Imagine a time way within the future once intelligent mechanisms are ruling the globe! Humans are defeated and are currently operating for robots with no pay. There are several remaining pockets of human resistance against robots everywhere in the world however they're quick being destroyed. Humans have lost the war against technology and now there's no hope in sight!!!
Will this sound like a fantasy? That’s as a result of it is! this can be a lot doubtless to be the story of a science fiction moving picture and not the particular future reality. However, robot pessimists powerfully believe that such a future could become a reality if there aren't severe restrictions placed on AI now. Will be not helped by the varied predictions of Elon Musk like “AI could be an elementary risk to the existence of human civilization”. On the opposite hand, some individuals believe that these doomsday predictions are taking things too way with Mark Zukerberg even commenting that these predictions are “pretty irresponsible” and artificial intelligence can solely improve our lives in the future.
Some of the world's smartest people are scared. Or if not outright scared, then alarmed. The list is long. It includes Bill Gates. Stephen Hawking was there in it. And so is Elon Musk. They all warn against smart machines, which are so good at doing things due to their advanced artificial intelligence (AI) that they will do bad things to humans. The concerns are apparently real enough for some of these smartest people to take a public stand on this topic. In 2015, a letter penned by 150 top scientists, entrepreneurs, writers and others warned against the dangers of AI. There have been warnings outside that letter too. Late British physicist Stephen Hawking warned that the emergence of AI could be the "worst event in the history of our civilisation." Tesla founder Elon Musk believes that AI research would create an "immortal dictator".
The problem, however, is that beyond the obviously over-the-top and fantastical Hollywood films like Terminator and Matrix, there is not much clarity on what sort of direction AI is going to take. It is possible that with extremely smart machines one-day humans may end up looking like meat bags or monkeys that can talk. But that future is so far off that it is impossible for us, even the smartest of us, to envision it in 2018. Instead, the best we can do is see 50 years down the line. Or maybe even 100, which would be prodigal considering that 100 years ago some people in Europe were predicting that in 2018 we would be riding whales to cross oceans, a prediction that we can now call ridiculous.
Instead of predicting a far-off future -- and by far-off imagine 1000 years -- it is easier to read how AI may end up evolving in the next 15 years. And it is comforting to know that in 2035 there is not going to be a time-travelling Terminator. Instead, the danger of AI is not that people will lose control over smart machines and systems. The real danger is closer home, and it is that some people will end up having more control over AI and smart machines.
Think smart machines, not killer robots
There is no doomsday scenario, but AI is going to get better and smart machines will be exponentially smarter by 2030. All signs point towards such a future. New York-based research firm Loup Ventures recently published a report detailing the IQ test results of four popular personal digital assistants - Google Assistant, Siri, Alexa and Cortana. The firm asked a set of 800 questions to each of the digital assistants and then graded the answers based on two metrics - did the assistant understand what was being asked to it and did it deliver the correct response?
Out of the four digital assistants, Google Assistant stood first by answering all the queries accurately while responding to them with an accuracy of 85.5 per cent. Next in line was Siri, which understood 99 per cent of queries and answered them with an accuracy of 78.5 per cent. Cortana and Alexa bagged third and fourth spots in the study, understanding the questions with an accuracy of 98 per cent and answering them with an accuracy of 52.4 per cent and 61.4 per cent, respectively.
The results were impressive. But a key takeaway was just how much these machines improved compared to two years ago. Notably, what has changed in the past two years is not just the efficiency with which these AIs understand and answer the queries but also the skill set, which has improved manifolds owing to the ongoing cycle of innovations.
These virtual assistants already live inside our phones. And now they are slowly growing their sphere to everything "smart". This means they are getting into smart speakers, smart TVs, smart lights, smart doors, smart shoes, smart jackets, and smart sex toys (yes, why not).
But even as the use of AI and smart machines grows, it is important to note that in all these scenarios, AI is being used as an assistive tool or a technology that can widen the scope of impact while reducing human effort. At no point is AI being deputed as an independent entity, which is allowed to make decisions at its own discretion.
"The AI that you and I actually will be dealing with regularly, this whole idea that robots will become self-aware and they will decide that they should kill all humans, this is all science fiction," Professor Mausam from the Department of Computer Science, IIT-D told India Today Tech.
Mausam adds it is possible that smart machines, for example, an armed and smart drone, will be used to kill people in future. "Will it be a human deciding what the AI should be doing or will be AI decide for itself? I think it will be a human deciding what the drone will do. In my mind, the hypothetical scenario that someday robots will be killing humans, and robots will be one team and humans will be a different team, and that there will be a war or a scenario where robots have taken over the earth is pure science fiction," he says.
