MRSM Conference: Man vs Machine
Is the struggle towards the future a case of Man versus Machine? Or will humans and machine one day be the same? Adapted from a presentation given to the Marketing Research Society of Malaysia.
· Good morning. My name is Matt Armitage.
· A lot of you may not know who I am or why I’m speaking to you this morning.
· I am a futurist and a communications strategist. I have a company called Kulturpop.
· Some of you may have heard me on BFM
· I have a weekly show where I talk about the implications that technologies have for the way we live our lives.
· Of course, the race towards an AI powered future is an increasingly important strand of that.
· One of the reasons I was so keen to speak to you all today is because we’re doing some of the same things.
· You are all from an industry that is invested in finding out what people thing and what they believe and asking that most difficult of questions: why?
· And it’s a task you undertake, even when those answers are difficult or uncomfortable for your clients.
· That said, I have been reminded that it’s first thing in the morning, you have a lot of speakers to hear today, so maybe I shouldn’t throw you straight into the fires of the apocalypse, which is where my radio show often ends up.
Visiting the WEF.
· So I thought I’d start with something upbeat and light-hearted.
· Like the World Economic Forum
· In September, the WEF published a report on the global impact of technologies like AI and automation.
· Unsurprisingly, at the very top of the list of professions that would not be affected by AI were CEOs and Managing Directors.
· The theory seems to be that the rest of us is fair game.
· I don’t actually agree with that, I think that MDs and CEOs are at just as much, and possibly more risk than the rest of us, but that’s probably a conversation for another place and time.
· If we go back to the report
o It estimated that automation would lead to the creation of roughly the same number of jobs as it sheds in the next ten to 15 years.
o Most of the jobs it creates will be in data linked professions. No surprises there.
o And for those of you in accounting, legal and client servicing, who don’t have the security of being the Chief Executive, the future is looking less rosy.
· Of course, the WEF’s report is largely a short term one.
· A lot of it focuses on what will happen between now and 2022.
· That’s an important benchmark for companies…
· But for society, when we talk about AI, it’s really 2032 that’s important. And 2042 and on and on.
· And the next 10 or 15 years are going to see incredible social changes, many of which will be powered by AI.
· These changes will transform societies and may even see us alter the definition of what it is to be a human being.
· Which, if you’re pursuing a career in a research industry that extracts information from human beings…
· You’re in for an interesting ride.
· In the spirit of that interesting ride, today I want to take you on a quick trip around the outer reaches of technologies like AI.
· This isn’t a prediction of the future as much as a projection of where the technology may take us.
· And maybe you’ll sit there thinking that some of the things I’m talking about don’t apply to you or your industry.
· But making use of AI is also about understanding how this technology will change the fabric of society and alter and influence people’s daily lives.
· Looking at that picture is not always easy.
· Especially when it comes to really complex technology.
· It has a way of getting us stuck.
· It’s understandable, the task of figuring out how technology applies to our own industry can be so exhausting and consuming that we don’t have the time or the mental space to examine any implications it might have has beyond that.
· So where do we start?
· Maybe by looking at how dumb AI is rather than how smart.
· AI and machine intelligence is so different from our own that’s it’s pretty much meaningless to throw out analogies like:
· As smart as Stephen Hawking? As smart as a ten year old kid? As smart as a beagle.
· And that makes it really hard to create a mental picture of what we mean by inteligent in this context.
· My favourite analogy is the 5 year old boy and dinosaurs.
· That kid knows everything about dinosaurs.
· Their size, their weight. Their Latin names, species and genealogy.
· They know their predatory characteristics and feeding habits.
· That kid is a dinosaur genius.
· Ask him to make you breakfast and you’re likely to get a buffet of gummi bears.
· For the foreseeable future, by which I mean probably 3-5 years, AI is going to be extremely stupid.
· And Stupid machines make stupid decisions.
· What do I mean when I say AI is stupid?
· Deep Mind’s Alpha Go is the world’s greatest go player.
· Its younger sibling, Alpha Go Zero, is even smarter.
· It’s entirely self-taught
· Starting with random moves, it played itself over again, Learning from every game and permutation.
· And unlike a human, it never forgets anything that it learns.
· After only three days playing, the Zero was able to defeat Alpha Go decisively.
Breaking the Board.
· Those machine are smart. Smarter than any human being at the single task of playing Go.
· That intelligence comes with a huge caveat.
· Change the dimensions of that board by even one square and not only will those machines not be able to win against any human player
· They won’t even be able to play.
