My aim is to enable every human being on the planet to be able to make their values tangible and be able to use them for any purpose they wish to.

When it comes to our internal systems, we’re still apes.

To reach that goal, here is how I’m thinking about effort/accuracy as trade-offs for values.

If we could administer a 500-question psychometric test to someone and observe them with a camera for a year, we’d know very accurately what that human being’s values are.

Of course, this is not possible.

So, we employ methods that are less accurate than this but require significantly less effort. Then we refine our AI model + increase datapoints to get closer to accuracy without increasing the effort beyond a certain point.