Fear humans with AI and not machines with AI
This is where the complicated answer comes in. As Mausam highlights, the danger of AI is not that it will become autonomous. The danger is that it will be used by humans, in a menacing way only humans can use something. This is also a danger that Yuval Noah Harari highlights so aptly in his book Homo Deus, in which he argues that few elites will be able to rule over -- and this will not be the rule of merciful benefactors -- over masses using AI-assisted machines. The danger of AI, argues Harari, is that it completely devalues humans.
Or seen from a different context, the immediate danger of AI and smart machines is not that they will make the world more efficient but that they will make it more bureaucratic, more chaotic. The smart systems will also put more power into the hands of a few. One example is India's Aadhaar, which uses an advanced biometric recognition system to identify people. But instead of empowering people with an ID, it instead puts its trust in smart machines, which when they fail to identify a person treat that person as non-existent.
In future, more of those smart systems will come online. Amazon, for example, is working with the US government on implementing face recognition technologies in various places. The intent is to create systems that will keep an eye on criminals, and make society safe. But that is where AI and smart machines falter. In a test recently by American Civil Liberties Union, Amazon's Rekognition falsely matched 28 members of Congress with mugshots of criminals. Not only that, the people it falsely matched were disproportionately people of colour. Imagine, an airport using smart machines and AI to ID fliers and then imagine that 10 per cent of matches fail. Once AI is everywhere and is automated probably monitored by security who have full faith in his machine, imagine the kind of havoc that false positives may end up creating at an airport, in a bank, or inside an office complex.
Then there is AI, which is expected to break down the reality as we take it for granted. Already, we have seen how much chaos smart algorithms of Facebook, Twitter, YouTube and Google have generated in the world. Fake news is rampant. Trolls have a say and a voice. Misinformation is easy to spread. Now imagine even smarter systems coming online, and completely destroying reality.
Some of these are already in testing. Google recently showed its Duplex AI that can make phone calls and talk like humans. After much outcry, Google announced that in future whenever it deploys Duplex, it will ensure that AI identifies itself as a machine. Then there are technologies like Deepfakes in the works. Deepfakes allows people to create AI-generated videos using voice and footage of real people. The result is that Deepfakes can be used to impersonate people online, it can also be used to sway public opinion. These are the concerns raised by a group of researchers from Cambridge, Oxford and Yale universities in a joint study published earlier this year.
Ethics will save us? May be
So, what's the way out? Machines will get smarter. Algorithms will rule our lives. Driverless cars are coming. Deepfakes will spread. Gradually, we will want Google Assistant to control the lights in our house. Governments will scan public places for criminals using smart cameras. Drones will be used for crowd control.
The march of AI can't be stopped, largely because it is also a solution to many of our problems. DeepMind, an AI system now owned by Google, recently developed a mechanism that could accurately read and analyse eye scans to identify over 50 eye diseases and recommend people for immediate medical attention. This system not only gave accurate results but was also able to supply a reason for its diagnosis and recommendations.
In yet another case, AI is being used to predict and save lives in events of natural disasters. While IBM's AI-powered Outage Prediction Tool is being used to predict power outages in case of storms up to 72 hours in advance in the US, one concern is an AI tool being used to inform first responders to know where they are needed the most after earthquakes, fires, floods and other natural calamities.
Mausam says the answer to the pressing problem of AI could be ethics. "Different uses of AI demand different levels of accountability. And some uses of AI demand a lot of accountability So yes, I believe there should be some limitations and we should be thinking about the ethics of AI very carefully," says the IIT-Delhi professor.
Silicon Valley to is throwing ethics into the mix, hoping that that will allow the development of AI and smart machines that are beneficial to humans. And these ethics go beyond the simple laws of Asimov. "A robot may not injure a human being or, through inaction, allow a human being to come to harm." Fair enough, that will work for robots. But more nuanced ethical guidelines are needed for human handlers of AI.
Google recently came out with some ethics that will govern its research and use of AI, and more tech companies are expected to join the bandwagon. Yet, more is needed.
When talking about ethics one important question that looms around the entire debate is if these guidelines and limitations should be self-enforced or should be decided upon by the government. "Some of these decisions will have to be taken by the government. And then (on occasions) companies like Apple, Microsoft, Google and Amazon will have to decide what is best. But I don't think this would be an either-or scenario, all of us will have to work together in creating the right atmosphere," says Mausam.