You submitted your questions about artificial intelligence and robotics, and we put them – and some of our own – to The Conversation’s experts.
Here are your questions answered (scroll down or click on the links below):
- How plausible is human-like artificial intelligence, such as the kind often seen in films and TV?
- Automation is already replacing many jobs, from bank tellers to taxi drivers in the near future. Is it time to think about making laws to protect some of these industries?
- Where will AI be in five-to-ten years?
- Should we be concerned about military and other armed robots?
- How plausible is super-intelligent artificial intelligence?
- Given what little we know about our own minds, can we expect to intentionally create artificial consciousness?
- How do cyborgs differ (technically or conceptually) from A.I.?
- Are you generally optimistic or pessimistic about the long term future of artificial intelligence and its benefits for humanity?
Q1. How plausible is human-like artificial intelligence?
A. Toby Walsh, Professor of AI:
It is 100% plausible that we’ll have human-like artificial intelligence.
I say this even though the human brain is the most complex system in the universe that we know of. There’s nothing approaching the complexity of the brain’s billions of neurons and trillions of connections. But there are also no physical laws we know of that would prevent us reproducing or exceeding its capabilities.
A. Kevin Korb, Reader in Computer Science
Popular AI from Issac Asimov to Steven Spielberg is plausible. What the question doesn’t address is: when will it be plausible?
Most AI researchers (including me) see little or no evidence of it coming anytime soon. Progress on the major AI challenges is slow, if real.
What I find less plausible than the AI in fiction is the emotional and moral lives of robots. They seem to be either unrealistically empty, such as the emotion-less Data in Star Trek, or unrealistically human-identical or superior, such as the AI in Spike Jonze’s Her.
All three – emotion, ethics and intelligence – travel together, and are not genuinely possible in some form without the others, but fiction writers tend to treat them as separate. Plato’s Socrates made a similar mistake.
A. Gary Lea, Researcher in Artificial Intelligence Regulation
AI is not impossible, but the real issue is: “how like is like?” The answer probably lies in applied tests: the Turing test was already (arguably) passed in 2014 but there is also the coffee test (can an embodied AI walk into an unfamiliar house and make a cup of coffee?), the college degree test and the job test.
If AI systems could progressively pass all of those tests (plus whatever else the psychologists might think of), then we would be getting very close. Perhaps the ultimate challenge would be whether a suitably embodied AI could live among us as J. Average and go undetected for five years or so before declaring itself.
Q2. Automation is already replacing many jobs. Is it time make laws to protect some of these industries?
A. Jonathan Roberts, Professor of Robotics
Researchers at the University of Oxford published a now well cited paper in 2013 that ranked jobs in order of how feasible it was to computerise or automate them. They found that nearly half of jobs in the USA could be at risk from computerisation within 20 years.
This research was followed in 2014 by the viral video hit, Humans Need Not Apply, which argued that many jobs will be replaced by robots or automated systems and that employment would be a major issue for humans in the future.
Of course, it is difficult to predict what will happen, as the reasons for replacing people with machines are not simply based around available technology. The major factor is actually the business case and the social attitudes and behaviour of people in particular markets.
A. Rob Sparrow, Professor of Philosophy
Advances in computing and robotic technologies are undoubtedly going to lead to the replacement of many jobs currently done by humans. I’m not convinced that we should be making laws to protect particular industries though. Rather, I think we should be doing two things.
First, we should be making sure that people are assured of a good standard of living and an opportunity to pursue meaningful projects even in a world in which many more jobs are being done by machines. After all, the idea that, in the future, machines would work so that human beings didn’t have to toil used to be a common theme in utopian thought.
When we accept that machines putting people out of work is bad, what we are really accepting is the idea that whether ordinary people have an income and access to activities that can give their lives meaning should be up to the wealthy, who may choose to employ them or not. Instead, we should be looking to redistribute the wealth generated by machines in order to reduce the need for people to work without thereby reducing the opportunities available to them to be doing things that they care about and gain value from.
Second, we should be protecting vulnerable people in our society from being treated worse by machines than they would be treated by human beings. With my mother, Linda Sparrow, I have argued that introducing robots into the aged care setting will most likely result in older people receiving a worse standard of treatment than they already do in the aged care sector. Prisoners and children are also groups who are vulnerable to suffering at the hands of robots introduced without their consent.
A. Toby Walsh, Professor of AI:
There are some big changes about to happen. The #1 job in the US today is truck driver. In 30 years time, most trucks will be autonomous.
