Former Google AI ethicist Timnit Gebru launches her own AI institute
protocol.com
protocol.com
Part of me is more pessimistic. Advocating for restrictions on things like facial recognition and pointing out biases so models can improve are fine goals that I would support, but I'm worried they are going to promote policies more like OpenAI's (overhyping the potential harm of AI, unfair discrimination in who to grant API access to, an insidious amount of arbitrary input/output filters and censorship) under the familiar justification of "ethics". I haven't seen a lot of evidence that said policies do any good, and I've seen quite a bit of evidence that they do real harm.
The Future of Life Institute has had some good podcasts on AI alignment, etc.