Comment by methodover | Hacker News Reader

Comment by methodover | Hacker News Reader

Let’s get one thing straight: If your company is saying they use AI to solve their customer’s problems but they do not, that is lying. 100%. It’s fraud. You are absolutely right to be concerned.

I would argue that you have a moral obligation to do something about this while bearing in mind your own needs and obligations (e.g., as a breadwinner for your family).

Others have been in your situation before. Try seeking advice from Tyler Schultz, one of the first Theranos whistleblowers: https://www.linkedin.com/in/tyler-shultz-450923126

There may be others, I’d love to see comments giving other suggestions.

You could also get in touch with the Wall Street Journal— tip line is here https://www.wsj.com/tips

You could send a message directly to reporter John Carreyrou, the WSJ reporter who worked on the first big Theranos story.

Going directly to the press has some serious risks — legal risks especially. That’s why you might first begin by merely seeking advice and input from people who have worked on this before.

Many people will tell you that the legal risks mean you shouldn’t do something. I understand that position. At the same time, I believe we human beings have an obligation to each other to deal fairly and honestly, and to help bend the world in that direction. It’s not an easy position for you to be in. Good luck.

It's possible OP is in a Theranos type situation, and that's a good thing to look out for.

It's also conceivable the company is doing Wizard of Oz exercises (e.g., having humans simulate what they ultimately want software to do), to refine their model and build ML training data, which is an entirely valid approach, but they'd have to be upfront about that.

Upfront to users, and especially, investors. For example, if someone were telling an investor that they have a DNN servicing the customer requests, but it's actually humans, that would create a very difficult situation, and fixing it would probably include consulting a lawyer. In that scenario, there might be others in the company who are pissed at the person who ran off their mouth, but not want to take a hit of fixing it in a good way (and then things can get worse; "the coverup is worse than the crime" is a thing).

I suspect OP's situation is much more easily corrected, or might even just be a small misunderstanding.

Wizard of Oz experiments don't work if you tell the user (from a user study perspective). You could tell them afterwards, but depending on the product / service that may not make sense.

This isnt necessarily an experiment, users' perception of how might not be a concern at this time, and there's different kinds of Wizard of Oz exercises.

For example, when I first learned the term (in HCI, or human factors engineering), they used the example of a mockup by photocopier designers, in which obviously the people using it knew there was a human moving sheets of paper to the slots.

Maybe there's a better term?

This is silly. Companies call all sorts of things AI that aren’t - from IBM to your smallest startup. It’s a nothing phrase

furthermore, it's not a lie if they believe it

Reply on news.ycombinator.com