Writing Right with Dmitri: Designing Thought Experiments That Make a Difference
One day in June, some of us from the h2g2 Core Team were part of the virtual audience for a panel discussion at the CogX conference. The moderator was our Robbie Stamp (who did a great job, of course), and the topic was 'The Day of Judgement: What does it mean to have a meaningful working relationship with a non-human actor?'
(I suppressed memories of the director who told me, 'Never work onstage with animals.' I further suppressed the memory of the story he told me about the cat and The Diary of Anne Frank, not to mention the dead pigeons and the musical in the park…)
Robbie's panel consisted of the following knowledgeable people:
Adah Parris: Futurist at The House of Griots
Lorraine Dodd: Director of Research at Cranfield University
Louise Hooper: Barrister at Humancentech
Tony Fish: Chief Digital Officer at Digital20
Robbie Stamp: CEO and Founder – Bioss
Robbie and his colleagues were talking about AI, of course: Artificial Intelligence. It's being widely used in a variety of high-tech enterprises that will soon affect our daily lives, if they don't already. And there are some serious issues with the way AI is employed. Some of these issues are practical. Some are ethical and moral. Some might even be life-threatening. The think-tank experts take these things seriously, as they should.
The panel offered us some thought experiments, which I will try to reproduce here from my notes. The thought experiments were used as springboards for discussion among the panelists and to provoke response in the Q & A session that followed. I'm going to suggest to you writers that you:
- a. Pick one. and
- b. Write a short story around it, or a song or poem, or create an image or cartoon to illustrate your thoughts on the subject.
Robbie started by reminding everyone of Douglas Adams' parable of the puddle. In this thought experiment, Adams imagines a puddle waking up to its situation one day. 'My,' says the puddle. 'This space around me fits perfectly! The universe must be designed with me in mind!' Robbie thinks the moral of this is that humility is a good thing in puddles and humans. Discuss.
Louise Hooper: This speaker addressed the issue of accountability in AI systems. If they make the decisions, how can you control them if they do wrong? You can't make a computer go and stand in the corner.
Thought experiment: Countries A and B train an antiterrorist algorithm. They contract with a private company to carry it out. Country A makes changes in the alogrithm. Two terrorists in Country A, get past AI. Country B stops the terrorist because, when a question arises, final responsibility is passed to humans.
Lorraine Dodd: How do AI systems function in an advisory capacity?
Thought experiment: An experienced bomb disposal officer in Afghanistan has had a long day: 5 incidents (successfully concluded), one more job to do (looks easy). His vehicle stops: IEDs ahead. He decides to dispense with a bomb suit – after all, it's hot as hades out there. The bombs are all laid out in a row, nicely. This should be a piece of cake. Warning, Will Robinson, says his AI. That's not procedure. You could get yourself blown to smithereens. Who's right here, man or machine?
Adah Parris: Who has Authority?
Thought experiment: An HR department hires freelancers based on computer interactions. A human operator is aware of complaints that a freelancer has told racist jokes, but regards them as 'harmless'. The AI is not informed, and reassigns the freelancer to the offended client. Loss of business results.
Tony Fish: Agency. (Human in the loop.)
Thought experiment: Your smartphone is given agency in your life. It is allowed to make decisions for you. After all, it has your best interests at heart. Can you trust it to work on your behalf?
Robbie Stamp concluded by reading the poem 'All Watched Over by Machines of Loving Grace' by Richard Brautigan.
Thought experiment: Prisoners in a jail are allotted cells and cellmates by algorithm. A guard has a bad feeling about one cell assignment. How does she react? Who should be trusted, the human with intuition or the machine with the scientific algorithm?
You can put your story here, or send it to the Post, or post it in an entry and give us the number. Please consider giving it a try: we can send these to Robbie to share with his team. Here's your chance to have some input with the experts.