Thursday, 23 December 2010

Why ST doesn't use skype

Sunday, 5 December 2010

Walking robots

I don't want to belittle the work and research of robotic engineers who have designed and built robots like the Asimo. The point is, there is no point. They are, frankly just publicity stunts. Asimo does not walk like a human being - it is kinematic i.e. the angular movements of each joint are calculated, programmed in real time. This is a huge achievement but it is not walking; walking is a controlled falling using 3D balance and constant adjustments to terrain, gradient and obstacles. The technology is not leading anywhere useful.

Our robot arms are simple (we think so) and are designed to automate simple repetitive tasks consistently and efficiently with the minimum of programming. We make no other claims about them.

Robots like Asimo are the result of a hundred years of accumulated technology. The designs must be vast, especially when you take into account the designs of the motors, sensors, processors, memory, circuits, software and so on. Yet compared to a real human being they are not much further forward than the first automatons. And still yet the entire design of that vastly complex system called a human being can be locked up in the DNA of a single cell, too small even to see.

The same problem exists with artificial intelligence. Imagine a device that can fly away from base, seek and detect using light, color and chemical analysis of the air, obtain the product it seeks and find it's own way back to base. It's called a bee. It's brain/processor is only 1mm in diameter. A human brain is 1.4 liters - 140,000 times bigger.

The very word robot is deceiving. George Devol did not call the first one a robot he called it a programmable transfer machine. Much more realistic. The drive to imitate humans has no useful purpose. And this is proven by the market - if it has no useful purpose no-one will buy it.


Thursday, 18 November 2010

words words words

I recall reading about a girl who was blind and deaf so she never knew language. I think it was Helen Keller but I am not certain. Someone taught her language through some means probably Braille and taught her how to write. Years later she wrote that the most marvelous thing about language was the way everything has a name - every object, every action, every description and that these words are pieced together to make concepts. You might suppose there would have been other ways for language to develop. Maybe we would have ended up with long strings of numbers or letters to express a single concept, kind of like the Welsh village name Llanfairpwllgwyngyllgogerychwyrndrobwllllantysiliogogogoch which has a meaning I won't bother with here.
RoboForth is like natural language. Simple easy words you just string together to make a robot concept, like TELL SHOULDER 1000 STEPS. No curly brackets, no dummy arguments in parentheses, just words.
OK, I know some Forth words have been short-formed to single characters but they are still words, just one character long.

Friday, 12 November 2010

TRUST in software.

TCG defines the notion of trust as “the expectation that a device will behave in a particular manner for the specific purpose”. For example, it is the expectation that our kitchen refrigerator will cool its contents to the level specified by its thermostat, no more (interfering with the contents placed within, spying on users opening the door, borrowing electricity to transmit wireless messages) and no less (failing to cool during early morning hours, changing the thermostat settings unpredictably)

When a user contacts me with a software problem it's amazing how often they seem to blame the software. For example "CAN'T REACH" must be some sort of bug. The fact that the robot can't actually reach because of the coordinates specified somehow is not trusted. So many users have been acclimatized to Windows applications that fall over with great regularity that they expect rock solid embedded software like RoboForth to do the same.