Is the AI-powered human the next stage of evolution?

Artificial intelligence will complement, not replace, humans - but we have to help each other

I've long contemplated the relationship we humans have with AI, and how it will evolve as machine intelligence continues its breathless progress. Will technology steal a march, leaving humanity in the dust, or will we merge further with these tools to become an entirely new species?

We're already more AI-powered than you think

We've actually been heading in this direction for a while now, increasingly offloading what we used to do with just our noggin not that long ago. How many of us remember phone numbers anymore? How many times do you interrupt a conversation to quickly Google that article you were talking about? Is it really important to remember the facts we'd previously have had educational certifications based upon, as long as you can quickly search for them? How long ago would "being able to summon a private car to where I'm standing in under five minutes" been considered magic? Now we complain when our private drivers take a minute longer than expected on the way to pick us up.

All around us, we've witnessed paradigm shifts that have completely transformed our day-to-day existence, and might once have been referred to ‘artificial intelligence'.

Now? It's just the way things are.

Replacement vs. augmentation

I often get asked if AI will replace humans in the near future, and my answer is always that I don't believe it will. While there are examples of certain processes and professions being absorbed by AI, this is the nature of all technological advancements, and we've seen those changes over the course of recorded human history. In terms of humanity being eclipsed and overcome by a rogue (always anthropomorphised) superintelligence? I'm skeptical, because everything points to a closer and closer merging of our selves with the tools we are creating every day.

One of the challenges with this merge is the very limited bandwidth between us and our devices. We use our fingers, speech and some of our vision for the most part. But when you think about it, humans can talk at a rate of roughly 120 words per minute - but our brains can actually process the information stored in four 4K movies every second. There's a huge discrepancy between what the brain produces and our ability to interact with AI.

While human-generated data increases the accuracy of most AI systems today, bridging the bandwidth gap will require an understanding of the human brain that we simply do not have today. We just don't get it. Neural models are inspired by the neural connections in our brains, but they come nowhere near the computation we perform so effortlessly at the tail end of this particular branch of animal evolution. Transfer learning seeks to apply advancements in one branch of machine learning to others, but we are only just beginning to piece these models together, and at a far more simplified level.

There are still so many severe limitations and we are many, many fundamental breakthroughs away from a world where machines threaten to replace our role as humans.

The beginnings of the AI-powered human

The history of AI is one of incredible surges of progress, followed by a natural plateauing where we figure out how to apply new developments to real-world problems, disrupting and creating new use cases and business models in the process.

For all the excitement behind recent advances, I think where we are right now is pretty primitive. We are only just starting to see the beginnings of the true AI-powered human, particularly if you take the business of translation as an example.

If you train a machine translation engine on news text, as Google, Microsoft and others do, it will perform quite well at translating news articles in general, but if you feed it another content type, like a chat conversation about a last minute flight booking, or an email exchange over a lost purchase, then it will more than likely perform quite poorly, and reflect poorly on those who have opted for it as a communication layer. If you feed it something like video subtitles, the effect can be disastrous.

Traditionally, translation is a craft. When you tell a human translator, "Please translate this," you're actually triggering a set of tasks in their brain that, to be honest, are not very well defined. What are the cognitive processes taking place here? It's not just pattern recognition, and a memory for parallel words between languages. There's a whole creative process happening, and we are far, far away from fully understanding it.

But we can start to chip away at the problem by sharing it between humans and machines alike. We can reduce the amount of human effort and automate all the rest. We can offload the mundane, predictable and repeatable tasks to machines, and free up the humans to work on what requires nuance, cultural understanding and human ambiguity.

The future of the AI-powered human

What we should be trying to do is to offload a lot of the tasks into a machine, the AI component, in a way that augments humans. But it essentially works both ways; machines helping humans, but also humans helping the machine, moving from a role of being responsible for delivering the quality in production, to being the data generation engines that help to train a machine on a continuous basis.

Where I believe we're headed in the future is the continued merging of AI with humans and the combination of the two leading to the next evolution of the human being. If you look at macro trends and how the increasing rate of change is driving a lot of insecurity across the globe, technology is progressing too quickly for humans to keep up. With technologies developing at an exponentially faster level, people are outdated by the time they come out of college.

But even here, AI will also be an enabling tool which might just help us develop fast enough to keep up with the rate of change. If we tackle issues of bandwidth and get smart with individual application, the age of the AI-powered human will truly be upon us.

Vasco Pedro is the CEO of Unbabel