with No Comments

Post No.: 0056robots


Furrywisepuppy says:


As of posting, we can build artificial intelligences that can beat the best of the known natural intelligences fairly and consistently at intellectual games like Go, chess or Jeopardy! yet we cannot currently get anywhere close to building artificial machines that can beat the best of the known natural machines at playing football, gymnastics or many other sports or activities that involve complex physical movements such as dancing or ballet, for instance. They cannot even beat merely average human performers as long as the humans are truly giving it a proper attempt and not just letting the robots win – so maybe football players don’t have quite as small brains as many people think?!


For me, this is partly a philosophical issue because it questions how much we think we know about what makes someone or something ‘clever’, and why mental pursuits are typically considered more admirable, highbrow and sophisticated than physical pursuits? Although not always the case, we generally need to understand something first before we can replicate or build it ourselves – and we seem to better understand how to build robots that are good at complex ‘cerebral’ tasks than complex physical tasks at the moment, suggesting that our intuitions have underestimated what it takes to be good at performing complex or delicate physical activities.


It’s arguably not to do with the bodily mechanics or hardware per se – we can design and build machines that mimic human-like joints and movement quite well, and we have available a variety of electronic sensors and actuators. But programming a ‘central nervous system’ and ‘peripheral nervous system’ that can make this body balance and move as fluidly as humans and other natural animals can is currently proving to be extremely tricky. (Maybe miniaturisation is not quite as good as what nature can do yet though e.g. skin-like touch and temperature sensor sensitivity and density.)


‘Moravec’s paradox’ states that, contrary to traditional assumptions and what it’s like for humans – high-level reasoning (e.g. playing a game of checkers) requires relatively little computation, yet low-level sensorimotor tasks (e.g. perceiving and manipulating a range of objects with fluid dexterity) require relatively vast computational resources. Robots currently struggle to do what 1-year-old toddlers can already physically do pretty well.


But machine designs and artificial intelligences with their algorithms that adapt/learn can and are evolving (in the sense that ideas and designs are memes, and memes evolve) at a much faster rate than natural life can and has been, and impressive (or scary!) progress is constantly being made by some specialist robotic firms – but there’s still currently a very long way to go before they can beat natural animals at such fluidity, speed and range of movements, and can beat humans at such sports (without cheating that is e.g. goalkeeper robots that are exactly the same size as goals and midfield lines that are as tall as stadiums, as wide as pitches and are full of ‘legs’ constantly kicking forwards(!)) I do believe that robots will one day in the future be able to beat the best human football players at football in a fair game (with the need to determine what makes a football-playing robot design ‘fair’ first), but that day isn’t here yet even though they can beat the best humans at many directly intellectual games today.


There is a lot of computing brainpower involved in physical movement and coordination. Some experts also claim that cognition is somewhat embodied i.e. it’s not just about the ‘central nervous system’ but also the design of the body overall – after all, the brains and bodies of natural organisms co-evolved together rather than separately. A lot of evolution has gone into getting organisms to sense and move in their environment, and this shouldn’t be surprising in the long arms races between predators and preys.


Physical coordination is much poorer when the brain is under the influence of drugs, when tired or mentally distracted. This correlates with the finding in neuroscience that physical exercise for humans is also good exercise for cognitive improvement as well as of course for physical improvement (hence one should include regular physical exercise in one’s life, on top of challenging cognitive pursuits, if one wants one’s brain to be as fit as it can possibly be – which means that we can only speculate what the late and great Stephen Hawking could’ve cognitively achieved if he could’ve been more physically active!) Whether it’s down to improved blood flow to the brain, being a direct cognitive test in itself and/or physical coordination uses far more computing power than one may think, or whatever other reason for it – exercise is good for the brain and is a good test for the brain too – in particular exercises that are intense enough and challenge one’s coordination, such as dancing. And it’s another reason why the elderly should keep regularly physically active for as long as possible too. Woof.


So maybe robots and AIs still currently have a long way to go before they impress ‘dumb’ sportspeople! Of course, if you want to be good at a particular task then regularly practice that specific task, so e.g. footballers who don’t play chess won’t be good at chess just for playing football, and vice-versa for people who just play chess and want to improve their football skills (read Post No.: 0031). But it would be erroneous to say that athletes have ‘no brains’. There are simply different types of intelligence. But then you’ll have to ask if athletic ability is good or enough for being a well-rounded person/worker/partner? Although this can also be asked about being good at chess or whatever else, or if you’ll even need a well-rounded person at all if a person just needs to do one job and nothing else, for which robots (at least currently) are very good at doing just one specialised job or role extremely well (including precisely repeated manual physical tasks, as well as ‘intellectual’ mental tasks such as predictions), thus are displacing many humans in the workplace. (Multifunctional, generalised intelligence AIs that are like humans are currently a long way off though – and progress on this front has been slow so far.)


In the main, you may only need to be good at one thing at work, but if that describes your job well and your job isn’t that mentally or physically complex then your job is probably most at risk of being taken over by artificial machines. It could even be the case that robots will be the masters of humans in the workplaces of the future because robots and AIs are relatively good at decision-making tasks whilst they’re still relatively poor or slow at fiddly manual tasks such as picking delicate berries from bushes without bruising them!


Furrywisepuppy promotes a life that is full of both mental and physical stimulation, mental and physical challenges, as well as knowledge diversity. Where possible, everyone should participate in both regular mental exercise as well as regular physical exercise – your brain will appreciate both kinds of exercises in the long run. So even if you’re bookish, please give your body some love and care too.




Comment on this post by replying to this tweet:


Share this post