I am also interested in this thread - particularly the bit about preventing bias and discrimination being coded in to AI (and other digital technology), and how we can possibly find algorithms for spotting this bias and flagging it up, which assumes in the first instance that we recognise that it’s there; and also the use of other bias interrupters.

Any other real examples of this would be gratefully received, as I am trying to put together case studies for undergrads at Aston at the moment. Thanks for the examples in the message below.

I agree with Thomas, there are things to worry about, concerning machine learning and AI. Sometimes these are the things reported, sometimes not.

I've recently read a few books on the topic, and highly recommend "Weapons of Math Destruction" by Cathy O'Neill. One of her case-studies is how algorithms for policing and sentencing encode and perpetuate prejudices such as racism.

Also:
"Digital Exhaust" by Dale Neef
"The rise of the Robots" by Martin Ford
"Postcapitalism" by Paul Mason
"Data and Goliath" by Bruce Schneier

And on somewhat related topics:
"The Internet is not the Answer" by Andrew Keen
"The Dark Net" by Jamie Bartlett
"Programmed Inequality: How Britain Discarded Women Technologists and lost its Edge in Computing" by Marie Hicks

I am intrigued by our use of social media in public engagement. Should we encourage the public hand privacy over to Twitter, Google and Facebook, so they can find our thoughts or sign up for our events? This could be argued either way. I would like to see more debate.

- Daniel

Steve,

I don't want to point fingers, but I feel that there are many aspects to this problem.

The first is that there are genuine fears of an existential risk that might arise from superintelligent AI (which we are at the very least decades away from even if it's possible, by a general consensus of experts. https://en.wikipedia.org/wiki/Superintelligence:_Paths,_Dangers,_Strategies is a good source for some of the academic work surrounding this.) These genuine concerns are then expressed by popular figures like Stephen Hawking and Elon Musk, and turned into soundbites of a single line that are splashed around by the media (usually with some ridiculously misleading picture of the Terminator attached.) To read some accounts, and certainly some headlines, you'd think that we already had superintelligent AI cooked up in some Google lab somewhere that's on the verge of destroying the world. This is being recklessly exaggerated - and it doesn't help that, e.g., the focus on robotics and artificial intelligence is often to blow developments out of proportion.

Take two prominent stories of late: Saudi Arabia "giving a robot citizenship"
http://www.independent.co.uk/life-style/gadgets-and-tech/news/saudi-arabia-robot-sophia-citizenship-android-riyadh-citizen-passport-future-a8021601.html

Facebook's AI "invents its own language":
http://www.telegraph.co.uk/technology/2017/08/01/facebook-shuts-robots-invent-language/


Both had major flaws in the reporting, especially when you go to less reputable publications. Nowhere, for example, in the Sophia article do they mention that all of its responses are pre-scripted; it's hardly as intelligent as you might believe at first glance.

Facebook's AI story was an interesting scientific paper about how the particular algorithms they were using deviated from what they were expecting (and, eventually, English); but "AI experiment shut down for getting too intelligent and inventing its own language" (as it was widely expressed); highly misleading.


So much for strong AI. Then we have to get onto the topic of what actually exists: weak AI (optimization algorithms) and robotics, automation. 

I think in many ways people are right to be skeptical about how well these fields can benefit them. We're being constantly warned that many jobs are in danger of becoming automated, and you don't often see many  economic or social arguments for the benefits to society as a whole. Algorithms are not always used responsibly; they can take away the human element in decision-making and hence also the responsibility. The fact that they're often used to sell us products and direct advertising in a way that some view as manipulative also gives people cause for concern. Like any new technology, there is always skepticism about the negative impacts before there is widespread acceptance; in this case, it is exacerbated by failures to understand what AI can and can't do, and various looming dystopian futures that get more coverage for being more lurid.

If anyone is working in robotics or artificial intelligence and would like to be interviewed for a science communication project that discusses the benefits and risks of artificial intelligence/robotics, do drop me a line on this email and we can get in touch.

Rant over...

Thomas

Robots scare Facebook - The Telegraph www.telegraph.co.uk Facebook shut down a pair of its artificial intelligence robots after they invented their own language.

