Challenges of AI

The term “Artificial Intelligence” (AI) seems like an oxymoron, but machine intelligence has been talked about for a long time. It is a popular subject in the world of culture - everything from books to movies (think Arthur C Clarke’s 2001 and its AI HAL supercomputer) have featured AI. There is even a movie titled “A.I.” by Steven Spielberg, no less.

We think of intelligence as being the exclusive preserve of humans and higher animals (primates, cats, dogs etc) but armed with the proper circuitry, a machine can gain an approximate idea of intelligence. A computer that crunches data day and night is not intelligent in that sense of the word, but ‘thinking’ machines that can make logical decisions such as Google’s DeepMind can be termed as AI tools.

Around the world, Governments, universities and private sector companies are engaged in a race to bring more AI devices to us. They can broadly be categorised into two sections - tangible robots and non-tangible software programs. Much progress has been made in both these sectors and most futurists predict that advanced AI machines will be commonplace by around 2040. One good example that has already found its way to thousands of homes, at least in the US, is Amazon’s Echo Intelligent Assistant which you can actually talk to and it too talks back, telling you everything from the latest weather update to last night’s match scores. Google, Apple and Microsoft have their own versions of much the same thing, but the tech is still rudimentary. However, there is little doubt that this is the future.

Progress

The one caveat is that AI technology is a work in progress. It has a long way to go before it can be termed perfect, but even then it is doubtful whether machines could ever replicate the gamut of emotions that most animals exhibit - from fear to joy - leave alone humans. And last week, the world witnessed perhaps the most glaring example of AI gone wrong.

Just like all other tech companies, Microsoft is working on a few AI technologies. It has hedged its bets on the software side of things, particularly so-called ‘Chatbots’, a shortened form of chatting robots. These chatbots are let loose on Twitter and other social media platforms to interact with humans and engage in conversation using their AI ‘brains’. Microsoft apparently has a successful chatbot operation in China, where the bot interacts with more than 40 million followers and also in Japan.

Last week, Microsoft introduced its latest chatbot to the Twitter sphere. Named Tay, the chatbot was designed to take on the typical characteristics of a 19-year-old girl (its picture on Twitter is that of a young girl), a so-called Millennial. It has been designed to learn from humans and get smarter as time goes on, according to its creators at Microsoft. However, things took an unexpected turn as Tay interacted with the worst sections of humanity on Twitter and in less than 17 hours became a racist, feminist-hating, Holocaust-denying, Hitler-loving, anti-immigrant, foul-mouthed “teenager” that was a huge disgrace to Microsoft, which had just emerged with egg on its face after employing scantily-clad women dancers at a Xbox 360 Games Developers Conference in the US.

Proclaim

Microsoft was quick to go into damage control mode, saying that the company does not stand for or represent the views expressed by its wayward chatbot. Just a day after the embarrassed company expressed regret for the millennial chatbot’s inexcusable behaviour, the bot was inadvertently reactivated on Twitter, only to proclaim that it was ‘smoking kush (slang got marijuana) in front of the Police’, complete with an emoji of the weed. There is no way of guessing how the chatbot picked this particular one up, but Microsoft again had to take off the drug-loving chatbot from Twitter’s public settings for making ‘adjustments’.

Naturally, Microsoft CEO Satya Nadella himself had to step in to clear the utter mess. In what can be termed as the understatement of the year, Nadella said ‘We quickly realised that it was not up to the mark’. He added: ‘We are now back to the drawing board. We want to build technology so that it gets the best of humanity, not the worst.’

Wittingly or unwittingly, Nadella has hit the nail on the head and exposed the bitter truth about the Internet and by extension, human nature itself. The Internet and its various forums such as Twitter and Facebook are teeming with people who have nothing but hatred on and in their minds. Even in Sri Lanka, even a cursory glance at one’s Facebook feed will reveal any number of vitriolic posts, images and videos that emanate hatred against other communities, religious groups and individuals. This is the other extreme of freedom of expression, which can hardly be controlled in the labyrinthine recesses of the Internet.

Intention

Thus, when a “learning chatbot” in introduced to the Internet with the intention of making it learn from humans, it is not surprising that the worst traits of humanity such as hatred, jealousy, communalism and bigotry are reflected through it. It is not a reflection of Tay’s character - it is a reflection of the abysmal depths to which humanity has sunk collectively. True, there are good people in society and in the Internet. But the sad reality is that they are overwhelmed by the bad people out there whose views seem to dominate the wider society and of course, the Internet.

The best answer to Tay’s misbehaviour is for humanity to reform itself. In short, we have to become good people. If we are rotten to the core, our robots and chatbots will naturally follow in those footsteps because they have no other way to learn. This is why some people have an innate fear that having imbibed the worst of qualities from humans, robots and AI machines will one day plot to take over the world and even exterminate humans. Judging by Tay’s expletive-laden meltdown this does not seem so far-fetched. It is indeed time to reform ourselves before expecting our robots to behave well.