Why has it been so difficult to build the kind of humanoid robots that have been made popular in science fiction films and books?
It's funny, because robotics is one of those fields that has been around for a long time—in a sense it kind of mirrors the development of artificial intelligence. AI came to the fore in the 1960s when we saw HAL in the movie 2001 and all of these sci-fi robots, and at the time there was a lot of optimism. Computers were getting faster every year and people were thinking we'd have machine translation nailed in five years and common sense reasoning in 10 years. Then they kind of just hit this brick wall; they fundamentally misunderstood how robust and intricate human-level intelligence really is. Human intelligence, like a lot of biology, is a highly parallel process. To use a paradigm for how intelligent decision making works, a human is fundamentally different from the way we actually engineer AI. We're actually closer to being able to mimic what nature has done [mechanically], because I think we fundamentally understand it at a much deeper level than we do intelligence. We understand the way most of our body works. Probably the most mysterious object in the known universe is the human brain.
So the difficulty is more a software (or artificial intelligence) problem than a mechanical one?
Artificially intelligent systems like computers work very well because their environments can be formalized with symbolic representations of what they're dealing with. But the real world [where robots must function] is inherently resistant to high levels of symbolic representation.
In order for a robot to function in the real world, its intelligence has to be more like that of the human brain than a computer?
Generally speaking, the key is sensory awareness. Humans have kind of evolved to fit into their environment [by filtering out information they don't need]. If you actually look at the amount of data coming in through all your senses, there's something like 100 million bits of information coming in every second through your visual system and another 10 million bits coming through your auditory system and another one million bits coming through your tactile system. We're basically at any given time absorbing hundreds of millions of bits of data per second through our senses. We can manage this, because our conscious stream is only aware of a very tiny fraction of that sensory input, maybe a few hundred bits per second. Most of our intelligence is really a filtering process. Which of those bits are most relevant at any instant? Our sensory awareness is really much higher than we perceive.
Another common theme, particularly in fiction like The Terminator, is the idea that robots can build subsequent generations of improved robots. How realistic is this?
This gets into a lot of issues of nesting. Can something recreate itself without having an overview of the process? When an organism is developing as an embryo, there's no master planner that assigns cells to be a bone or an eye. The overall design of that organism is a [by-product] of genetics and embryology. Each cell is making its own independent decisions all along the way. And due to the nature of that system, they happen to differentiate correctly, move to the right place and develop correctly. That's the approach that a self-replicating robot would have to take, a bottom-up, distributed collective intelligence approach. That's also the holy grail of nanotechnology—creating self-assembling devices. It is theoretically possible to build a self-replicator, but a lot of that will have a lot to do with context. A nanosize machine self replicating in your blood stream is going to be an extremely different thing than an interstellar space probe traveling out amongst the stars trying to replicate into more space probes.
Subscription Center
See what we're tweeting about
7 Comments
Add CommentInteresting interview - and good on Mr. Wright for emphasizing that consciousness is mostly about knowing what to ignore.
Reply | Report Abuse | Link to thisIf Will Wright has had so much success in this domain, perhaps it is precisely because he is self-taught, learning from the complexities of the real world, rather than having been completely pre-programmed by robotics experts?
Reply | Report Abuse | Link to thisI think it is worth asking what Will Wright's real agenda is here. Although his work to advance gaming has been well documented, many of us have good reason to suspect he is not in favor of further advancements in robot technology. Wright, with the help of the Minnesota League of Women Voters, has been doing everything in his power to slow critical development in human modeling in robotics. Why, Mr. Wright? Why?
Reply | Report Abuse | Link to thisKudos to Will Wright for an insightful and innovative perspective on the nature of robots and AI, distilled from his own first-hand experiences. Hes right on that we wont truly understand the complexity of human life until we actually learn to mimic those processes mechanically. The ability to reverse engineer the human brain really will be the primary determinant for advancements in robotics and especially, humanoid robotics, in the next several decadesas it leads to the development of sophisticated machines with the ability to learn and acquire new information independently. Thoughts? "Synthetic Neural Circuits and the Humanoid Brain" at http://www.zygbotics.com/2009/03/15/biomimetic-neural-circuits-and-the-future-of-robotics/
Reply | Report Abuse | Link to thisInterestingly enough, I am preparing concrete irrational documents for posterity along with a "Nouveau Mathematics" which is more relevant to the era of A.I. and the Android as it has its sources in the origins of the species... Symbolically an old man can never understand this but a child could easily absorb it and become a truer genius than any PHD. If I wanted to build the perfect Android I would have to do it in this way, leave the mecha blueprints to those who can read mecha and take on the impossible aspects of designing the internal weave structure of the Android Brane, my goal as a one of History's Painters is to provide the bloody irrational organic mimesis that will one day devour all A.I...
Reply | Report Abuse | Link to thisConciousness is much more than filtering. This just happens to be an area that is being very focused on in robotics: what to pay attention to/spend more cpu cycles on. But conciousness is a complex thought and feedback process. The filtering seems to occur beforehand.
Reply | Report Abuse | Link to thisHe also briefly discusses whether a system can build a better system than itself. It seems unlikely since to understand "oneself" you'd have to have a meta-system for analysis/comprehension, which is not possible. Therefore, a truly "humanoid" robot would be beyond our understanding as to how it functions. Much like ourselves ;)
will make sims 3 for a console you will make more money and i bet a lot of people will buy it you rock will!!!!
Reply | Report Abuse | Link to this