How we cope with this change is a question not for technologists like myself but for society as a whole. History would suggest that protectionism is unlikely to work. We would, for instance, need every country in the world to sign up.
But there are other ways we can adjust to this brave new world. My vote would be to ensure we have an educated workforce that can adapt to the new jobs that technology create.
We need people to enter the workforce with skills for jobs that will exist in a couple of decades time when the technologies for these jobs have been invented.
We need to ensure that everyone benefits from the rising tide of technology, not just the owners of the robots. Perhaps we can all work less and share the economic benefits of automation? This is likely to require fundamental changes to our taxation and welfare system informed by the ideas of people like the economist Thomas Piketty.
A. Kevin Korb, Reader in Computer Science
Industrial protection and restriction are the wrong way to go. I’d rather we develop our technology so as to help solve some of our very real problems. That’s bound to bring with it economic dislocation, so a caring society will accommodate those who lose out because of it.
But there’s no reason we can’t address that with improving technology as long as we keep the oligarchs under control. And if we educate people for flexibility rather than to fit into a particular job, intelligent people will be able to cope with the dislocation.
A. Jai Galliot, Defence Analyst
The standard argument is that workers displaced by automation go on to find more meaningful work. However, this does not hold in all cases.
Think about someone who signed up with the Air Force to fly jets. These pilots ma y have spent their whole social, physical and psychological lives preparing or maintaining readiness to defend their nation and its people.
For service personnel, there are few higher-value jobs than serving one’s nation through rendering active military service on the battlefield, so this assurance of finding alternative and meaningful work in a more passive role is likely to be of little consolation to a displaced soldier.
Thinking beyond the military, we need to be concerned that the Foundation for Young Australians indicates that as many as 60% of today’s young people are being trained for jobs that will soon be transformed due to automation.
The sad fact of the matter is that one robot can replace many workers. The future of developed economies therefore depends on youth adapting to globalised and/or shared jobs that are increasingly complemented by automation within what will inevitably be an innovation and knowledge economy.
Q3. Where will artificial intelligence be in five-to-ten years?
A. Toby Walsh, Professor of AI:
AI will become the operating system of all our connected devices. Apps like Siri and Cortana will morph into the way we interact with the connected world.
AI will be the way we interact with our smarthphones, cars, fridges, central heating system and front door. We will be living in an always-on world.
A. Jonathan Roberts, Professor of Robotics
It is likely that in the next five to ten years we will see machine learning systems interact with us in the form of robots. The next large technology hurdle that must be overcome in robotics is to give them the power of sight.
This is a grand challenge and one that has filled the research careers of many thousands of robotics researchers over the past four or five decades. There is a growing feeling in the robotics community that machine learning using large datasets will finally crack some of the problems in enabling a robot to actually see.
Four universities have recently teamed up in Australia in an ARC funded Centre of Excellence in Robotic Vision. Their mission is to solve many of the problems that prevent robots seeing.
Q4. Should we be concerned about military and other armed robots?
A. Rob Sparrow, Professor of Philosophy
The last thing humanity needs now is for many of its most talented engineers and roboticists to be working on machines for killing people.
Robotic weapons will greatly lower the threshold of conflict. They will make it easier for governments to start wars because they will hold out the illusion of being able to fight without taking any casualties. They will increase the risk of accidental war because militaries will deploy unmanned systems in high threat environments, where it would be too risky to place a human being, such as just outside a potential enemy’s airspace or deep sea ports.
In these circumstances, robots may even start wars without any human being having the chance to veto the decision. The use of autonomous robots to kill people threatens to further erode respect for human life.
It was for these reasons that, with several colleagues overseas, I co-founded the International Committee for Robot Arms Control, which has in turn supported the Campaign to Stop Killer Robots.
A. Toby Walsh, Professor of AI:
“Killer robots” are the next revolution in warfare, after gunpowder and nuclear bombs. If we act now, we can perhaps get a ban in place and prevent an arms race to develop better and better killer robots.
A ban won’t uninvent the technology. It’s much the same technology that will go, for instance, into our autonomous cars. And autonomous cars will prevent the 1,000 or so deaths on the roads of Australia each year.
But a ban will associate enough stigma with the technology that arms companies won’t sell them, that arms companies won’t develop them to be better and better at killing humans. This has worked with a number of other weapon types in the past like blinding lasers. If we don’t put a ban in place, you can be sure that terrorists and rogue nations will use killer robots against us.
For those who argue that killer robots are already covered by existing humanitarian law, I profoundly disagree. We cannot correctly engineer them today not to cause excessive collateral damage. And in the future, when we can, there is little stopping them being hacked and made to behave unethically. Even used lawfully, they will be weapons of terror.
