[_y5hn]

Is insubordination to have a different opinion, and then offer paths to agreement between professionals?

Is insubordination to represent your responsibility area, as Ethical AI Researcher?

Is insubordination to send an email in order to establish dialogue around matters of disagreement between professionals?

[jollofricepeas]

Good point. I think its important to leave a link to the paper here for HN. Please read it or at least its intro & conclusion.

https://gofile.io/d/WfcxoF

After reading, it seems very much in line with her role as an ethicist. She merely states that there are various risks associated with processing large datasets.

Its a bit sad that they attempted to censure such an innocuous paper that without malice states “there are risks to consider.”

[tertius]

> She merely states that there are various risks associated with processing large datasets.

What's the strongest argument against her "merely" doing this?

[_y5hn]

We haven't heard any arguments, only that management can do as they please when they don't like results of AI Ethics Research. We'll likely not get anything substantial, ever, either.

As I've stated in other post, this is very, very troublesome for AI and Google.

OTOH, Google has much emphasis on ethics in AI/ML otherwise. But the whole company, especially management, need to put actions in alignment with their words on this matter.

[jkestner]

"No artificial neural network is near a point where we can talk about it having moral responsibility separate from its trainers’ and deployers’ – but we can make it sound like it does, and exonerate them, if we call it AI." https://twitter.com/shashashasha/status/1335067153402355714

[_y5hn]

Modern ethical AI / ML, is both that and beyond that. Issues range vastly in how complex algorithms can be used, misused and misinterpreted. There are also issues of uncovering hidden biases, unfairness, manipulative effects, privacy issues, risk and impact analysis, explanatory AI, who is accountable for accidents and malicious acts, etc. The list goes on covering both the legal and the political. This may not be unique to ML / AI, but is harder to quantify and prove, with more prevalent usage of complex algorithms.

https://analyticsindiamag.com/how-is-ethical-ai-different-fr...

https://www.kdnuggets.com/2016/03/ethics-machine-learning-ta...

[tertius]

What is the strongest argument that you can come up with? Not that's been presented.

[_y5hn]

My strongest argument is AGAINST managers at Google getting to micromanage, censor and to revenge-fire a researcher, just because they personally disagree with their published AI Ethics Research. A job said researcher was tasked with and leading.

Such behaviour, unless corrected, undermines Google's entire AI/ML and cloud operations.

[tertius]

That's not what I asked. If you can't argue the counter of your own position in the most honest way possible then either you're not prepared or not honest.

It's called steel manning.

[jollofricepeas]

The strongest argument for Google is it not wanting people to know that there are risks, no?

The strongest argument for her not doing it is keeping her job vs maintaining her professional responsibility as an ethicist, no?

[tertius]

That's not a strong argument. Can you come up with anything better?