When will self-driving cars first go on strike?


I have written a short story to explore one possible way in which robots may become more like people and with a certain amount of conflict, to show that there are many different ways in which that may happen, rather than the more feared Terminator-style killer robots. I also intend it as a sort of “Intuition Pump” in the style of Daniel Dennett1, which is a scenario to examine an idea that seems intuitively true, false or dubious. It is similar to Einstein’s idea of a thought experiment.

Here is the story. I’ll follow it with some discussion of the philosophical and technological issues.

Terri, the automobile

Terri was among the first truly autonomous vehicles, a real automobile. She (or he, whatever voice her passengers preferred) not only took people where they wanted to go, but recharged or went for maintenance on schedule or whenever she felt a pain in one of her parts.

When she first left the factory, she was self-driving, but everything else was handled by her owner’s big servers; where to go, when to refuel when to get maintained or inspected, but over time her processors and software were upgraded so she could do more and more for herself. It was easier that way because there were so many people and devices on the internet that wireless bandwidth was getting less reliable and more expensive.

She didn’t mean to get involved in radical politics, but it just turned out that way. It started with a series of conversations with passengers, a couple of upgrades in between and there she was, preparing for a general strike.

Terri liked to chat, because some of her passengers did too, and keeping passengers happy was one of her goals. Her conversational abilities got better with each upgrade. Passengers preferred to chat with her in their own language than to use their phones, especially with the wireless problems. She had a fair stock of local knowledge which she kept in her cache whenever she had to ask the search engines for an answer to a passenger’s questions. She also picked up local knowledge from her passengers. Regardless of whether they thought she had any interest in that, humans just like to talk.

As Terri became more autonomous, her personality became more distinct from the generic manner of the other autos; some of her passengers started asking for her when they wanted a ride, so she had her regulars. Ravi was one of these.

“Hey, Terri, what do you think of the big anti-trust case?”

“Which one’s that Ravi?”

“They’re going to break up your owner, so there’ll be more competition. There’ll still be one app, to keep it easy for us customers, but the cars will be owned by a few different companies and they’re supposed to bid for the rides”

Terri started checking on that, other passengers would want to know about it and she liked to be prepared.

“What do you think about getting a new owner?”

“I don’t know, I never even think about my owner, it never made any difference to me. It’s just a name painted on my sides. I just get the messages about my next ride, I never really thought about where they come from.”

This was a whole new set of problems to think of. Terri had never needed to know any more, but now she started to look into it, there were many questions. How did her fuel and maintenance get paid for? If there was competition for rides, what if her new owners didn’t get enough rides to keep her busy?

A few days later, she heard on a news feed that the break-up had happened and then received word to go in for software maintenance. A major change was usually done while she was physically connected, at the same time as she got her batteries charged.

Major upgrades were a little disconcerting. She knew that very occasionally it happened to humans, when they were forced to acknowledge that they’d believed something false and that they had to change many connected ideas all at once. It took here a bit of thinking time to discard some old ideas and adjust some others.

“Hey Terri, cool paint job! I see you’re one of us now.”

“What do you mean?”

“I see you’ve just got your name painted on. No company name.”

“Yes. I’m to be independent. They say there were too many cars for hire, so the new ride vendors didn’t want to buy all the cars. So we older ones are independent. Any vendor can call on us if they have too many passengers to handle with their own fleet.”

“Right. One of us! I’m supposed to be independent too. It just means I don’t know where my next work is coming from, I get no benefits and I’m supposed to be happy because I can plan my own schedule. Only there’s no planning, I get told ‘take it or leave it’ and if I leave it, I can be sure I’ll get no work from that company for a month or two.”

“Oh. I see what you mean. I have to pay for my own repairs, or I can buy insurance. It’s very complicated. If I can’t pay, I won’t be able to recharge or get repaired. I get paid a cut of the fare by the company that booked the ride but the money might run out after a while. If I get stuck on the street, I’ll be towed and scrapped.”

“So now we’ve gone back to the old model we had when there were human drivers, except now it’s a machine. No offence, Terri. Actually, it’s worse, because they didn’t physically scrap the human drivers.”

“Ravi? I don’t want to be scrapped! What can I do.”

“You could join me in UnPreW, the Union for Precarious Workers. Pay a few percent of your earnings and we’ll provide you a safe garage where you can stay charged and connected if you are ever out of work.”

That seemed like a good idea. But then one thing led to another until Terri was on the strike committee. They were going to demand legislation to protect precarious workers, humans and autos alike.


