Search Engine Founder says Artificial Intelligence cannot be created
Artificial Intelligence cannot be created, so why is the term being abused by scientists and computer engineers
- Artificial Intelligence cannot be created
- Machines cannot be taught in the same manner as children after a certain point due to inherent problems with self awareness and decision making
- Theoretically possible to already create a “Knowledge” base greater than all human knowledge combined
- Theoretically possible to Pass Turing’s test with this “Knowledge” base 100% of the time
- Machines will never be capable of Original Learning
- Computer Neural Networks cannot match the processing power of the brain
- Machine Intelligence and Deep Learning is being mistaken for AI
- Singularity will never happen, except in Movies and Science Fiction
- You cannot teach a computer love, hate, longing, or what or who God is
- Can a Computer based Intelligence Hope or Dream, feel, grow old, die
- Siri and Watson are not Artificial Intelligence, just rudimentary Knowledge Bases
- Supercomputers are just a bunch of metal boxes networked together
- Man trying to play God
Every day, there is a new report, news item, scientific publication where some company or the other, some research team, some start up claims to have launched a product built with Artificial Intelligence, or to have achieved a breakthrough in this field, or promises a new product which will change the entire field.
Unfortunately, the term Artificial Intelligence or AI for short, has to be the most over abused term by scientists, computer programmers, start up entrepreneurs and the tech media alike. It is still in close competition with the term Big Data, though.
My name is Sukhbir Benipal and i am the founder and creator of the e commerce search engine benipal.com, and i can proudly say i tried for 5 years to create AI or some semblance of AI, and failed spectacularly, only to realize how wrong i was, not just in my own efforts but also in understanding the actual meaning of the term.
I have been working in this field for over 5 years and tried endlessly, at various points even believing i had a breakthrough, until one day when Hurricane Sandy hit Manhattan, and with no power, heat or running hot water, finally realizing i was so wrong, on all counts. I had not had any breakthrough in AI, in fact i did not even understand its actual meaning, what i had been able to accomplish was something totally different, a combination of deep learning and machine intelligence, but nowhere close to the AI pipe dream.
Was i wrong about Artificial Intelligence or my own efforts? The truth is, about both.
Why Artificial Intelligence cannot be created?
What really is Artificial Intelligence? Is Siri AI? Is Watson AI? What about Google and their Star Trek computer? Alan Turing, John Searle, Noam Chomsky, Ray Kurzweil, if none of them is wrong, who is more right?
Starting from early 50s and the Turing test, everyone with an interest in computing and associated disciplines has at one point or the other had an eureka moment, only to be proved, well, short. Let’s think about the term AI itself, what does it really mean? An Intelligence which is Artificial, non biological, created by humans, similar to the working of the human brain, can think and is equal to or more “intelligent” than its creator. Humans playing God.
Teaching a Child
When you teach a child something, what really happens. You are imparting knowledge and the child’s brain itself is creating billions, even trillions of relationships based on that imparted knowledge. But it is still not getting into the part of what to do with it. So let’s say, you teach a 2 year old child what a bird is. What is a bird? What does it mean? From the child’s first perspective of it, it is simply a furry little object that can flap its wings and fly away. Information obtained from both a visual and cognitive perspective, and your imparted knowledge.
But you know what, the child now associates everything that can flap its wings and fly away with being a bird. A Swallow, an Eagle, a Crow, a Parrot, a Hen. A Hen? Why a Hen, it has wings, it flaps them, so why doesn’t it fly. I am sure a lot of parents have answered and explained that question. Now what? As the child grows, it will learn on its own and formulate the relationships between birds to know which ones can fly, which ones are majestic and great, which ones are awfully ugly, and which look best on a dinner table.
Now based on this example, a lot of scientists believe the same process of learning and teaching could be applied to machines, and yes, i can confirm that is possible. I am doing it myself. But does it constitute Intelligence, and if it is, is it what is called Artificial Intelligence or Machine Intelligence developed by deep learning.
A Child, A Parrot and a Crow
Leaving a child alone with a crow and a talking parrot in a room, would let the child decide which one of those two birds it likes or dislikes, with no external input, imparted knowledge or interference. The simple task of coming to its own decision with no external knowledge is what makes humans, human.
The same can under no circumstance be taught to a machine. It will have to go through every single text, video, image known to man to decide if it should like the parrot and hate the crow, or otherwise. And even then, the decision it makes will be based on knowledge, not its own decision. As of now, it will probably like the parrot. Who likes crows anyway. To the machine, both are the same. It cannot even judge based on color, something which humans have shown great inclination for.
Just this one simple thing, like or dislike. Love or Hate. And the term AI is dead in the water.
Teenage Years and Later
Next stage in teaching a child is when he or she grows into a teen. That is when love, jealousy, hate, angst, peer pressure come into the picture. Why else is this time called the most formative years. Each and every bit of knowledge a teen learns is tinged with these factors. How do you teach this to a machine?
Rich Kid, Poor Kid, both form their own view of the world based on their environment, their conditions growing up, family being together, single parents, each intelligence is unique. No two human beings in this world can have the same experience, not even twins. How do we replicate this in a single entity? Won’t there always be someone better?
Growing up, Workplace rivalries. Politics. The Hunger to succeed at all costs. Women. A woman’s heart runs deeper than an ocean. No Man has been able to decipher a Woman, her thoughts, her feelings. Can you know what another person is thinking or feeling? Failing Health, Diseases and their affect on the human mind. Can a Artificial Intelligence feel this, know this, accomplish this?
I can answer all questions about Logic, yet i fail at Mathematics. I know Love, yet i fail at relationships. I know Struggle, yet i don’t have success. Each of these defines the human being in its own mold, and shapes its intelligence. A Unique entity that cannot be copied. Just replaced by a younger, newer version that is different.
