Totally agreed! There’s no way we can build effective AI if it is only built by an unrepresentative subset of people (and their data).
My friend Sara leads Cohere For AI, which is an open source research group who are trying to bring together researchers and scholars from all over the world without the typical prerequisites that act as a barrier to entry. It’s really cool:
I think that what is powerful about AI today, especially for the most visible and talked about use cases, is that is a force multiplier and amplifier of the current inputs and the values embedded in them.
I think LLMs are largely fed by western cultures and western value systems. And reinforce those. Because these are places where the investment flows. And because of that it’s largely shaped but those views and biases - intentional or not. For me that is a bit scary.
How do we make sure the builders, the regulators, and the inputs are diverse and varied and reflect human values and not just the values of the most resourced humans.
Love that you used a term from baseball to reference the action of the many workers at OpenAI who “stepped up to the plate” - the MLB player union has a very interesting history and its ongoing impact (MLB, MiLB) has wide implications even beyond baseball. Highly recommend studying that bit of collective action history even if baseball is not your thing.
Thank you—that is very cool, I will definitely read about it! Strong proof that even well-compensated employees can work together to shape their industry and the way it operates in the world.
Totally agreed! There’s no way we can build effective AI if it is only built by an unrepresentative subset of people (and their data).
My friend Sara leads Cohere For AI, which is an open source research group who are trying to bring together researchers and scholars from all over the world without the typical prerequisites that act as a barrier to entry. It’s really cool:
https://txt.cohere.com/c4ai-scholars-program/
I think that what is powerful about AI today, especially for the most visible and talked about use cases, is that is a force multiplier and amplifier of the current inputs and the values embedded in them.
I think LLMs are largely fed by western cultures and western value systems. And reinforce those. Because these are places where the investment flows. And because of that it’s largely shaped but those views and biases - intentional or not. For me that is a bit scary.
How do we make sure the builders, the regulators, and the inputs are diverse and varied and reflect human values and not just the values of the most resourced humans.
Love that you used a term from baseball to reference the action of the many workers at OpenAI who “stepped up to the plate” - the MLB player union has a very interesting history and its ongoing impact (MLB, MiLB) has wide implications even beyond baseball. Highly recommend studying that bit of collective action history even if baseball is not your thing.
Thank you—that is very cool, I will definitely read about it! Strong proof that even well-compensated employees can work together to shape their industry and the way it operates in the world.