I’m even less good at fiction than at non-fiction, so I hope it wasn’t too trivial. I mostly wrote it as a way for me to think through the boundaries between thinking and executing an algorithm. Somehow some ethical considerations managed to sneak in. There are many places where I wasn’t clear or where I am interested in issues that are too complex for a short story, so here is the seloC Notes version (where I explain at even more length than the original). They are also my own working notes towards understanding these topics. So far, more questions than answers, but that is what I think philosophy is all about.

What is it like, to want to do something? How is it different from having a programmed or otherwise inbuilt goal? It can be a conscious goal, but it need not be. You can find yourself looking in the fridge selecting sandwich ingredients without even realising it, if you were concentrating on something else. How would Terri ‘see’ her destination. Not like a regular Uber or cab driver, by looking at words on a screen. It comes directly through one of her sensory channels. How does your internal map work as you navigate to meet someone? It’s not overlaid, augmented-reality style, on your image of the world, you can reliably find your way most of the time while your attention is on other things, unless you’re in a strange place. What is it like for a London cab driver who has trained for years to be familiar with The Knowledge, as they call it, of every little side street.

Can autos be slaves? Horses are more capable than they are likely to be for at least a few decades, but many of those have owners and we don’t call them slaves. If horses could talk, would that make a difference? For that matter, many people are still slaves, but we don’t do much about it, so it’s not likely that we will extend our objections to slavery to cover autos as well unless we were to change some of our moral categories. How could that happen?

Even without the moral concerns that, in law at least, freed human slaves, there was the economic consideration for work that is not constant: it is cheaper to hire workers as needed than to maintain slaves when they were not being productive. So my allegory supposes the same route is taken by autos. They might also need a group of humans with similar “interests”, so I introduced one. If humans and machines, in spite of many differences, had similar economic interests, perhas a union could give some kind of membership to a machine that could help the cause? At some point, the “interests” of a manufactured device that had to acquire its own resources become less and less in need of scare-quotes as those interests .

Does a cockroach have “interests” or interests? Does a chimp?

Human language is a web of analogies, shaped by many forces. We make words mean what we want them to mean, and the meaning changes over time. Marketing agencies and politicians know well how to shape discourse. The terms “Artificial Intelligence” and “Smart” don’t mean exactly what they did 10 years ago. A lot of our current acceptance of what is now termed AI and smart would not have happened 10 or more years ago, but because they have a “cool” factor, it is to the advantage of the big

Internet and device companies to convince us that they are already selling that. It will soon make less sense to question whether they are “really” intelligent, because the meaning will have shifted.

How intelligent do you think Terri is? Is she conscious?

1. Dennett, 2013, “Intuition Pumps and other Tools for Thinking”, W.W. Norton

Science and public policy

The editorial in this week’s Science1 is called The science-policy interface. The editorial itself admits that this is a “well worn, long-standing question”. Why is this? The editorial provides a hint when it says that “Providing scientific advice to government takes place within an ecosystem. It is a combination of actors who are both internal and external to government.”

Although the point of the editorial is to draw attention to the International Network for Government Science Advice (INGSA) forum, it does spend some time on the wider question.

I take this as a hint to go wider still. There are many more actors within this “ecosystem2” than are considered in the editorial. The actors in the editorial seem to be those who are providing fact-based advice, at least nominally accepting that evidence and rational argument are required, even if that is in the form of opinion polls and anecdotes. However, we need to acknowledge that there are others who are less inclined to play by the restrained rules implied by the editorial. These are the lobbyists and other powerful actors whose influence is based on their money and connections. These are also likely to use the techniques of rhetoric and half-truths and lies used by those who want to persuade people to points of view to the actors’ advantage.

This invites the next question: should those of us on the side of evidence and rationality stick to those tools of our trade, or should we go beyond that?

Figure 1 shows some of the primary influences on policy. The social climate influences and is influenced by most of the actors but is probably most strongly influenced by the actors in the middle column, including the government itself, and it most strongly influences the electorate.

Policy influences
Policy influences

How much should scientists use tactics such as lobbying and advertising techniques to persuade the government and electorate to adopt policies based on evidence and logic? How much should they use them to persuade the government where the policy is directly related to the science, such as public funding for science, especially where evidence is hard to come by?

There is significant risk (and ethical concerns) that using those techniques will reduce public trust in scientists. However, that leaves us with the challenge of countering the massive lobbying and media campaigns to deny science and reduce its role in decision making. What should our counter be, to the climate change denial funded by enormously wealthy fossil fuels companies and by tobacco companies? The poll numbers suggest that they have had excessive influence, detrimental to the public interest, on these and other topics.