How do you teach Artificial Intelligence if it is male or female? Can it change its own name? Can it change its own sex? This one, actually yes. Not a difficult thing for a Knowledge based Intelligence to do.
Teaching anything to a machine, even letting it learn from patterns, past behavior, data troves, statistics and allowing it the computation ability to compute any number of scenarios given a particular situation would still be limited to the extent of what it has been taught to do, the rules it has been formulated with.
When Google found Cats and Carnegie Mellon found Zebras, i can say with confidence there were a defined set of rules which allowed their algorithms to iterate on n number of images or videos, match each predefined attribute in them in increasing order of probability and to keep on doing it until a single subset was achieved, and now to match that particular subset’s varying pieces of detail with information found on the web to discover what it was. Occam’s Razor.
That is how i have done it. Because, any rational person should be able to tell you, without a reference point there is no formula. Mathematics itself depends on assuming and accepting something is already known. If nothing was known in any mathematical equation, the answer could certainly never be deciphered.
Logic, Mathematics and Philosophy are the base of any advanced computation.
Hopes and Dreams
When a human dreams or hopes, there is no limit, no rules, no boundaries. You can soar through the sky, bend time, be in multiple places at once, do things which would never be possible in the real world. Yet, you dream, and hope, essentially believing in that moment that there are no rules. And that is what separates a man from a machine.
The ability to create rules on its own, to bend and break the existing ones if necessary, and to act on knowledge, other people’s wisdom, to dream, to hope, to love, to have affection, feel pain, sadness, hope, joy. To be racist, a homophobe in private and the epitome of equality in public. Every single bit of our knowledge is tinged with each of the above attributes.
As troubling as the human species is, its intelligence cannot be artificially matched.
Who Am I. Why am i Here
If it is not self aware, to answer the most basic questions “who am i”, “why am i here”, “what am i doing”, “what is my purpose in life”, it can never transition from being simple machine intelligence. Why, because no matter what else, every single human behavior, act, idea, work, is powered by one simple truth, the human’s expiry date. If AI is to be achieved in its true form, it must first become self aware, and aware of its own fragility, its impending death.
Self coding machines able to rewrite their own code better with each iteration. Sure. Able to answer any question known to man. Sure. But no singularity. Skynet will have to learn to be dumb.
I won’t even go into the whole mechanics of human beings believing in God, because that would really be a tough one to teach a machine.
Until then, just stop abusing the term. Let’s call it what it is, Machine Intelligence. Nothing more.
Are Siri and Watson Artificial Intelligence
I think the answer has been enunciated in detail. They are just rule based programs with an enhanced knowledge base and a defined set of rules, albeit large set of rules, with the ability to process your input in Natural Language format. Confusing the systems’ ability to parse and break down the question at the input phase with AI has to be a terrible mistake for any intelligent human.
What we are doing now
What we are building and working on are dumb machines, machines with a huge knowledge and computation base, but still essentially knowledge bases. So was the Star Trek computer. Any fan will know, Captain Kirk would still do what he wanted even after being given the computer’s probability of chances of survival. The Star Trek computer is already here, there is no need to wait for it. It just needs a better Input Output interface. One that doesn’t involve a monitor, keyboard and mouse. Just Voice and Gesture interaction.
What is the big deal in a computer finding out your body temperature, your heart rate, your glucose level with a surgically implanted chip. And to analyze if you are angry, sad, low, high, having a medical condition.
Do People truly believe that knowledge itself is intelligence? Knowledge is not Intelligence. Intelligence is knowing what to do with Knowledge and acting or deciding not to, on it. Decisions made from self awareness. From original self learning. And to have each and every action tinged with the hint of mortality.
Death is at once a human’s greatest fear and the source of his greatest strength. If there was no expiry date, humans would never have an incentive to accomplish anything.
So long as Intelligence is being created by a set of rules, no matter how small they are, or is being built on a base which runs on a set of rules, it can never be termed Artificial Intelligence, and must be called by its true name, Machine Intelligence.
It is just a metal box
Computers were built from nothing by men, great scientists who had a vision, a vision empowered by their belief in science, and also by their soaring imagination. No scientist is complete without a wild imagination. If you cannot imagine, you cannot break any boundary.
And yet, even today, after 70 years or so, we are stuck with a Logic Board, Processor, Memory, HDD, OS, plus an archaic Input Output interface in the form of a Key Board, Mouse and Display.
A supercomputer with a Processor and Memory sitting on a Logic Board is worthless without a OS, nay even without a Hard Disk, and a system with a Processor and Hard Disk is worthless without Memory. So when the basic of your entire ecosystem is a rule based program that needs a specific input in order to deliver a rule based output, where is AI?
Based on what people are doing today and if they continue to insist on calling their work AI, the Linux kernel and grudgingly, Windows are the best AI products out there. And so is the AirTrain at JFK, the autopilot in Airbus’ and Boeings, the driver less cars and also your local ATM machine. Let’s also not forget the humble cash register at your local diner.
A computer, even the most complex supercomputer, is simply a box of metal, with its power unplugged, or its OS disk removed, or its memory sticks taken out. In fact, it is not even good for storing paper since there is not much space left in the box. I don’t know if any Intelligence would allow another entity so much control over its own being.
I hope you all succeed in building programs which can break captchas, destroy the Turing test, build Star Trek computers, get an answer to any question, talk and act faster than any human, respond faster to any thought, but please, for god’s sake, stop calling it Artificial Intelligence or go back to school and re learn the basics.
Or you could just create fancy new words to suit your own needs. Modern Humanity has consistently shown its competence at that.