Saudi Arabia becomes first country to make a robot into a citizen www.independent.co.uk Saudi Arabia has become the first country to give a robot citizenship. The move is an attempt to promote Saudi Arabia as a place to develop artificial intelligence – and, presumably, allow it to become a full citizen. But many pointed out that those same rights aren't afforded to many humans in the country. The robot, named Sophia, was confirmed as a Saudi citizen during a business event in Riyadh, according to an official Saudi press release.

---

From: psci-com: on public engagement with science <[log in to unmask]> on behalf of Steve Pritchard <[log in to unmask]>
Sent: 01 November 2017 14:02
To: [log in to unmask]
Subject: [PSCI-COM] Guardian: Public backlash to AI

 

Hi Sci-Com peeps,

Spotted this in today's Guardian - https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.theguardian.com%2Fscience%2F2017%2Fnov%2F01%2Fartificial-intelligence-risks-gm-style-public-backlash-experts-warn&data=02%7C01%7Cthomasa2z%40hotmail.com%7Ca442a19d91a5472d390a08d5213154ed%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C636451418106662541&sdata=fTfQtEqfvqVh8pi6trfCH5%2BLuNKQQFNGtEhbOIlkSUk%3D&reserved=0

Now I'd hope that sci-com professionals could help in this case, but it does seem that we're not learning from previous experience of failure to engage over new technology.

What's the cause of this? Funding? People not seeing the importance of good communications? Or is the message getting across - given that this is a warning in advance of the problem?

Answers on a postcard...

Steve

**********************************************************************

psci-com how-to:
Once subscribed, send emails for the list to [log in to unmask]. If not subscribed, either subscribe here https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.jiscmail.ac.uk%2Fcgi-bin%2Fwebadmin%3FA0%3Dpsci-com&data=02%7C01%7Cthomasa2z%40hotmail.com%7Ca442a19d91a5472d390a08d5213154ed%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C636451418106662541&sdata=OO3sQvZX85U5V46cX08i7LJocu5NhuRCN6J1KvOZ2wo%3D&reserved=0 or send requests for items to be posted on your behalf to [log in to unmask]

To unsubscribe (or silence messages while away) send an email (any subject) to [log in to unmask] with one of the following messages (ignoring text in brackets)

• signoff psci-com (to leave the list)
• set psci-com nomail (to stop receiving messages while on holiday)
• set psci-com mail (to resume getting messages)

Contact list owner at [log in to unmask]
Small print and JISCMail acceptable use policy https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fsites.google.com%2Fsite%2Fpscicomjiscmail%2Fthe-small-print&data=02%7C01%7Cthomasa2z%40hotmail.com%7Ca442a19d91a5472d390a08d5213154ed%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C636451418106662541&sdata=63ARQBsjdbSovxwY7Sln0q0FLAglMWTHv5csAlfahkY%3D&reserved=0

**********************************************************************

**********************************************************************

psci-com how-to: Once subscribed, send emails for the list to [log in to unmask]. If not subscribed, either subscribe here https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=psci-com or send requests for items to be posted on your behalf to [log in to unmask]

To unsubscribe (or silence messages while away) send an email (any subject) to [log in to unmask] with one of the following messages (ignoring text in brackets)

• signoff psci-com (to leave the list) • set psci-com nomail (to stop receiving messages while on holiday) • set psci-com mail (to resume getting messages)

Contact list owner at [log in to unmask] Small print and JISCMail acceptable use policy https://sites.google.com/site/pscicomjiscmail/the-small-print

**********************************************************************

-- 

Dr Daniel Barker

Institute of Evolutionary Biology

University of Edinburgh

Charlotte Auerbach Road

The Kings Buildings

Edinburgh

EH9 3FL

United Kingdom

The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. 

**********************************************************************

psci-com how-to: Once subscribed, send emails for the list to [log in to unmask]. If not subscribed, either subscribe here https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=psci-com or send requests for items to be posted on your behalf to [log in to unmask]

To unsubscribe (or silence messages while away) send an email (any subject) to [log in to unmask] with one of the following messages (ignoring text in brackets)

• signoff psci-com (to leave the list) • set psci-com nomail (to stop receiving messages while on holiday) • set psci-com mail (to resume getting messages)

Contact list owner at [log in to unmask] Small print and JISCMail acceptable use policy https://sites.google.com/site/pscicomjiscmail/the-small-print

**********************************************************************