· So when we think of AI as a cure-all for business, It’s worth remembering that even the least capable employee in your team or your company is far more adaptable than the most expensive AI you could build or lease.
· Of course, in the consumer space, AI has become almost synonymous with autonomous cars.
· This is the pie that everyone from Volvo to Google to Uber wants a piece of.
· But self-driving systems have already led to a number of human fatalities.
· Misuse of Tesla’s self-driving mode has resulted in a couple of fatal collisions in recent years.
· As many of you will know, earlier this year one of Uber’s self-driving test cars killed a pedestrian.
· So, while we’re being told that self-driving cars will be on our roads within 2 to four years.
· Some machine intelligence experts disagree.
· They say the systems are not smart enough.
· Because the systems we have now are great at parsing data.
· They can analyse traffic flows and deduce that as all the cars stop at that intersection, therefore there must be a stop sign there.
· But knowing there is a stop sign there at an intellectual level is very different from being able to spot a stop sign and react to it.
· And then there’s us.
· Human being are unpredictable.
· We do things on the spur of the moment, and for no logical reason.
· Like breakdancing in the road.
· So we have to think very carefully about how we share our world and our public spaces with autonomous machines.
· In any confrontation, several tonnes of metal usually has the advantage over the 70kg of mostly organic human.
Smarter & Better AI
· What’s the answer? Well, we need AI that is much smarter than we have now.
· We need machines that are intelligent enough to make genuine choices. To adapt. To be able to think beyond the confines of their programming.
· Which, as I think everyone in the room knows, we’re well on the way to achieving, at least in first generation form.
· We don’t need – and we probably don’t want – machines that think like us.
· That contradiction has led some commentators think that AI poses an existential threat to humanity.
· Guys like Elon Musk fear that AI will go rogue and enslave or eradicate us.
· But I think that tells you more about Elon Musk than it does about the intentions of clever machines.
· I don’t think that sentient AI poses an existential threat to humanity.
· I think it’s the dumb AIs we have now that are a greater threat.
· Look at the role that HFT algorithms in the financial markets may have played in creating our last global recession.
· Algorithms created by trading houses to automate purchases and sales of stocks transactions that would deliberately send out contradictory requests to try and confuse their competitors’ algorithms and disguise their own positions, leaving a trail of destabilising price spikes in their wake.
Why Am I Telling You This?
· You may be thinking, well, what does this have to do with me?
· A lot of the people in this room will be thinking about adopting AI now or in the very near future.
· A lot of very smart people will be telling you that their machine systems will do incredible things for your business.
· What they probably won’t be doing is telling you what happens when their AI rubs up against someone else’s AI.
· And there’s a very simple reason for that: they don’t know.
· I’ve got some more examples for you of what dumb AI can do.
Dumb AI: Bofors
· Back in 2012 a Bofors autonomous artillery system killed its own crew.
· A mistake was made with the gun’s targeting area and as a result it identified the crew as valid targets.
· The gun didn’t malfunction, in fact It did exactly what it was programmed to do.
· What it wasn’t programmed to do, and what it wasn’t able to do, was to realise it had made a mistake and not fire.
Dumb AI: Cambridge Analytica
· Look at the damage that Ai assisted political targeting has done on platforms like Facebook and in terms of fuelling the fake news epidemic.
· Again, you can’t blame the AI: it did nothing wrong.
· But would it have been any different if that AI had even a semblance of sentience, the ability to make a decision whether what it was doing was right or wrong?
· That maybe it shouldn’t target people with fake news?
Dumb AI: Amazon AI
· Just last week, Amazon had to step back from an AI it was hoping would revolutionise human resources.
· It could analyse thousands of resumes and accurately and without bias select the top dozen candidates for interview.
· It wasn’t unbiased. Because Amazon had previously been a male skewed organisation, the historical data caused the machine to perceive women to be inferior and favour men.
· So, our AI is dumb.
The World of Sentience
· Unfortunately, there isn’t time today to go through all the steps to actually getting that smart AI.
· So let’s jump straight to a world where we have AI that can think independently. AI that can reason. AI that is sentient.
· At this point the discussion about what that world looks like and how people and machines integrate and behave becomes very relevant to the kind of business that you all do.
Can an AI be Human?
· First of all, how should society treat sentient machines?
· Should an AI have rights? Should it have the same rights that we have?
· I know those might sound like ridiculous statements but we should be trying to answer these questions now.
· When you look at the legislation gap, we’re already behind the curve.
· For example, Europe introduced the GDPR data regulations this year.
· And those regulations would be fantastic if this was still 2014.
