Lisa Braun29
The idea of bias in AI is really interesting (and scary). @GlenPRobinson can you talk more about that. How can you spot negative/harmful bias in AI and address it? @JerryAOverton ?
Glen Robinson
What a great question. We are seeing the emergence of AI ethical audits which is a sorely needed approach to ensuring the potential bias doesn't create risk.
Glen Robinson
Just recently Amazon found that their cv screening bot was filtering out women, because it had the bias in recruiting men, as that was the majority of the gender recruited recently. That was a poor outcome.
Lisa Braun
@GlenPRobinson So this has to do with ML and training and retraining the AI. But I have a feeling that, just like the news, there will always be some amount of bias even though the aim is to be neutral/balanced.
Nathalie Vancluysen
@GlenPRobinson so that's the reason why we have so few women in IT! :-)
Dan Hushon
I think that the best that you can do is continue to audit to find the gaps and then reinforce. Machines Learn from the data given so we may need to train with some change intentional data?
Sukhi Gill
This is where Sociology meets Technology. There is a need to understand the Diversity Bias as you train.
Graham Chastney
I think that there's also a need to recognise the amount of bias already built into organisations and the decisions that they take. We tend to forgive ourselves as humans and judge the machines more harshly.
Graham Chastney
@sukhigill Agreed, I think that there are huge opportunities for those with sociology and psychology skills in the future platform world.
Lisa Braun
@GrahamChastney Interesting point...judging the machines more harshly. Indeed we call come with biases.
Frederic Auberson
I was reading "Ethical by Design - A Manifesto", which raises that very point: Not only deploying biased AI, but also letting it drift because we're not continuously monitoring its performance is an ethical issue (https://goo.gl/V4KiQR)
(edited)
Bill Murray
Natalie, I can think of five horrible reasons there is not enough gender equality in IT, and my colleague Caitlin Macdonald and I will be putting out a piece on diversity it Tech very, very soon.
Lisa Braun
Thanks for that link, @Frederic. Looks interesting.
Chris Swan
I was talking about this recently at a customer event in Vienna. Some of the problem comes from our (natural) language, and the fact that it carries many biases (e.g. patriarchy)
Nathalie Vancluysen
@BillMurray500 looking forward to that! we can think about plenty of reasons... but how do we solve it?
Bill Murray
Frederic, can I commend a thoughtful piece by my colleague Caitlin McDonald ...The Winter of AI discontent!
https://leadingedgeforum.com/publication/emergent-...
https://leadingedgeforum.com/publication/emergent-...
The Winter of AI Discontent: Emergent Trends in Algorithmic Ethics
We’re not fighting a battle on tech ethics: we’re fighting a battle on ethics, period.
We’re not fighting a battle on tech ethics: we’re fighting a battle on ethics, period.
Dan Hushon
@GrahamChastney totally agree on organizations innate bias... reasons the next challenge which is understanding the need for dimensions in which you evaluate diversity
Bill Murray
Interestingly Natalie the Norwegians just experimented with quotas and it went well. It's not just an issue of fixing the pipeline; the leaders are fixing progression, retention, pay equality and representation all at once.