Skip to content

By Michael Szollosy on

Five reasons why you don’t need to fear the Robopocalypse (and two why you might…)

The question we’re most often asked when we talk to people about the future of robotics is, simply: ‘When will robots take over the world?’

People usually ask, I think, looking for reassurance, that this won’t actually happen, though some are genuinely wanting to be prepared, and wondering if now is the time to start stocking up on tinned food.

Of course, if you talk to any roboticist (or just about any roboticist), they will assure you that there is nothing to worry about. And should you be so lucky to meet an actual robot, you’ll realise that they aren’t at all that scary.

But, for anyone that still has lingering doubts, here’s a quick primer on why you don’t need to be afraid of robots and artificial intelligence (AI) taking over the world any time soon:

1. Robots aren’t all that smart (Yet)

Despite our remarkable advances in robots and AI, we still have a very long way to go before we are anywhere near close to creating the sort of robots and AI that you’ve come to fear in movies like the Terminator.

Whenever you doubt this, just play this video:

2. Robots and AI require a LOT of power to run

If robots do suddenly decide to take over the human race, it is Sheffield Robotics’ policy to… simply unplug them. (Seriously. We’ve sneaked that into our risk assessment forms. We’re not sure if anyone has noticed. If they have, they’ve been too afraid to question it.)

Worst case scenario: run up a flight of stairs and wait for their batteries to run out. That should take only a couple of hours.

While the human body is very efficient at converting energy, our smart machines are much less so. For example, AlphaGo, the Google DeepMind computer program that beat professional player Lee Sedol at Go in 2016, consumed almost as much energy in the week-long match as 34-year old Lee Sedol had consumed in his entire life.

This is especially true of humanoid robots. If an AI were serious about world domination, bi-pedal, laser-toting robots like Terminator are probably the least efficient way to go about it. (It would be much easier, say, for an AI to simply toxify our atmosphere. For example. But let’s not mention this. We don’t want to be giving AI ideas. Just in case.)

So if we are ever threatened with the Robopocalypse, we can probably just starve them out.

3. It’s all in your head

Simply, those robots that you’ve learned to fear, that story-tellers and film-makers have been telling you for decades are coming to get you? They don’t exist. Fritz Laing’s Maria, The Terminator, Ex-Machina, Robocroc (yes, that was a real film). None of them.

All of these robots tell us much more about we human beings and what we fear than they do about real robots. All the monsters we create in our imaginations—vampires and zombies and aliens and robots—are nothing but projections of particular things that we fear. The robots that you are afraid of aren’t real machines that may one day walk the earth. Those robots represent something else: our fear of being transformed into unfeeling machines (a fear that has been with us since the start of the Industrial Revolution 250 years ago), or our fear of our own genocidal tendencies. But there is no reason real robots or AI would evolve to be anything like this. So there’s no reason to be afraid, no matter how often The Daily Mail uses pictures like this:

Not a real thing. Just a mirror image.

4. Robots aren’t really into killing stuff

As I explained in the previous point, all those genocidal thoughts are uniquely human impulses. We imagine that since we think that way, any other rational, thinking being must as well. Our violent impulses—our hatred and fear and envy and all that—are intrinsically human, sure, but they are products of millions of years of evolution. But however artificial intelligence evolves, even if robots one day become as smart or even smarter than humans, there is no reason to think that their intelligence will be like ours, and come wrapped up in so many negative impulses.

The exciting thing is that artificial intelligence may become something wholly different from our own intelligence (or lack thereof). My bet is that they’ll be quite relaxed, and much more reasonable. In a good way. Hopefully.

5. Robots are not going to replace humans; that’s never been the idea

At Sheffield Robotics, we work a lot with ‘social robots’—robots that are built to help care for us, educate us, even be our companions, like pets. But despite a lot of fear from people and bodies (some of whom really should know better), social robots are only ever designed to supplement human relationships, never replace them. Just because we might have a robot teacher, or a robot carer or even a robot friend, that doesn’t mean that we won’t also have a human teacher, human carers or human friends.

Come meet Miro and make a new friend. And you can keep your human friends, too!

There are some things robots and AI can do very well, but there are some things that humans will (almost certainly) always do better. Roboticists are trying to build robots that will work alongside humans, to make us more efficient.

Robots won’t replace humans. But they may make us better humans.

(And two reasons why you might want to worry, just a little bit…)

1. AI is getting smarter, and it’s learning to learn…

The greatest advancements in robots and AI are being made by giving our smart machines the same properties that have driven human intelligence for so many millennia: the desire to learn, a curiosity about its environment and to understand its place in the world. In our lab in Sheffield, we are trying to create a robot with exactly these characteristics. It’s a long way from being a finished product, but we believe that in trying to build such a robot we can not only make important strides in artificial intelligence, we can also learn a lot about what it means to be human.

The version of AlphaGo that beat Lee Sedol in 2016 was trained by mimicking the moves of thousands of human players. A newer version, AlphaGo Zero, created in 2017, was programmed to teach itself the game of Go by playing games against itself, starting with completely random moves. In just three days AlphaGo Zero had surpassed the original Alpha Go, winning 100 games to 0.

AlphaGo Zero also demonstrates tremendous improvements in power consumption, using much less energy than its predecessor.

Some people speak of the technological Singularity: the point at which artificial intelligence surpasses our own to the point that it becomes self-improving, grows exponentially sparks runaway technological change far beyond our human comprehension. Some welcome such a explosion of AI; others fear it deeply.  The potential magnitude of such a development might be why in the 2014 film, Automata, there is a prohibition on robots repairing or modifying themselves.

A white robotic face with a mischievous expression
Not self-aware. Yet.

2. …and they’re learning from us

There is a very popular saying in robotics and AI:

Artificial intelligence is no match for natural stupidity.

We might take some solace in the fact that robots and AI still need human beings to program them, and to plug them in. And we might fear the day that robots and AI no longer need human beings to fulfil these basic functions. But in the meantime, robots and AI that are under human control or learning from human behaviour are demonstrating that maybe human beings are a bigger problem, and that we might not be entirely trustworthy masters of our robot slaves. (The word ‘robot’ comes from the Czech word robota, meaning ‘forced labour’, and was first used in the 1920 play R.U.R.)

There is ample evidence that AI systems inherit their programmers biases, giving us AI that is as racist or sexist, for example, as its programmers. Famously, Microsoft created a chatbot named Tay, which was sent out into Twitter to act with other (human) users. A mere sixteen hours later, Microsoft was forced to shut Tay down after it started posting vile sexist, racist, anti-Semetic and genocidal tweets, at one point calling for a race war.

All that after learning from humans for just 16 hours.

The face of Tay, a racist, sexist AI. Again, only a mirror of ourselves.

Maybe a self-taught, sentient robot completely independent of human bias and control isn’t such a bad idea after all. Maybe we should even put one in charge.

[NB: I’m not the first to suggest that. Isaac Asimov made the same suggestion, way back in 1950.]

Sheffield Robotics will bring their robots and hands-on activities to Pi: Our Robot Friends – Powered by Siemens on Saturday 20 January. Click here for more information about the event. 

Leave a comment

Your email address will not be published. Required fields are marked *