Search powered by Funnelback
Funnelback’s Managing Director Brett Matson muses on his own interest in machine learning, robots, and how AI will impact search.
What is the foundation of your interest in AI?
It comes from the new class of solvable problems that were previously unsolvable, in all aspects of life. AI is about to change everything, but there are conflicting views on how this will happen. I needed to understand it myself, so I completed Stanford’s Coursera Machine Learning course, which covers the fundamentals like neural networks, recommenders and anomaly detection systems.
What do you say to the belief that AI will lead to human-like experiences in the near term?
Current algorithms don’t support anything close to artificial general intelligence, but they are getting better at mimicking it in some domains. I’m excited about the research happening in reinforcement learning, where algorithms continue to learn the more they’re used. However, particularly within industry, it’s important to have a balanced view, not an elongated view. I risk sounding short-sighted, but AI currently has limitations that we need to be mindful of while planning for the short-to-medium term. In the longer term, we might have more meaningful human-like experiences as well as having to deal with potential threats from AI.
So what IS the real threat with AI?
Experts perceive a two-tier threat: a short-to-medium term economic threat, and a longer term existential threat.
The economic threat is well understood. There are going to be far fewer jobs soon, particularly in the transport and service industries.
The existential threat is less well defined. The theory is, as soon as researchers develop AGI (artificial general intelligence), the AI will attempt to gain access to as much computing power as possible and train itself to be exponentially smarter than a human. Think of AlphaGo, which learned to play Go better than any human. If the AI decides that humans are a threat to its existence or goals, it might do whatever it can to keep humans subjugated. Think of a farmer killing a fox; the farmer isn’t evil, she just needs the fox gone from her property.
In the less sci-fi kind of scenario, there will always be that cry of ‘what about our jobs?’. What do you say about that?
That’s the short-term economic threat of the next 10 years. Throughout human history, the economy has adapted well to improvements in technology many times, but what’s most concerning is the rate of change we’re heading into. Maurice Conti describes it in terms of an exponentially increasing rate of change in the way humans work. He points out that the hunter-gatherer age lasted several million of years, the agricultural age lasted several thousand years, the industrial age lasted several centuries, and the information age has lasted several decades. We’re now entering a new age where the rate of change is going to be an order of magnitude faster than we experienced in the information age. This extreme pace will limit our ability to adjust in the way civilisation has in the past, which will likely lead to widespread unemployment. Combine that with a large global population, fragile economies, and limited natural resources, and we have big problems to solve.
At least we have self-driving cars and the next iPhone to look forward to, but I’m hopeful that new technologies will help solve the bigger problems too.
In that case, how can companies continue to invest in innovative and interesting technology, keeping up with the trend, so to speak, but still contribute to a society that has the majority of its population gainfully employed?
It’s well known that companies that successfully adopt AI will thrive beyond those that don’t. But it’s less well understood how companies can use efficiencies gained in automation to reallocate their human workforce to jobs for greater competitive advantage. i.e. to creative, strategic, or human interaction roles.
For example, over the past few decades there has been a movement within the banking sector to replace human effort with machines. ATMs, apps, and online banking are all part of that movement. This century we’ll see greater automation behind the scenes, which might make it feasible to gain further competitive advantage by replacing customer-facing automated systems with humans. That is, we might see companies using the efficiencies gained by AI to fund the redeployment of humans back into the roles they used to do in order to compete in an increasingly competitive environment.
The value in a two-tier workforce will be in what the humans are doing. It’s about customer experience, and human contact is a part of customer experience, which is why there’ll likely continue to be a role for humans in bartending.
Not to mention the small side point of all the humans that will have their current skills redirected into the maintenance and innovation of the software AND hardware of these robots!
That’s right. And there’s also the related issue of shareholding. In the case of an extremely AI-affected economy, those who don’t own the robots will likely struggle to generate income. It’s beyond my expertise but there’s a philosophical and economic debate about how this may affect the redistribution of wealth.
So, the million dollar question (for us anyway) is: Where is this taking search?
The dream of the semantic web and knowledge graphs has been around for decades. Where it has failed is the amount of human effort required to make it happen. Machine learning has already significantly improved how search technology helps us find information. There are still many areas of improvement, but one is using natural language processing to understand the intent of the user, the meaning of digital information, and how to bring the two together. We still have a long way to go.