You can learn more about these issues by watching my TEDx talk on this topic.
A. Sean Welsh, Researcher in Robot Ethics
We should be concerned about military robots. However, we should not be under the illusion that there is no existing legislation that regulates weaponised robots.
There is no specific law that bans murdering with piano wire. There is simply a general law against murder. We do not need to ban piano wire to stop murders. Similarly, existing laws already forbid the use of any weapons to commit murder in peacetime and to cause unlawful deaths in wartime.
There is no need to ban autonomous weapons as a result of fears that they may be used unlawfully any more than there is a need to ban autonomous cars for fear they might be used illegally (as car bombs). The use of any weapon that is indiscriminate, disproportionate and causes unnecessary suffering is already unlawful under international humanitarian law.
Some advocate that autonomous weapons should be put in the same category as biological and chemical weapons. However, the main reason for bans on chemical and biological weapons is that they are inherently indiscriminate (cannot tell friend from foe from civilian) and cause unnecessary suffering (slow painful deaths). They have no humanitarian positives.
By contrast, there is no suggestion that “killer robots” (even in the examples given by opponents) will necessarily be indiscriminate or cause painful deaths. The increased precision and accuracy of robotic weapons systems compared to human operated ones is a key point in their favour.
If correctly engineered, they would be less likely to cause collateral damage to innocents than human operated weapons. Indeed robot weapons might be engineered so as to be more likely to capture rather than kill. Autonomous weapons do have potential humanitarian positives.
Q5. How plausible is super-intelligent AI?
A. David Dowe, Associate Professor in Machine Learning and Artificial Intelligence
We can look at the progress made at various tasks once said to be impossible for machines to do, and see them one by one gradually being achieved. For example: beating the human world chess champion (1997); winning at Jeopardy! (2011); driverless vehicles, which are now somewhat standard on mining sites; automated translation, etc.
And, insofar as intelligence test problems are a measure of intelligence, I’ve recently looked at how computers are performing on these tests.
A. Rob Sparrow, Professor of Philosophy
If there can be artificial intelligence then there can be super-intelligent artificial intelligences. There doesn’t seem to be any reason why entities other than human beings could not be intelligent. Nor does there seem to be any reason to think that highest human IQ represents the upper limit on intelligence.
If there is any danger of human beings creating such machines in the near future, we should be very scared. Think about how human beings treat rats. Why should machines that were as many times more intelligent than us, as we are more intelligent than rats, treat us any better?
Q6. Given what little we know about our own minds, can we expect to intentionally create artificial consciousness?
A. Kevin Korb, Reader in Computer Science
As a believer in functionalism, I believe it is possible to create artificial consciousness. It doesn’t follow that we can “expect” to do it, but only that we might.
John Searle’s arguments against the possibility of artificial consciousness seem to confuse functional realisability with computational realisability. That is, it may well be (logically) impossible to “compute” consciousness, but that doesn’t mean that an embedded, functional computer cannot be conscious.
A. Rob Sparrow, Professor of Philosophy
A number of engineers, computer scientists, and science fiction authors argue that we are on the verge of creating artificial consciousness. They usually proceed by estimating the number of neurons in the human brain and pointing out that we will soon be able to build computers with a similar number of logic gates.
If you ask a psychologist or a psychiatrist, whose job it is to actually “fix” minds, I think you will likely get a very different answer. After all, the state-of-the-art treatment for severe depression still consists in shocking the brain with electricity, which looks remarkably like trying to fix a stalled car by pouring petrol over the top of the engine. So I’m sceptical that we understand enough about the mind to design one.
Q7. How do cyborgs differ (technically or conceptually) from A.I.?
A. Katina Michael, Associate Professor in Information Systems
A cyborg is a human-machine combination. By definition, a cyborg is any human who adds parts, or enhances his or her abilities by using technology. As we have advanced our technological capabilities, we have discovered that we can merge technology onto and into the human body for prosthesis and/or amplification. Thus, technology is no longer an extension of us, but “becomes” a part of us if we opt into that design.
In contrast, artificial intelligence is the capability of a computer system to learn from its experiences and simulate human intelligence in decision-making. A cyborg usually begins as a human and may undergo a transformational process, whereas artificial intelligence is imbued into a computer system itself predominantly in the form of software.
Some researchers have claimed that a cyborg can also begin in a humanoid robot and incorporate the living tissue of a human or other organism. Regardless, whether it is a human-to-machine or machine-to-organism coalescence, when AI is applied via silicon microchips or nanotechnology embedded into prosthetic forms like a dependent limb, a vital organ, or a replacement/additional sensory input, a human or piece of machinery is said to be a cyborg.
