Roger Attrill Profile picture
Jun 19, 2019 10 tweets 3 min read Read on X
A layman's thread about #AI and #ethics.
This morning as I was driving in to work @BBCr4today discussed a framework for ethics in AI after the £150m Schwarzman donation to Oxford Uni for a new Institute on AI and Ethics. Whoever was speaking was asked for examples... 1/9
Example given: A self driving car bombing down the road and the brakes fail - should the car go ahead killing pedestrians in its path or swerve to avoid them and instead kill the driver (such as the occupant might be). 2/9
Seriously - how may automotive engineers groaned as the classic trolley problem was brought up as an example. I really hope they don't fritter £150m trying to 'solve' the trolley problem! But there we are - it got mentioned... 3/9
And therein lies a real problem that I, as an layman, can't get to grips with. In order to feel comfortable with any AI, I need to feel that it has no bias, no ulterior motives or 'preferences' that I do not align with. 4/9
I need to feel that the decisions made by an AI can be explained with a 'human way of thinking'. I don't want an autonomous black box that I have no understanding of. And yet what happens if an AI *can* explain itself - and I don't like the way it thinks...? 5/9
What if in the case of the trolley problem, the AI explains that the vehicle it controls would more often swerve thus killing me and my family as the occupants. Would I buy that car? 6/9
I'm torn between needing to feel that an AI is open and transparent; needing to feel that an AI would be aligned to my own needs; and needing to feel that an AI would be truly unbiased for the greater good. 7/9
Whose AI is it? And in whose best interest is an AI acting? How can multiple AI's interact with each other? Maybe no better than how people interact today already? 8/9
As the layman hearing and thinking this, it does make my head spin, because - like the trolley problem - I don't feel there is 'a solution'. It will always, and forever more, be 'more complicated than that'. 9/9
@threadreaderapp please unroll this thread, thank you :)

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Roger Attrill

Roger Attrill Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @think_ui

Dec 13, 2021
I wanted to understand more about what affects color contrast: lightness, saturation, hue, choice of foreground (black or white), and algorithm (WCAG 2.0 or APCA ).
The interaction between all these is far messier than I could have imagined.
A thread 🧵 ImageImageImageImage
So what are we looking at here.
The charts show how contrast varies for black or white text on a range of colored backgrounds when using the WCAG 2.0 color contrast ratio, or the proposed APCA color contrast algorithm coming in WCAG 3.0 (which is still years off)
There's a different chart for white or black text on a range of background colors, either varying by lightness (at constant saturation = 255) or varying by saturation (at constant lightness = 128).
Read 13 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(