the fixes say the numbers can be way off.
The FDA just approved an AI-powered tool that could help suss out colon cancer before it strikes.
***
Until the Jackie Snow corporation hires a chief advocacy officer that insists on a review of my weekends,
Jackie