Tilden and BEAM robots - This is Not intelligence, artificial life, or neural nets

I'll give the guy this: Tilden has implemented reflex into his BEAM robotics However, I don't see any of these bots doing much more than this. It seems to me that these things are just mechanical zombies. They don't do any thinking, they have no memory... they simply react to different stimuli...(which yes that is one part of the search for AI but it has been done before... this is not Tilden's discovery) These bots are no more than fun moving art projects.

At best, it seems to me that he is essentially just doing neural net learning manually... i.e. if one thing doesn't work he physically changes it to see if something else will work.... That isn't too natural. I agree with Boskovich, in that this is alchemy of the mid-evil period... not science.

All in all, let's face it. BEAM robotics is far from realm of artificial intelligence. BEAM will amount to little more than cool looking mindless mechanical devices wandering around, bumping into things and perhaps doing simple tasks with no memory of what they are and have been doing. In Tildens world an oven (which reacts to a set heat level) is intelligent.

The only promise to Tildens BEAM robotics, and all others in the field of artificial intelligence is if it is taken to the next step... that is much more complex networks *on the order of billions* which can self adapt (much like the brain) and carry and utilize memory.

Tilden seems to have achieved little more than building very simple analog microprocessors from the ground up.

Also, the A in BEAM... does it really need to be there? If the goal here is intelligent evolving robots... who cares about aesthetics?

BEAM robotics is great for the amateur robotic enthusiast, but it doesn't offer much promise for true future development in the area of artificial intelligence. Comments?

- Annan Mozeika

Reply to
annan
Loading thread data ...

Everyone is certainly entitled to their own opinion, but who says the only future for useful robotics development is in the direction of artificial intelligence? We are surrounded by examples of successful organisms, who you could just as easily refer to as "mindless zombies", reacting to stimuli to survive. If a robot doesn't 'need' to remember to perform its task, why give it a memory? AI will doubtless play an important role in our future, but in reality it is often best to have 'just' the right tool for the job.

I do understand where you are coming from, as I'm a convert myself. Far more than traditional(?!) robotics, BEAM has taught me to remember there is so much more to a robot than just the processor.

To address the aesthetics issue... well... I guess beauty is in the eye (or photodiode) of the beholder. I've always assumed it was meant as a "beauty of form" or a "beauty of function" thing, and not necessarily a "pleasing to the eyes" thing. Take a spider, for example. Kind of a hideous looking thing, unless you get a chance to really study one up close. Everything is so balanced, so perfectly evolved for its role in life. No extra parts, or wasted thought processes with a spider, just a living machine, reacting to its environment.

-Anthony Martin

Reply to
Anthony Martin

One of the things I like about BEAM is the simple mechanics. Take one of his simple walkers, add a microcontroller, and you have a simple, programmable walker. Not a 12 or 16 axis walker, but a walker nontheless.

Reply to
Mark

I've never been a huge fan of the "AI" aspect of BEAM either, although in the aspect of adjusting its own behavior based on motor loads and stimulii, it works well, and simply.

Bug intelligence is all I'm after, and Mark's building blocks get me a good way there.

It's also been the very successful building blocks of the Robosapien (and other) walking robots. I've got video of the prototypes using BEAM tech where the protoSapien really scoots across a table, much faster than the commercial version. Regardless of the perceived rhetoric, it's very hard to dispute the results.

Aesthetics is a HUGELY overlooked aspect of robotics. A clean, neat design is naturally more robust and able to weather the wear'n'tear of a real environment. Of course, our goal is to build self-sufficent devices, not necessarily "intelligent evolving robots", in which case, a laptop duct-taped to an inverted garbage bin with wheels is more necessary.

If A.I. doesn't seem to be a suitable "classification" of BEAM-style robotics, that's ok. It's being well received by the Neuromorphic community, which is more concerned with implementing the same principles in real nervous systems into artificial ones.

S'ok with me - I use BEAM as I get more satisfaction out of building with a soldering iron than with a keyboard (not that the two are mutually exclusive). I build robots to see what they can do, not to develop an advanced AI.

Regards, Dave Hrynkiw President, Solarbotics Ltd.

Ref: Protosapien -

formatting link
-
formatting link

Reply to
Dave Hrynkiw

We just developed & released a 2-motor walker with an option for a microprocessor "rider". It isn't needed, but a uC can totally subsume all walking functions, or just influence them when need be. "Horse and Rider" topologies have worked for us in a few designs so far, and seem to be effective.

Regards, Dave

Reply to
Dave Hrynkiw

I couldn't have said it better, nor agree more. Thanks for stepping up to the bat! Steve

Reply to
srgrimm

I agree with the 'horse and rider' concept. The less we have to worry about basic actions, the more we can concern ourselves with complex interactions and layered behaviors. As far as modelling the human mind in our creations is concerned, we should also consider the learned behaviors we take for granted as complex beings. Our creations should be able to walk and chew gum, even if one or the other is only an innate reflex. Then we can pop a bubble while waving goodbye to the alchemy of this growing field. It would be interesting to see a learning model in which PLD or some similar device was used to convert a 'soft' behavior into a 'hard' one, bridging the gap between BEAM and conventional software AI practices. The human brain does this to convert short term memory into long term memory. In short, BEAM has its place among the under-appreciated monotonuous, repetitive and boring actions that have become the foundation of an exciting new science.

Reply to
sinner

I'm not sure, but I would not be at all surprised if the A in BEAM came about simply because it made a nice acronym. BEM somehow doesn't sound as good. I am unfamiliar with the history, but I can easily imagine Tilden making little robots until one day he noticed that they all tended to mimic aspects of nature (the B) while using mechanics and electronics (M and E). And you know what, they looked pretty cool, too (the infamous A).

You will notice that some of the world's greatest engineers broadened their horizons beyond the simple function of their designs and took into account form as well. Burt Rutan is my favorite example. He has designed some of the prettiest aircraft in the sky. And the most functional: Take Voyager, for example, which flew around the world non-stop, non-refuelled. Or the team of aircraft that won the X-Prize recently. Much of what we design is meant to interact with humans, so its form, the aesthetic, is part of its function (to please it's human masters). I know that BEAM robotics is like that for me.

It should be noted that form and function often go hand in hand anyhow, particularly for engineers. (I think we often subconsciously recognize a design's likelihood of success and sense it as beauty - but that's for another discussion.)

Not to get too far off topic, I must say that I am unfamiliar with Mr Mozeika's opinions on the AI aspect of BEAM and so cannot really comment on them. And I think that was the main part of his post. Anyhow, I was just commenting on the part of the post that most affected me. I hope I am accurately and adequately expressing the feelings of others interested in BEAM robotics as well as my own.

Andrew Barthle

Reply to
Andrew Barthle

anyhow,

recognize

Dave has some nice pictures of the early R.sapiens on his site. I missed these the 1st go round .... I assume the commercial version has a microcontroller in it. Anyone know about this? I also notice the solder sucker near the Biodroid, so someone else gets some fun out of those projects too.

formatting link

Reply to
dan

PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.