· Right now we’re stuck at ground zero, talking about who is legally responsible if an AI powered car has an accident.
· I’m not saying that those aren’t important considerations: but what we really should be deciding is how sentient we want these machines to be.
· This isn’t an esoteric discussion. In the US company personhood is legally enshrined.
· If we want to have machine intelligence that is self-aware and self-determining, we have to decide what freedoms those machines enjoy.
· Can it own property?
· Is it responsible for criminal acts? If so, do you turn it off or do put it in a firewalled prison?
· Because let’s be honest about this: AI may be smart but it won’t always be right.
· We have laws and codes of behaviour that help to guide us when people make mistakes.
· We are going to need those same systems to work for machines as well.
It’s just a machine.
· A lot of people ask me: Why does any of this matter, these are machines. They don’t need rights.
· If society decides that: fine.
· But if machines don’t have rights, they shouldn’t be self-aware.
· I often get asked about sex robots and AI and my answer is simple: if the machine is self-aware and it isn’t free to give informed consent and make its own decisions about those human partners, then that machine is a sex slave.
· And that goes for the AI powering your car, your house or your company.
· If it’s sentient and it has no rights, it’s a slave.
· Because these aren’t just machines. Soon they will be us and we will be them.
· We are already starting to blur the lines between human and machine.
· People have robotic limbs and organs.
· We can implant computers in nerve endings to switch off pain signals and short circuit conditions like arthritis.
· Brain implants are already in the works for people with dementia, enabling them to store and access the memories that their organic brains can no longer write or retrieve.
Dual Use Technology
· Looking at where we are now, I know some of this sounds implausible.
· Part of the reason it sounds implausible is that we have a tendency to compartmentalise things that sometimes stops us from seeing all the applications that technologies have.
· If we build implants that can replace memory for people with medical conditions, then someone is going to take those devices and use them to enhance healthy brains.
· Someone will also connect those implants to the cloud.
· If you connect that implant to the cloud, why not connect it to Siri or Alexa?
· Suddenly you don’t just have a chip in your brain, you have an AI
· Advance that thought by a few tech-generations and that chip is no longer Siri or Alexa, it’s a machine with its own thoughts and its own identity.
· A second personality sitting in your brain.
· This is what some commentators are defining as the age of the posthuman.
· We could start to see a divergence between enhanced and unaugmented humans.
· A divergence that would be socio-economic to begin with but could become tribal and existential.
· Because we have to accept that those enhancements may be physical as well as mental.
· Because, why wouldn’t you use gene modification or biotech alongside AI implants?
· This is where that legal framework surrounding AI becomes critical.
· If you implant a sentient AI chip in a person’s head is that person now one or two individuals?
· Will we define humans according to the proportion of homo sapiensto machine in their body?
Who owns my brain?
· Then there is the question of how we access these AI services?
· Will we own our own minds, or will it be like a smartphone, where you own the phone and a company like Apple owns the operating system?
· Of course, all of this has interesting implications for the research industry.
· If an AI is in someone’s head, does that mean that the companies that own or lease that data will have real time access to our thoughts?
· Just imagine what that means for marketers.
· No second-guessing. No focus groups or online surveys,
· The task then becomes how to actually sort and categorise and prioritise that data in ways that are meaningful to your clients.
· So if you’re sitting here thinking about how you should roll out AI for your business.
· Maybe you should also spend a little time thinking about how to access and analyse the data from AI sources embedded within consumers.
Getting ahead of ourselves.
· Of course, I’m getting ahead of myself.
· Right now, AI is that toddler with the dinosaur obsession and the gummi bear buffet.
· All of this is really food for thought.
· Much of what I’ve talked about today may never happen
· But it isn’t science fiction.
· Only last week it was reported that a team of scientists at the University of Washington linked three people up to a machine via EEG caps where they were able to collaborate and play a game of Tetris using only their thoughts.
· It’s already being touted as Hivemind technology, and of course, the next step to increase linear processing of tasks would be to add an AI to that network.
· So who knows what fantastic places we’ll find ourselves in in 10 years time.
· Thank you for listening.
· If you’d like to hear more, please listen to my show on BFM at 10 AM every Friday morning.
· If you can’t listen live because, hey, you’ve got a job, you can download the podcasts from EFM, Spotify, iTunes, Google play and most of the other places that podcasts are available.
· My email and website address are also there, I’ll put up a transcript of this presentation should you want to go and read over any of it again.
· I guess the final thing is If any of you have any questions, I would love to try and answer them.