AI is inherently biased because everything the AIs know was built on this massive data set of inherently biased information.
What do I mean when I say biased information? Biased information is things like all CEOs are white middle aged men, That all nurses are women, that all labor is a foreign, right? This is biased societal norms that has found its way into this data. And it's there because if you go online, you see this bias in all of the conversations that I had, and it's not true,
it's a misrepresentation of the actual reality that we live in.
And it's just a perpetuation of these stereotypes. But when companies are trying to figure out how to deal with bias, in some cases they end up just going way too far. It is absolutely something that has to be dealt with. Because if we can deal with this bias in the data, it allows us to start to weed out through the use of AI as AIs will become just commonplace used tools. We can remove these stereotypes from the understanding that that's the reality of the situation.
And as kids start to use AI more and more in their schooling and their education and leaning on that, like we did with Google,
it gives us an opportunity to teach them that the world is like a melting pot of differing cultures and people and styles and choices and tastes.
One of the key things that I find super interesting is that when companies started looking at this biased data and the effect it has on the output from these language models.
Cause when you start to make changes to the output, to deal with this bias,
it really starts to uncover the societal cultural bias that is just ingrained in everything.
And all of that sounds bad. It's not right. It is just an uncovering of the reality of how AI's work and how they're built.