California Lawmaker Introduces the "No Robo Bosses Act" | Hacker News Reader

California Lawmaker Introduces the "No Robo Bosses Act"

contracosta.news

11 pointsPopAlongKid6 comments

> The legislation would also prohibit the use of ADS systems that use personal information of workers to “predict” what they’ll do in the future.

Don't we want to use objective systems to "predict" how employees will perform?

All this anti AI legislation is confusing. There are things like "you can't use face detection" but you can rely on humans to do it which is much more prone to error.

It's just silly. It's like a bank can make loans based on some kind of arbitrary formula but if it's "AI" its wrong. Even though the AI has the same exact inputs that can be in your arbitrary formula and you can control what inputs it has access to and even test to see if there is bias.

This fear of using technology/statistics for prediction screws over a lot of people and buckets people into overly broad buckets. It's why federal student loans for an engineering degree in the same school charge the same interest rate as a basket weaving degree from NYU. Under some bizarre perversion of egalitarianism it treats everyone and everything equal.

I think the biggest fear is lack of accountability.

We have already seen this causing real problems for real people. Please see Jeremy Howard's lecture on ethics in AI for some real world examples.

> Don't we want to use objective systems to "predict" how employees will perform?

Not really, because choosing 'objective' is inherently biased in some way

There is nothing truly "objective", since everything has to be viewed through human lenses that are by their very nature subjective.

Just as we cannot actually perceive objective reality, we cannot truly take actions based on objective data, as the mere act of viewing and interpreting that data is inherently subjective.

I don't know. An AI system is pretty objective. I mean you might not be able to interpret the weights, but I can't interpret what someone thinks of me when I'm standing in front of them. At least you can tell an AI to ignore things like race, color, religion, etc since you won't be telling them what it is.

So I'd take an AI when it comes to making decisions any day of the week.

The problem comes when information, like race or gender gets imputed in other ways, like zip code or purchasing habits, and the computer, whether it's a traditional algorithm, an ML model, or AI, outputs a decision against a protected class.

The "my algorithm isn't racist" issue isn't a new one.

Reply on news.ycombinator.com