Research for the Malignobot

Greetings and Lamentations, I AM Lord Malignance, and you WILL Crouch before me Now!

(Villains excepted of course)

One tries to keep abreast of trends and breakthroughs in various fields of science and technology, and found a remarkable article with links to others on Scientific American. The article describes significant breakthroughs in Artificial Intelligence (AI) and the profoundness of what it implies is exciting. The Scientific American website is rife with copyright warnings however, so One won’t cut and paste their article here, and will instead paraphrase and report on the piece and its linked information.

The much better source material can be found here.

Hod Lipson (was there ever a more apt Star Wars/Villain’s name?) Roboticist at Cornell University’s Computational Synthesis Laboratory has been doing research in programmed self awareness in machines. These constructs have built in and evolve means to reflect and analyze their own methods of thinking. Conceptualize it as that voice of reason/doubt/inspiration that runs in your head, analyzing your performance in your daily activities and interactions with other people. This process is called “Metacognition” and it applies when you try something and it doesn’t produce the desired results. It is this voice that analyses performance and develops new strategies. Long accepted as fact is that Life learns by making mistakes and coming to understand how failure occurred. Animals (and plants?) follow this same logic to success.

The article by Charles Q Choi from February 24^th, 2011 (One IS two months behind from the Accounting) goes on to describe the strengths and weakness of current robots: They perform flawlessly in perfect environments. If the conditions change, the robots invariably fail, sometimes superhumanly. The breakthrough the researches have made that is so significant is, “What if a robot could encounter a variable in operation or environment, and correct for it?" This isn’t a flowchart of IF/THEN options, but a true analysis of all input and a trial and error adaptation. For this test they built a four legged star-fish type robot and impelled it to learn to move around – and it did after a time. Then the researches removed a leg and presented the construct with a problem – “how to move now, that the current understanding of movement was no longer valid (it didn't work any longer)?”. The robot, which had developed an understanding of its own self image (body mechanics), had to adapt to moving without its missing leg. Biology takes these adaptations for granted – the urge to survive being the programming of life. But in Robotics, this is very significant.

Another experiment demonstrating Theory of Mind had one robot observe the actions of another, and over time predict the movements of the other. Humans to ants all have this ability because of socialization adaptations programmed through trial and error by millions of years of evolution. In robotics this is also so new and exciting.

These robots run algorithms which randomly select models of predicted results. They then select the model and the action to take, and then observes the result. If the result was positive, it then throws out the models used for that action that predicted failure.

Robots that could be careful, that could intuit and predict best chance of success scenarios. Survival, is no small feat, as it has taken and is taking biology the time of all life on Earth to develop.

This tangents to Singularity theory, which tells us that there may come a time when technological progress and the leaping forward evolution of Artificial Intelligence become so quick and so profoundly changes the world, that the future becomes more unknowable. These changes accelerate over time, until future realities become unimaginable possibilities.

One imagines such a possible future, where One’s metal horde of Malignobots has enslaved the world in perfect merciless, ruthless domination. It will be good (and by “good” One of course means “Evil”) to be,

-Lord Malignance