There are already early experiments with such cybernetics. In 1998 Professor Kevin Warwick named his first experiment Cyborg 1.0, surgically implanting a silicon chip transponder into his forearm. In 2002 in project Cyborg 2.0, Warwick had a one hundred electrode array surgically implanted into the median nerve fibres of his left arm.
Ultimately we need to be extremely careful that any artificial intelligence we invite into our bodies does not submerge the human consciousness and, in doing so, rule over it.
Q8. Are you generally optimistic or pessimistic about future of artificial intelligence and its benefits for humanity?
A. Toby Walsh, Professor of AI:
I am both optimistic and pessimistic. AI is one of humankind’s truly revolutionary endeavours. It will transform our economies, our society and our position in the centre of this world. If we get this right, the world will be a much better place. We’ll all be healthier, wealthier and happier.
Of course, as with any technology, there are also bad paths we might end up following instead of the good ones. And unfortunately, humankind has a track record of late of following the bad paths.
We know global warming is coming but we seem unable not to follow this path. We know that terrorism is fracturing the world but we seem unable to prevent this. AI will also challenge our society in deep and fundamental ways. It will, for instance, completely change the nature of work. Science fiction will soon be science fact.
A. Rob Sparrow, Professor of Philosophy
I am generally pessimistic about the long term impact of artificial intelligence research on humanity.
I don’t want to deny that artificial intelligence has many benefits to offer, especially in supporting human beings to make better decisions and to pursue scientific goals that are currently beyond our reach. Investigating how brains work by trying to build machines that can do what they do is an interesting and worthwhile project in its own right.
However, there is a real danger that the systems that AI researchers come up with will mainly be used to further enrich the wealthy and to entrench the power of the powerful.
I also think there is a risk that the prospect of AI will allow people to delude themselves that we don’t need to do something about climate change now. It may also distract them from the fact that we already know what to do, but we lack the political will to do it.
Finally, even though I don’t think we’ve currently got much of a clue of how this might happen, if engineers do eventually succeed in creating genuine AIs that are smarter than we are, this might well be a species-level extinction threat.
A. Jonathan Roberts, Professor in Robotics
I am generally optimistic about the long-term future of AI to humanity. I think that AI has the potential to radically change humanity and hence, if you don’t like change, you are not going to like the future.
I think that AI will revolutionise health care, especially diagnosis, and will enable the customisation of medicine to the individual. It is very possible that AI GPs and robot doctors will share their knowledge as they acquire it, creating a super doctor that will have access to all the medical data of the world.
I am also optimistic because humans tend to recognise when technology is having major negative consequences, and we eventually deal with it. Humans are in control and will naturally try and use technology to make a better world.
A. Kevin Korb, Reader in Computer Science
I’m pessimistic about the medium-term future of humanity. I think climate change and attendant dislocations, wars etc. may well massively disrupt science and technology. In that case progress on AI may stop.
If that doesn’t happen, then I think progress will continue and we’ll achieve AI in the long-term. Along the way, AI research will produce spin-offs that help economy and society, so I think as long as it exists AI tech will be important.
A. Gary Lea, Researcher in Artificial Intelligence Regulation
I suspect the long-term future for AI will turn out to be the usual mixed bag: some good, some bad. If scientists and engineers think sensibly about safety and public welfare when making their research, design and build choices (and provided there are suitable regulatory frameworks in place as a backstop), I think we should be okay.
So, on balance, I am cautiously optimistic on this front – but there are many other long-term existential risks for humanity.
Toby Walsh, Professor of AI, Research Group Leader, Optimisation Research Group , Data61; David Dowe, Associate Professor, Clayton School of Information Technology, Monash University; Gary Lea, Visiting Researcher in Artificial Intelligence Regulation, Australian National University; Jai Galliott, Research Fellow in Indo-Pacific Defence, UNSW Australia; Jonathan Roberts, Professor in Robotics, Queensland University of Technology; Katina Michael, Associate Professor, School of Information Systems and Technology, University of Wollongong; Kevin Korb, Reader in Computer Science, Monash University; Robert Sparrow, Professor, Department of Philosophy; Adjunct Professor, Centre for Human Bioethics, Monash University, and Sean Welsh, Doctoral Candidate in Robot Ethics, University of Canterbury
This article was originally published on The Conversation. Read the original article.
Related reading: 2017 in artificial intelligence
