So I’ve recently realized that I have a growing love for BEAM robotics and a great sadness that technology has marched on without it. I was just wondering about the following:
Did BEAM robotics fall out of fashion because (for many) it’s much easier to program a microcontroller/computer than to build and debug analog circuits? I feel the same, but with neuromorphic chips on the rise, BEAM robotics and neuromorphic chips are still things I’d like to experiment with in the future.
What if we could “enhance” BEAMbots by combining them with some digital electronics and maybe even neuromorphic chips/PCBs as well? Sure, they wouldn’t be true BEAMbots anymore, but I skimmed a webpage where someone actually augmented a BEAM robot with a microcontroller. It was programmed with adaptable simulated circuitry. I might post the aforementioned webpage to the links section if I come across it again.
I personally think that integrating the BEAM methodology with digital microcontrollers/computers/circuits might bring about some more interest in BEAM-like robots again, just in a different form suitable for today’s hobbyists and engineers.
You are correct that it is easy to program a microcontroller these days. Understanding each component in a mostly electronic project takes time and a level of patience which most people don’t have these days. Projects and aspirations have gone beyond simple / dedicated tasks and as you eluded to, BEAM robots are becoming the vynil records of the modern age. Neuromorphic chips are still “chips” in that you can add them to a complex PCB. How / if / when neuromorphic chips are integrated will depend on how much the end user needs to understand about their inner workings. BEAM allowed people to gain soldering experience with a variety of components and create something which could actually perform a basic action, but commercial BEAM robots were / are not inexpensive either.
That makes so much more sense! I had thought that BEAM robots had untapped potential… but if they went out of style, then perhaps they just didn’t have enough potential. Thank you, cBenson!
Strictly my opinion, but I see BEAM robots as very limited in function & expandability. Their behaviors are severely limited by the hard wiring of the components used. Whereas even a basic processor has memory and can do computations to compute distance and many different branch routines based on a given situation. You can also completely redesign the personality of your robot without ever lifting a soldering iron.
I love BEAM, such elegance with every part having a function. It couldn’t be more different than using a Raspberry Pi for an object avoidance robot yet its functionality is the same. But, we live in a digital age and we take simulations for granted. We have to, a digital based system will always be a simulation, an approximation of the real world.
It’s for example for a reason that analogue synthesizers and record players still are manufactured and sold, often for large amounts of money.
Guys, I’ve just had a Guys, I’ve just had a mostly-off-topic but interesting revelation. I don’t know what you’ll think of it, though I’d like to know.
Pascal Kaufmann has recently declared that artificial intelligence is not really true intelligence, with a major reason being that AI is just human intelligence coded into a program. But, in my opinion, I half-heartedly disagree. If the code is an extension of human intelligence, isn’t that technically exactly what we’ve been aiming for for decades? Could that really not be a form of intelligence?
I say the same of analog circuitry, not to tear it down, but rather to uphold it as well. If circuits are based on the human brain, then that’s exactly what those circuits are also- extensions by nature, and therefore intelligent.
Do we really need all of these neural networks and other fancy algorithms to extend our human intelligence into a non-human form? I don’t think so. Not anymore.
I think we are years away from AI even approaching the human brain. Our brain processes thousands of bits of information every second that we probably don’t even think about but affect our actions. We are constantly aware of light levels, temperature, sounds, smells, and hundreds more. We are instantly aware when something in our environment changes and can react to it accordingly. Even the animal kingdom has more processing power than the most sophisticated computer when it comes to processing vision, smells, analyzing danger, etc.
Don’t discount our “sixth sense” when it comes to sensing danger whether it is a strange smell, being approached by a stranger, or just the hair on the back of your neck standing up. I don’t believe AI will ever posess these abilities.
I agree. Maybe such an AI I agree. Maybe such an AI could be intelligent to a degree, but it obviously wouldn’t be as intelligent as a human (for many, many more years at least).
AI is just human intelligence coded into a program
You could say that about programs that have coded rules. As soon as the robot learns autonomously (for example using a neural net), and also picks the examples from which it learns autonomously, and explores to find new ones and revise its models/knowledge accordingly, then I think you can speak of intelligence with a sufficient degree of independence (=true intelligence).
Currently, most AIs/machine learning algorithms have specific rules coded into them, also by hand picking and selecting the training data. So yes, a lot of bias and ideology goes into the resulting algorithm.
If an AI is capable of detecting its bias, finding new sources of information, and is able to create new theories, I think you can truly say it is intelligent (not aware or conscious, which is an entirely different matter.)
@extension of/building upon:
There is almost nothing which is totally independent of things that existed before, and I don’t think this is a requirement either for something to qualify as intelligent or not. What matters is if it can grow independently, get rid of its initial bias, and apply its knowledge/models to new situations, like people who are capable to see beyond their peer group, family, and other social environment, and truly think independently and creatively.
I’ll be going forward with BEAM robotics soon. I plan to break some of the traditional rules of BEAM (i.e, no microcontrollers; but not right away) and see where that takes me. I believe that “breaking the rules” will allow the flexibility of the microcontroller’s program to interface with the adaptivity and robustness of BEAM architectures, thus resulting in a robot with the best of both worlds.
I’ve also been struggling with whether analog or digital brains are best for the AI job, but in the end, I think there’s always more than one way to do various things- including AI (but both methods have their ups and downs, of course!).
So those are my plans. I’ll try my best to post all my builds.
Being a beginner for robotics, I’d never heard of BEAM robots before this post, but having looked online at some of these robots, I think I really like them - the designs are lovely and it must take such care to make them! I agree, it’s a shame that they’ve become unpopular. I think, once I’m done with my next robot, I might try out some BEAM robotics - I’m doing a course on electrical engineering at the moment so it’d be interesting to see what I could learn from them.
I agree with you that microcontrollers seem to be a factor in them falling out of fashion - microcontrollers are much cheaper nowadays and easier to use. Using electronic components on their own (without a microcontroller) seems to be rarer nowadays.This may be different for people not in the UK, but here, coding is well taught in schools but electronics teaching is less common, and if you look on sites with electronics tutorials (i.e. Adafruit and Sparkfun), most of the projects use microcontrollers and some form of programming. Perhaps the decline in BEAM robotics is linked to this situation with electronics teaching?
