There’s a lot of fear-mongering about Artificial Intelligence and the very future of the human race. Heck, there will be some of it in the articles on this very site, and for all we know, they might eventually prove to be well-founded fears. However, this article looks at the flip side and plays devil’s advocate. Why? Because that’s how you arrive at truths, by looking at all the evidence, especially the evidence that goes against the premise that “AI is bad for humanity”.
Just a tool
Maybe a machine consciousness will someday take offence at this paragraph, but AI is still essentially a productivity tool. It is merely lines of code, written out by humans to enhance the abilities of machines to excel in areas that humans don’t. When we think AI we usually think of Skynet, the evil intelligence of the Terminator series, or else the AI of the Matrix world. Even if you think of AI as the characters from the movies A.I. or Bicentennial Man, you’re essentially feeding your own narcissism by assuming machines want to be human, when there’s no proof that machines even “want” anything.
“Cybernetic organism. Living tissue over metal endoskeleton”
AI today is so much more than the limited human-killer of fiction, or a poor little machine trying to be a human. AI is used for machine learning, perception, natural language processing, planning, social intelligence, manufacturing, etc. AI is such a broad term, that there’s no one discipline that covers it. It is each of the aforementioned fields of study, and all of them as well. AI, as we know it today, is essentially trying to replicate the functions of an intelligence – us.
AI is broadly divided into three categories – Narrow AI (specialized applications of the sub-fields mentioned earlier), General AI (a universal AI with the capabilities of multiple Narrow AIs), and Artificial Superintelligence (think Skynet). Today most AI are of the first category. Virtual Assistants like Siri and Cortana are slowly graduating from Narrow NLP and aiming to be General AI, though they are a long, long way off from even a 5-year-old child’s intelligence.
What is “Intelligence”?
How should we define intelligence itself? It stands to reason that understanding something complex requires an intelligence more complex than the thing that is being understood. This poses a problem. How can we ever understand our brains using our own brains? If you need something more complex and “higher” than the human brain to understand it, how can we ever hope to create an AI as complex as us? Richard Feynman said, “What I cannot create, I don’t understand”, so how can we ever create what we will never understand? This itself suggests that unless we can create an AI that learns and enhances itself (evolves), we can never hope to create something more complex than us.
Artificial Intelligence is no match for Natural Stupidity!
You cannot see your eyes with your own eyes, but you can use a mirror – you will see the general shape of your eyes and how they look and behave, but you will never see the inner workings of your own eyes. This is how we approach AI. We look at external manifestations of our intelligence, try and reverse engineer it using mathematics and logic, and then figure out how to translate that to code. However, it can’t just be intelligence in a computer. One school of thought (called the embodied mind thesis) feels that our functional aspects such as movement, visualisation, etc., inherently contribute to what we eventually call our ‘mind’. The experimental evidence in Daniel Dennett’s book Consciousness Explained also seems to agree with this idea. Eventually, in order to produce true AI, we might very well need to create the kind of humanoid robot that has been depicted in movies such as I, Robot and Bicentennial Man.
Reasoning and logic
It’s safe to say that the most describable aspect of our intelligence is reasoning and logic, and thus, it is the part most easily artificially replicated. Computers are built on a foundation of logic, which is why they are able to beat us at, say Chess and Go – logically-driven games.
Even the most intelligent AI robot today, if you permit us the oversimplification, still just follows lines – as our skills in coding increase the lines are becoming ever more complicated curves, but they’re still directives. But who’s to say humans don’t work exactly the same way, right?
Guess which dog can do more tricks?
The AI that exists today is logic-based Narrow AI, and excels in domains such as Chess or Go (as we mentioned earlier), but pales in comparison to humans, for example, in the area of facial recognition. The Turing Test still holds the key for General AI – testing whether or not an AI can successfully convince humans that they are interacting with another human. But…
Are we there yet?
The first industrial revolution was steam powered, the second was electricity and petroleum-driven, the third was internet and robotics/automation caused, but a fourth is coming in the form of AI. But why not as a subset of the third revolution?
The Fourth Industrial Revolution will be due to all our advancements in nanotechnology, biotechnology, robotics, 3D printing, IoT, and of course AI. However, each industrial revolution created new job categories, will the fourth take them all back?
Guess who just lost their job…
If we arrive at AI that is capable of mimicking human intelligence, but can be tied into all of the advantages of robotics, you arrive at super productive individuals who never tire, never eat, need not be paid, and cost much less than the average human in the long run. The need to make a buck and keep shareholders happy should see AI replace humans in every field, right?
Even without strong AI (one that can pass the Turing Test), robotics and AI are already taking away thousands of jobs. In fact, in a recent meeting, the World Economic Forum discussed this very issue. The bulk of the world’s stock trading is being carried out by Narrow AI whose algorithms react to changes faster than human thought. Companies like Ross Intelligence are already using AI to enhance their paralegal practice, as software can look through a ton of past files (as long as they are digitized) faster than any human associate or intern, and makes fewer mistakes too. Self-driving cars are going through hot and cold phases, but we can all see the eventuality of it being one of the first human jobs to be made redundant.
Even in healthcare, an industry one would think would be the least affected by AI and robots, doctors are off-loading the tedious task of looking at a patient’s genetic history to determine the most suitable drug (from hundreds available) for anti-cancer treatment. You can’t even be sure that you’ve been chatting with a human customer care agent these days.
Creativity and ingenuity is where AI fails, however. When faced with a new problem, the human mind is capable of thinking “out of the box” to find a solution, or even to take a call on a path that incurs losses, but minimises overall damage. AI is terrible at this (so far).
Kismet the emotive robot – Faking it like a pro!
Artificial Creativity can, however, be programmed by unnatural selection. Just like the evolution of life molecules, AI could evolve (a certain problem, if not itself) by making as many dissimilar random changes as possible and applying that to the problem-case to see if the solution is a better fit than the existing solution. This would be a kind of creativity captured mathematically… There is already research into AI that makes Art, thanks to Google. Neural networks, deep learning algorithms and scruffy techniques are paving the way, as evidenced by an AI developed at MIT that can ‘dream’ based on a given image (after having gone through millions of videos on Flickr) or the AIs at Google Brain who developed their own encryption systems by themselves.
Taken to its extreme, assume that AI can indeed replace humans at most professions… Let’s say Apple completely replaces all humans in its company (except say, designers and technical staff) with robots, as does every other big company… With so much unemployment, who is really going to buy an iPhone? Of course, there will always be rich people, but will they be the teeming millions who buy their phones now? They sell about 80 million iPhones a quarter, and if they were reduced to 1 million buyers, and even after their costs went down, they’d still need to price the iPhone much higher to make the same profit. Weak AI is good for business and bottom lines, but strong AI that makes billions jobless is terrible for business! You can count on greed to reduce the pay that you get, but their greed will also keep you employed and seeking retail therapy.
For now, it seems like acquiring multiple skills and using creativity will keep you employed for the longest… We’re just dreading the day they teach AI to write about technology…
This article was first published in April 2017 issue of Digit magazine. To read Digit’s articles first, subscribe here or download the Digit e-magazine app for Android and iOS. You could also buy Digit’s previous issues here.