Humanizing AI – Wait, Is that a Thing?

I have been noodling over this concept and reading about it randomly for a while now. But when it came up in one of the marketing discussions earlier this week, I decided to put it down here. It turns out that the response is relative to the context. Here, I attempt to put some high level dots on the map (looking at it from space), and identify what humanizing AI really means in relation to a few AI-centric mega movements.

Google’s Brain Team is on a mission to make machines more intelligent and has been hosting various thought leadership programs on this. The MIT Media Labs started the AI More Than Human drive, bringing artists, scientists, and researchers to showcase their interpretation of this concept. Microsoft has been leading some research and development on humanizing AI with deep learning to make machine interactions more human by detecting the underlying emotions and sentiments. These are just a few of the many initiatives that are underway.

Conversational AI – Machine Talks like a Human

Imagine that you’re trying to shop for a sympathy card and wanted to get some help to find the right one. You decide to work with a conversational bot. The big bot throws a plethora of card options out as soon as you rattle the word ‘help me find a sympathy card.’ A few other options that come up in the results are - congratulations, best of luck, and happy birthday?!

Source: Shutterstock

What would have been nice is if the bot parsed the context and filtered-in the word sympathy. And instead of giving a pre-built response via rule-based conversation processing, it could calibrate the emotion/sentiment/tonality behind the scenes to become more emotionally intelligent. At this point, the bot transcends from being a machine to becoming human.

So here's how an emotionally intelligent bot reacts. First, it feels sorry for the loss - it demonstrates empathy. It not only shows specific sympathy card options that I might like based on already profiled data, but it also shows other products like a flower ensemble, a wreath, or a black scarf for the occasion. With facial pattern and speech recognition, this result might be easier to achieve, but with pure text based interaction - we still have time.

Personification of AI – Machine Looks like a Human

You’ve all heard of Sophia the robot from Hanson Robotics that uses voice recognition and other under-the-hood AI tech to get smarter over time. This is a classic example of personification of AI with unique responses to every situation. A robot that looks and acts like a human. There are also inventions like Pepper, Atlas, Aibo, ASIMO, and others, which are human-crafted fiction characters guiding the future of where AI is heading – towards humanized AI. So yes, we’ve made a transition from the land of Hollywood science-fiction – think Wall. E, R2D2, Marvin, and Robocop to real robots like Sophia.

Source: CNBC.com

Ethicize AI – Machine Behaves like a Human 

Several debates on ethics and AI are making rounds in the various AI evangelism circles. You might have heard of things like AI for good, AI for better, AI for all. In a nutshell - to ethicize AI is to design systems that go beyond just first principles. They entail creating inclusive and unbiased models across its applications in industries and bringing a world order where human intelligence can be used for bigger and better things.

There are several stances to ethics and AI. One of the big ones is to make the data science and AI teams diverse so that the data is inclusive and well represented. And when this data is used in the fields of health care, supply chain, government reforms, and such, we will see well-represented and unbiased data patterns to make intelligent interpretations to benefit everyone equally.

The second one is around understanding the ‘cultural and social context’ when designing an AI system. A classic example is autonomous vehicles. When deploying, the AI needs to go beyond 'first principles' of driving and understand how the traffic ecosystem and ethical driving choices come together in a specific country. Driving in the United States is very different from driving in India or Mexico, for example, and the self-driving cars should have that cultural intelligence built-in.

Source: Shutterstock

There are a few other AI and ethics debates, but the last one I am going to talk about here is the argument on automation making humans obsolete, taking over their jobs, diminishing their value, etc. Corporations engaged in creating this new world order also share a moral responsibility to drive home the meta point – AI is here to allow you to grow, and these new AI inventions are a step to leap ahead and not fall behind. A few years ago, I iterated how CI or creative intelligence is very essential and will be a force that will complement AI. So, yes, robots and AI are here for us to reach our potential and do what we’re capable of, not just push papers.

Structured and Unstructured Data – The Heart line of this Movement

To conclude, we know that for an AI model to thrive, it needs heaps of data – both structured and unstructured. And to become more human, it needs to access, analyze, and model more unstructured data sets. Unstructured data patterns are the ones that reside in emails, in corporate communication channels like Slack, in blog posts, on social media sites, in photo and video files, etc. This data gets classified and encoded in a data lake setting and aids in making probabilistic assumptions around the underlying sentiment.

So, acquisition and nurturing of this data is a critical gating factor for a brave new world of humanizing AI be it in the context of conversational AI, invention of more human-like robots, intelligent automation of tasks, deploying self-driving cars across the planet, or building world-class diversified data science teams.

I am going to leave you here with a short talk from the CEO of Affectiva, an MIT spin-off company. Their mission is to humanize how we interact with technology. They bring a concept called artificial emotional intelligence. It predicts the emotional states via AI to solve some deeper problems like autism, depression, health, and education.

Source: YouTube