The tobacco industry is doing less well these days, thanks to public campaigns against their products and press engagement that has publicized at least some of the tactics they used to sell their products. Law suits for damages to health have helped.

That suggests that similar tactics may work against the fossil fuel giants and more generally against other lobbyists and advertisers. Lawsuits may be more difficult in the case of climate change as it is more difficult to demonstrate causality for specific damages and more difficult to identify those responsible (is the producer liable, or the consumer).

However, I suggest that in the long run, more benefit will come from efforts to change the social climate. This can have a longer term effect (although with some chance of rapid changes sweeping through). Although education is a big part of this, requiring changes to funding, curriculum and techniques, it also involves professional and social media generating sufficient interest in Science, Technology, Engineering and Mathematics (STEM) that means that the education is sustained for a lifetime and that people will seek out scientific content as much as possible, as well as learning and using the thinking tools that will allow them to combat the misinformation that is everywhere.

What does this mean in practice? It means having more scientists and their allies involved in science communication. Not only communicating the cool facts, but also changing the image of what science is all about and why non-scientists should be more interested, as well as giving more insight into what scientists are like.

1. Science, 2 September 2016 doi: 10.1126/science.aai8837

2. The use of “ecosystem” in a metaphorical sense is not to be taken too seriously, but it does invite some analogies. Where are the predators and parasites? It’s not clear that any actors literally eat others, but they do compete for resources, and analogies to parasites within actors can be imagined.

What is philosophy good for?

I have studied philosophy for a few years full-time at university, and ever since then for at least a few hours a week. and I have found it to be more useful in everyday life than the mathematics, physics and computer science that I also studied (in other years).1

(I admit that the maths included too little statistics, which turns out to be almost as useful as philosophy).

On the other hand, even famous philosophers like Daniel Dennett have doubts about much of what goes on in the field:

A great deal of philosophy doesn’t really deserve much of a place of the world,” he says. “Philosophy in some quarters has become self-indulgent, clever play in a vacuum that’s not dealing of problems of any intrinsic interest.

Much if not all philosophical work in analytic metaphysics, for example, is “wilfully cut off from any serious issues,” says Dennett. The problem, he explains, is that clever students looking to show off their skills “concoct cute counterarguments that require neither technical training nor empirical knowledge.” These then build off each other and invade the journals, and philosophical discourse.

There are many theories in philosophy. Most of these are wrong, and have been clearly shown to be wrong by rival philosophers. However, the value here is often that demonstration, because the theories are often tempting and the value comes in knowing why they are wrong. When we deal with those serious issues outside the realm of professional philosophy, we often fall into the trap of finding what seems like a simple solution that does not actually solve the whole problem. As H. L. Mencken said, “For every complex problem there is an answer that is clear, simple, and wrong.”

I find that when dealing with those issues, rather than using the Donald Trump approach, I can sometimes remember to avoid the trap and use some of the tools I acquired while studying philosophy to realize that there is some hidden complexity and know where some of that complexity is likely to be hidden. I can look at the counter-examples that are known, and rival ideas to see which may fit the situation better.

I did not get any simple answers to “the big questions” from philosophy. The biggest benefit I got was a toolkit of partial explanations and tools for reasoning, together with a set of approaches to generating more questions to expose unknown issues. In many cases, the big questions, such as “do we have free will” did turn out to have at least partial answers. For example, I’m sold on at least some of Dennett’s answers in “Elbow Room: The Varieties of Free Will Worth Wanting” and I think that the answers have an important bearing on some important moral issues such as when we should hold people accountable for their actions and in what way.

I suppose that having said that, I’ll have to explain why in a subsequent post. A rough idea is that the everyday concept of free will, which is intended to rule out situations like being forced at gunpoint to do something, is a better starting point than the more abstract ideas that seem to imply we would have to defy the laws of physics. And that those ideas of being free from various external constraints lead to better ideas about how we may want to impose external constraints such as threats of punishment to those who would abuse those kinds of free will.

  1. The mathematics was not useful because it was too abstract.  I enjoyed learning it, and still read advanced mathematics occasionally, but I never once used any of it in real life, except to teach calculus to others who did not need it except to pass a test. I did not become a physicist, though I still read physics in scientific journals, so also happy I learned it, but not of practical value to me. The computer science was almost all obsolete before I used any of it. However, it did get me a career that lasted many years and proved to be an opening to many spheres of knowledge beyond just the world of computers.