What We Can Learn from AI Learning to Read Like Humans

Over the last few years, advances in technologies such as AI and machine learning have taught us to think bigger and try to visualize the art of the possible. Machines can already beat the best human contestants at chess, Jeopardy!, and even a 2,500-year-old complex board game called Go. But, what skill could the machines learn next to defeat their human masters once again?

An AI algorithm, developed by Chinese e-commerce giant Alibaba and Microsoft, recently outperformed humans on a reading comprehension test for the first time. The Stanford Question Answering Dataset (SQuAD) test involved asking both human and robot contestants to provide answers more over 100,000 questions drawn from over 500 Wikipedia articles.

However, the big question was, “Has artificial intelligence advanced enough to process and digest vast amounts of data and then provide correct answers to questions?” The results revealed that the AI software just snatched a victory with 82.44%t correct responses compared to the humans’ 82.304%.

Although it's a little premature to claim that this is the beginning of a dystopian future and the rise of the machines, it's important to document that this is the first time that natural language processing (NLP) has outperformed humans. It's also worth highlighting that tech behemoths such as Google, IBM, and Facebook have yet to produce results with reading comprehension.

Disappointingly, many are focusing on scary headlines based on the fears of watching too many science fiction movies and lazily reporting worst-case scenarios as fact. We need to start thinking about how this technology can be put to use in the real world and transform the customer experience.

Luo Si, chief scientist for natural language processing at Alibaba’s Institute of Data Science of Technologies, told Bloomberg, "The technology underneath can be gradually applied to numerous applications such as customer service, museum tutorials and online responses to medical inquiries from patients, decreasing the need for human input in an unprecedented way.”

However, we must not run before we can walk, and we must understand that this technology is still very much in its infancy. According to AI experts who reviewed the test results, it was easy to confuse the system by adding gibberish into the passages. The most significant drawback in this latest example is that although it excels at reading comprehension, it doesn't yet genuinely comprehend language.

We have all met a grade A student who is gifted with a photographic memory but doesn't understand the subject matter or how to apply the skills in a real-life scenario. There is much more to genuinely understanding a topic and the many nuances of language than answering a question using data.

A lack of intonation in emails and online messages often results in confusion or misunderstandings. Other challenges are around regional accents, muttering, or even highly illogical arguments. As humans, we are continually searching for hidden meanings or reading between the lines in conversations.

Whether it's honing our reasoning and debating skills or trying to understand what somebody really meant after making a flippant comment, language is complicated. To progress, we need to remove fear and look beyond the scaremongering headlines that attempt to pour scorn on any technological accomplishments.

In a digital age of soundbites, many will see AI as a threat that is destined to take away millions of jobs. However, this latest story is an excellent example of how far technology has progressed, but also how far it still has to go. Rather than swimming against the tide, maybe we should be looking at how we can play a part in a future led by rising consumer expectations rather than corporations.

As old roles are retired, technology is also creating jobs and even entire departments within companies. There is also a demand for technical skills across multiple sectors that are in desperately short supply. There is an argument that technology will stimulate job growth by creating new positions for a digital era. Maybe we should think beyond headlines of “a robot is going to take my job” mentality and start thinking how we can be a part of the future.

For example, at IPsoft, led by it's mathematician CEO and AI guru Chetan Dube, the company is attempting to change the perception of AI utilizing cognitive and autonomous technologies to help redefine the customer experience. Amelia, the company's flagship platform (watch her in action here at the Gartner annual keynote), is the world's most human and fully conversational AI, entirely automating human-to-human interactions and process execution.

It was crucial that Amelia communicated using natural language that can respond to customers’ emotional states, just like a real person would, thus enabling a depth of conversations that reaches far beyond what Siri, Alexa, Cortana, or Watson can do. But what IPsoft is most proud of is putting people at the heart of these technology-based solutions.

It's not about pitching humans vs. robots, but how we can work together to build a better future. Maybe this is a story that does not get as many clicks or sell as many newspapers as a scary one does. But if helping connect conversations to data and processes to give every customer the best end-to-end experience possible isn’t a mouth-watering prospect, what is?