Comments on: Honda (barely) celebrates ASIMO’s 10th Anniversary

By: kert

kert — Mon, 01 Nov 2010 17:47:55 +0000

You probably know what i meant, but i will elaborate : a really basic, simple chess program checks each and every possible move.

Thats brute force. And it works, to a degree. There are tons of smarter approaches to it, but just throwing loads of computational power at problem helps with certain problems.

Certain vision algorithms can also be brute forced. Especially if you have a very large database to plug in from the backend ( see Google Goggles )

And so on. Basic _components_ of AI can be helped by just throwing massive computing power and/or very large databases at it. Thats why i put “intelligence” in quotation marks, you CAN solve subproblems of general intelligence by brute forcing it.

By: alex

alex — Sun, 31 Oct 2010 20:59:59 +0000

to this: “that a lot of \intelligence\ can be basically solved by brute force computing.”
I have to say: lol
nuff said…

By: Robotbling

Robotbling — Sun, 31 Oct 2010 19:37:53 +0000

@ kert

Sure, there are improvements. But I think Honda will probably wait until there are major breakthroughs before they redesign ASIMO. But who knows?

By: kert

kert — Sun, 31 Oct 2010 15:56:53 +0000

There is tons of improvements in basic tech still going on. MEMS sensors improvements, tactile sensors come to mind first.

Battery technology keeps improving at rapid rate, now driven further by automotive applications.

I have not seen many demos of what level of dexterity is built into ASIMOs hands ? Can it grab an egg without breaking it ?

ASIMO intelligence : i am not sure if anyone has really tried, but i would plug high-level AI processes of that robot into cloud/grid computing, immediately. There is so much processing power available for cheap ( a bit high latency, yes ) that a lot of \intelligence\ can be basically solved by brute force computing.

Leave low-latency tasks like basic sensor processing, actuator controls, balancing etc on board, and offload all higher level \thinking\ to very powerful grid computing backends.

By: alex

alex — Sat, 30 Oct 2010 08:17:03 +0000

yes, I meant that it’s not useful practically without the intelligence – it’s good for research and all but you can’t use the robot outside the lab for real work. And A.I. could be used without a robot too, just in a computer -> HAL :)
For humanoid robots the reality is they don’t know how to make an A.I. so they at least play with the electronics and mechanics so they have something to do and then some time when they have an A.I. they can at least use it directly with some robot :) – industrial robots are something else, they can be used now

By: Robotbling

Robotbling — Fri, 29 Oct 2010 19:54:02 +0000

@ alex

Gotta disagree. Think about how much tech is involved in just building a robot like ASIMO that can walk and hold its balance the way it can. There’s definitely value in that. Also, just building ASIMO has caused a huge wave of inspiration around the world to build humanoids. We need systems like ASIMO before we get the intelligence anyway, otherwise the intelligence will be useless.

By: alex

alex — Fri, 29 Oct 2010 19:38:03 +0000

humanoid robot without some intelligence is not so useful – you can mostly only use it for show and presentations
- the only possible solution now would be teleoperation – like those flying drones but a remote terminator :) , i.e. for the use in dangerous places…maybe humanoid robotic astronauts – like that nasa one but with full body…

By: Robotbling

Robotbling — Fri, 29 Oct 2010 19:14:14 +0000

True, but I think even in its current state it would be pretty much ready to go if it had the necessary A.I.. We don’t need to duplicate an exact human being for it to be fully functional. I’m not sure how much battery technology has improved in the last 6 years (2004 was the debut of “New ASIMO”). By the time we have better actuators (more like human muscles) Honda will probably have to go back to the drawing board or buy Boston Dynamics and have a truly dynamic biped.

By: kert

kert — Fri, 29 Oct 2010 18:59:25 +0000

ASIMO development has pretty much stopped, hasnt it ? Sensors and actuators, processing power and not to mention power sources have kept improving over recent years, but there is nothing new being integrated in ASIMO.

Its not like its on par with human body already, so there is tons of stuff still to be done apart from the intelligence research part.

If they realistically want to do a robot soccer vs human teams in 2050, there is a long way of improvements still to be done on the physical parts.