The Ethics of AI – A Fireside Chat

What do you think about like, the ethical use of AI in terms of training models and using models in the wild? Because it's something that concerns me that when we look in certain geographies, right, it may be that they allow AI to run wild and in other geographies they're like pounding down or right now until they understand it better. What is it that you hear about this?

You want to be a hundred percent transparent with your consumers and other stakeholders on what was involved in creating this AI artifact. And then there are other responsible AI aspects which are more downstream, which is, what is it, what is it that you are planning to use AI for?

Those ethical considerations are very important to outline so that when someone is actually building an AI system, they are thinking of data that will actually become part of this system that can be used for a variety of purposes. Especially on the training data side of the house, all the public data sets are pretty much gone. So the only data set that you're gonna have access to is your own data set.

But guess what? Your own data is not in the right place, and your own data is not discoverable. So discoverability is a much bigger pace.

And I think people do not necessarily connect the dots between responsible AI and finding data. What about generating data for training models? You can, as long as you are clear what that synthetic data is.

Synthetic data makes certain models run faster, you know, and we have seen synthetic data not being, not only being used for training, but for inference as well.

So synthetic data has its place, but synthetic data is not a replacement for real data, right? No, I agree. Data is messy, right?

And synthetic data is going to be clean because you've just created it for the particular purpose that you have in mind. So as you go out into the wild, it's not gonna be representative of what we're going to encounter.