Translate

Thursday 8 February 2018

Part 44 - I think, therefore am I?

The Ehang 184 which I have discussed right back in my first blog entry, is now carrying people in China as part of it's advanced testing and expects to have a commercial version of this available this year.
From the video it looks like they now have a two seat version as well.
 There appears to be a bit of a war to decide if Uber or Waymo is going to be the leaders in automotive technology and it has gone to court. Industrial espionage and all the juicy ingredients make this a battle to watch.

Meanwhile, back in Oz........

Robotics is a funny old thing.
By that I man it is mostly about perception.

If a robot looks and sounds human then may people think that this is an advanced intelligence that thinks for itself.

Not necessarily so.

I remember writing a program many years ago in basic.
What it did was take any input and turn it back into a question. I called it "Shrink".

I had it ask "what would you like to talk about"?
I then rephrased the input and elicited a response.
I had a list of keywords. If the response didn't have a keyword I would ask "Can you expand on that a little bit?"
This gave the impression that the program was listening and quite often the expansion would elicit a keyword that could get a response.

It wasn't listening as such, but psychiatry works on asking the right questions and having the subject follow their own path and come to their own resolution.
This program did just that.

It wasn't smart but sounded like it was even though it was as dumb as a brick.
It's all about perception.

AI is not sentient, it's just getting a machine to learn.

The dangerous part of that is having a machine make it's own decisions and allowing it to act on it without any control.

For instance, if you had a computer with the power to control every aspect of your environment with total autonomy and then introduced the ability to control your health, then you could have an issue.

For instance you could say " Computer, wipe out cancer in all humans".

Great!.

Maybe not. This is a machine that works on logic not a human with the sum of knowledge accumulated by years of growing and learning without being able to cause a lot of harm.
You have a machine that has total control, "Wipe out cancer"? sure. The logical way is eliminate all humans and the problem is solved!

Not the optimal result from our perspective but an elegant solution nevertheless.
There is no malice in this action and no intelligent response either. Just a machine response.

Coming back to robotics. If you build a robot it is a machine, just like a toaster and just as smart.
But perhaps with the ability to learn. That learning will be a logical response not a human one.

If you have software that controls it then it is just process control with no real smarts and no real danger, unless you let it be dangerous.

A car is a machine with no danger to anyone when it's sitting still. Once you start it and drive off then it has the potential to kill. Humans kill many people every year with cars either accidentally or intentionally and hopefully robotic vehicles will put an end to much of that.

But only if it is intelligently implemented.

Imagine the situation where you have a totalitarian government and they have control of all autonomous vehicles. Dissident? not anymore. Oops sorry his car malfunctioned and drove off a cliff.
My bad, we'll fix the software.

Anyway, back to the humanoid robot.

No matter how human it looks it's still a machine, just like a car and just as safe or dangerous as the controls around it.


I have now built a robotic arm as the first stage of the Imoov project (inmoov.fr) and downloaded the software to make it work using the preconfigured myrobotlab.

It looked really cool when installing, looked like a hacker sequence from a movie with the green text rolling down the screen for ages.

Once I got it installed I found that it is actually voice activated.
I then tried to respond when it asked my name and tried for some time to get it understand that my name was Clive, not hi or hello or any other variant except my name. I even had my wife try and she had no better result.

I then remembered when I worked in Taiwan at one stage no one could understand my Australian accent so I had to put on an American one and then people knew what I was saying.... Such is life.

I tried it on the computer as Cliiiive with the best American accent I could muster and, wouldn't you know it, got it right first time.

So OK..

Being voice activated I thought I would jump right in and after setting a few things up I said to it "close hand". It responded "closing my hand" and lo and behold it did. Cool.
Not to be outdone I said "open hand" and it dutifully did.

Oh the power....

I commented to my wife, "this is quite sophisticated stuff" and to my surprise the computer responded with "Thank you. I try."

So this is my entry into robotic software control. Years in advance of the basic programs I wrote with my first robot arm version only a couple of months ago. No longer will I try and re-create the wheel as this stuff is light years ahead of anything I could do.

So now, my programming skills shown as woefully inadequate, I will push on with the build and play with real software to get a better understanding and keep you posted as it happens.
And have fun in the process.

Who said getting old has to be boring...

Cheers











No comments:

Post a Comment