How will it unfold?
In 1952, the computer learnt how to play Tic Tac Toe
In 1997, Deep Blue beat Gary Kasparov, the reigning world champion at that time.
In 2011, IBMs Watson won at the game of Jeopardy.
Last year, the auto pilot of Tesla car was also launched.
Narrow or Weak AI
All these examples, incredible as they may have been in their respective moments in history, are all but examples of Narrow or Weak AI, designed for specific instances.
When initially, the auto pilot of Tesla was launched, it was a failure. It crashed! However, in just 2 weeks, the car had “learnt” and was able to navigate effectively. What is noteworthy, is that it wasn’t just one model that learnt, but this learning had replicated itself in all the existing models as well. This accumulation and aggregation, spectacular as it may seem, is still in the realm of Narrow AI.
The focus right now is to make the computer truly understand the nuances of language, and emotions. How to teach the computer to negotiate with another computer or even a step higher, with another human being.
Naturally the expected outcome of all these developments is likely to be replacement of jobs by machines. It is believed that 47% of jobs will directly come under threat, and some opine that this percentage could be even higher. As costs come down, the number of people using these technologies will proliferate.
To strengthen his argument further, Patrick cited the examples of Baxter the 20K USD Robot, Hilton Hotel Concierge, Atlas Robot, Boston Dynamics and a few others. All equipped with incredible computing power that is sometimes frightening, as much as it is exhilarating. Autonomous car is just 5 years away, he reiterated. Imagine the positive impact on traffic, if these cars are able to program themselves and double up as taxis. If they can be programmed to know exactly, the time and destination of multiple pickups and drops, a facility, which can then be used to ease out on traffic. The user is able to navigate the itinerary through a mobile device, for instance. The mobile device talks to the car in real time.
Big Difference is in General Purpose AI – The First Wave
Last year, AlphaGo the computer, beat the world champion in Chinese Go, a game which is touted to be infinitely more complex than chess. It wasn’t expected to do such a thing. Not in the next 10 years, at least. The computer was fed with data from 100000 games played by humans. AlphaGo played with itself 30,000,000 times, to learn how to overcome past errors, which eventually led to beating the reigning world champion. At that stage, the developers no longer understood what the computer was doing! Deep Reinforcement Learning.
Another stunning example. The Bose-Einstein Condensate (BEC) is theoretically the coldest state possible, at -273.15 degree celcius. Anu & ADFA, an AI was able to create a prototype through laser access and minimize particle movement, in 1 hour flat.
AI is going in a direction and at a pace, which early researchers had not thought to be possible. The first wave will be reasonably benign and early adopters would be: the govt, academia and business.
The Second Wave
This is when technology becomes ubiquitous and bad habits start to form, marked by data breaches, identity threats etc. It is the moment in time when bad actors starts misusing the power of AI. Increasingly, the world is getting filled with angry people who believe that they have been treated unfairly, and there are enough reasons to self- destruct.
Unarguably, AI is scaling at a massive pace. Just like in the earlier decades of 50s, 60s and 70s, humans kept adding electricity to every device they could get hold of, now it is the turn of AI. AI can actually help us solve some massive problems which are to do with energy, pollution, clean water, for example. It can actually usher in the golden age, if it remains in the right hands.
However, in the 2nd wave, bad actors cannot be avoided. Foundation of software will be expanded by them with an intent to cause explicit harm. As Elon Musk prophesizes, “it could be like summoning the demons.” Stephen Hawking says, it could sound the death knell of the human race. For example, think about the destructive power of drones.
Perhaps that time is not too far away, when machines in their bid to get faster and more accurate, starts viewing humans as “obstacles to efficiency.” Whether this is likely to happen will depend on the exponential growth curve. To draw an analogy: who would have thought that viruses like Ebola, Saars and Zika could multiply at an exponential rate.
The real danger is in the creation of one monolithic AI which controls all other AIs. But, who might want to create such a monster and summon the demons? If we can avoid doing that and restrict ourselves to many AIs which can think exponentially to address common enough problems faced by man, then we are strongly placed to experience the “golden age of society.”