In 1950 Isaac Asimov came up with the 3 laws of robots; they cannot harm humans or let them come to harm, must obey command given by humans and must protect its own existence however must not conflict with the first law. There are obvious problems with all 3 of these laws as roboticist Daniel Wilson “They are written in English. How do you program that?” Which is understood as you can’t program the artificial intelligence to follow the laws given because of conflicts that would possibly crash the system. Another reason is also how they are already being used in today’s world. You can’t have military drone that is not able to do its job because of these laws, there would be no point to it. Furthermore, dangerous people would have access to these robots that follow any command given for example sending it out on a suicide mission, which would go against its third law to survive and possibly the first law, if it is to defend itself to survive from a human attacker. The laws are not what militaries would want built into the AI system, they want the opposite of Asimov’s laws [1].
Creators of the AI will program how they want the robot’s use to be, most likely disregarding the 3 laws. When they come to programming it in their own way, the AI system will fall into 3 categories; weak AI, strong AI and Specialized AI. Weak artificial intelligence is programmed for the robot to respond to commands that are established beforehand, for example, you ask Siri a question and the AI will retrieve data relating to the question. These AIs have no understanding of their tasks, they only know to execute the command when given. Specialized AIs are programmed to do only specific special tasks and are limited to what they can do. An example is a robot arms in an industrial factory. Strong AIs can carry out their job on their own without supervision once their programming has been complete and uploaded. This AI is programmed to be in humanoid robots which can sense the area around them and interact with objects in their vicinity. They are also implemented into robotic surgeons which are becoming very popular as there is no point to interrupt the robot once it starts its job [2].
Carnegie Mellon University grad-student, Feng-hsiung Hsu was working on a project to create a chess playing computer, along with classmate, Murray Campbell. The project was called ChipTest. Both were hired to work for IBM and continued their development of ChipTest, with the help of other computer scientists to create Deep Blue. In 1996, Deep Blue lost against Kasparov at first, but in the rematch Deep Blue took the win. Deep Blue helped understand how far AI can go to solve difficult, complex problems [3]. In 2007 IBM created Watson, named after founder Thomas J. Watson Sr., an AI capable on answering questions. Watson was highly noticed after defeating Jeopardy’s two best champions in 2011 and from then on has developed expensively. Today IBM offer Watson as an AI assistant to help with complex situations and problems [4].
At the MIT AI labs in 1999 the first expressive humanoid robot, Kismet, was created. Kismet can engage in natural and expressive face to face communication. Programmed into Kismet are theories and concepts relating to psychology, evolution and ethology to help it learn and respond from humans. For this to be possible, Kismet was built with audio and visual sensor for social cues where it would then calculate an appropriate response, vocally, physically and facially [5].
Deepmind’s AlphaGo beat Go’s 18-time world champion Lee Sedoul at Go. AlphaGo has had 3 successors, AlphaGo Master, AlphaGo Zero and AlphaZero, all much more powerful than AlphaGo [6]. This shows how far AI has come and shows that there is no stopping any time soon and will continue to develop powerful than the last.
Ways of programming AIs into robots will become more complex as time goes on. There are robots programmed to assist the elderly, robotic surgeons, assistance robots for home and robot kitchen. There are also robot animals being developed [7]. Will AI and robots continue help us in the future or fight against us? It is uncertain but either outcome could become a possibility. An AI can hold a lot of information, be programmed for certain tasks and converse, how long until they become self-aware and are able to learn by themselves? Only time will tell as more and more progressive developments are made every day to create powerful machines.
Comments