Artificial Narrow Intelligence is something we’re all familiar with since the advent of smartphones, Siri, the Google Home Controller, and the Kuri Mobile Robot, just to mention a few of the intelligent objects inundating society nowadays. After school cartoons were favorite distractions, especially the Jetsons with their robot housekeeper, Rosey. She was a marvel to behold because if you’ve ever kept a household in order, you know that narrow AI would not be adequate enough to master the demands needed to make and serve coffee or organize meal preparation.
We think a lot has happened in Artificial Intelligence and pretty soon the fourth Industrial revolution will transform our world into a utopia where everything will be accomplished for us without much effort.. We’ll be able to relax because our machines will have everything under control and our lives will be similar to that fictional resort WestWorld where nothing can go wrong… wrong… wrong.
Much has happened in the AI domain, but the bulk has been in narrow AI, and not in general AI which is disappointing plus simultaneously confusing. With so many businesses investing in things like robotics, automated warehouses, driverless logistics vehicles and so forth, why do we believe a lot of progress is being made in the AI domain? Most AI developments are one-shot wonders capable of executing one task extremely well. Most AI products need human supervision to work without a glitch occurring sometime.
Euphemistically known as the father of artificial intelligence, the late MIT professor Marvin Minsky, once said AI was “the science of making machines do those things that would be considered intelligent if they were done by people.” Roughly stated then, one could say we simulate the conditions to fool us into believing that a machine is smart or clever enough to pass itself off as a person. The collection of bits and bytes organized, by a human, and packaged as software instruction sets, is inserted into a mechanical device and creates an illusion so convincing we believe that God is in the machine.
Infosys CEO, Vishal Sikka, also rephrases Minsky’s take on AI as “any activity that used to only be done via human intelligence that now can be executed by a computer,” including speech recognition, machine learning and natural language processing. The power of NLP was recently unveiled in beta tests of GPT-3. This language model developed by OpenAI Developers has an impressively diverse range of applications such as an all purpose Excel function, a recipe generator, a search engine and several others. Unfortunately, a bias of hateful racism and sexism spewed into documents the AI was generating, and the product is in abeyance until this problem is resolved. The over dimensional parameters make this AI an electronic scrapbook marvel with problems.
Anyplace you look, the machines we have now are generally based on “narrow AI,” which addresses single-task applications. This includes playing strategic games, language translation, self-driving vehicles, image recognition, and automated surgery. Even autonomous vehicles still need a human in the loop to make critical judgements when the software makes a critical judgement error that could cause damage to property or loss of life. Is the vehicle savvy enough to decide that the masked human, with a gun in hand and waving the vehicle to stop, means no evil? If the vehicle were programmed to stop upon recognizing certain hand gestures while underway, it would certainly stop, and with disastrous results.
Artificial Narrow Intelligence then is oriented toward specific tasks while Artificial General Intelligence is aiming at general cognitive abilities. This type of AI is a nascent concept and in the world of theory. Domains like language processing, vision, reasoning, ethical and moral judgement, and fairness are human-level cognitive functions. We aren’t there yet to say we have AGI in the palms of our hands.
Building an AGI system would entail coordinating thousands of ANI systems, all working in concert like the Borg from Star Trek’s fictional universe. Jane Jetson’s Rosey could be achieved today using IoT (Internet of Things) technology and some clever Python code. She could have a Wi-Fi interface orchestrating the coffee automat, refrigerator, lights, automatic window blinds, security systems etc. to command the household. All devices could be linked to a humanoid type-robot that would make Disney’s Small World look like a cheap midway attraction.
Unlike machines, our neurological system is much more complex and has an entirely different organization. We have 86 billion nerve cells that communicate with each other in ways we still don’t fully understand. Connecting these neurons are trillions of synapses which are the communication pathways. Man hasn’t built a machine that complex yet. Even IBM’s Watson took an amazing 40 minutes to simulate a second of neuronal activity. AGI will come, but it won’t be tomorrow.
Many simple things we do, without much thought, are often complex instruction sets for a device to accomplish.For example, deciding whether an animal is a cat or dog is easy for us after we have had experiences with the animals. We look, and know within a second if it is a dog or not.
In computer vision, the data sets are large and, if the background adds noise to the field being interpreted, a false identification is made.
Aiding human capabilities and helping us make decisions with sweeping consequences is the charter of Artificial Intelligence. Technically that’s correct. But what about philosophical considerations? Here, Artificial Intelligence can potentially aid us humans to lead lives devoid of toil and seek the meaning to life. The great hope of AI is to assist us in managing our interconnectedness to function in harmony that’s mutually beneficial to our race and planet.
When AGI finally arrives, embrace it fully. Either this will be the most marvellous tool man has ever made or it will be the beginning of our extinction as the world of IoT makes us a comfortable bed for our last sleep. Rosey can keep house for me then.