More often than not, the media hypes up certain news from the Artificial Intelligence(AI)world which it deems to be sensational. This is probably good for getting more clicks and views but it doesn’t work in a favorable way for the advancement of the field or to generate interest in the subject. Hollywood fans and doomsday believers like to dream of a dystopian world where AI and robots will have taken over the world and rule over humans in the future. Indeed, a few news articles have generated considerable buzz on the Internet and while they deserve our attention and appreciation, it is not something to be scared about.

Facebook chatbots inventing their own language was an example that created news sometime back
https://www.independent.co.uk/life-style/facebook-artificial-intelligence-ai-chatbot-new-language-research-openai-google-a7869706.html
Google’s Deepmind making AI fight against AI is another example
https://www.theverge.com/2017/2/9/14558418/ai-deepmind-social-dilemma-study
But nevertheless, here are the top three reasons why there is really no need to be imagining taking orders from your AI overlords in the future.

1) Ongoing research to create trust – Indeed, AI practitioners are thinking about this too. That is why they are currently trying to develop products that are reliable and acceptable for all. Across the industry, the first thing which researchers think about after making an accurate model is to make it unbiased and generalizable. Many big companies are investing a lot of money, time, and effort into ensuring this. If this responsibility continues towards making AI work, not just accurately but also reliably, then it could prove a game-changer and a significant step in human history.

2) Legislations and Regulations – More and more countries are coming up with legislations around the practice and development of AI. This involves the use of understandable algorithm predictions, and proper data protection and handling mechanisms. Since most of Machine Learning is data-hungry, having data protection measures, especially for personal data is very important.

3) AI can’t be generalized beyond a certain level – Up until now, and perhaps for the foreseeable future too, the kind of Machine Learning, Deep Learning, or even Reinforcement Learning that we have seen is very narrow and can’t possibly be generalized for all cases and under all situations. Moreover, extrapolation and recognizing things that are not in the memory of AI are quite hard challenges even today. Therefore, to imagine a fully self-conscious AI is nothing more than the imagination of some people, and quite impossible to create when we are still grappling with questions on consciousness, memory, and intelligence.

Therefore the question is, what should the role of AI be in our lives? How should it look like in the next ten to twenty years? This brings us to the Butler analogy, which the following paragraphs summarize brilliantly –

Ideally, a butler is always available, prepared to do the tasks he is asked for with few questions and no complaints. He rarely interrupts his employer to suggest ways in which he may be helpful and in case of a problem, he will find a way to fix it or work around it without much bothering the employer. Instead, he pays attention to the habits of the employer and observes what he has done in the past so he would be able to anticipate what may be wanted in the future – thereby carefully avoiding to exaggerate this anticipation because he knows that it may be more costly to do something unwanted than it is to refrain from taking the initiative. Besides this behavior, he makes a special effort to be courteous and respectful, even in a situation when asked for something he may not be able to do. They learn to collaborate.

The employer on the other hand will not ask the butler to tell him how to make requests, but instead observe the butler on what he can do and what he does well and thereby gradually learn, by picking up on his feedback, that asking one way rather than another will result in better outcomes. This complementary behavior may over time, if they have a good relationship, lead to the result that they learn to work together without noticing the interaction.