Artificial intelligence has been big news over the last year or so. While the idea of the technology is not new – Arthur C Clarke’s HAL 9000 brought it to the fore back in 1968 when 2001: A Space Odyssey was released – we have reached a tipping point as the amount of data we have access to has exploded and the cloud has made vast amounts of computing power accessible. But we are only now starting to think about the real impacts of AI on business and society.
Last month, I attended a panel discussion run by the ACS on AI. While most of what was discussed was not new, I did like the model one speaker suggested for thinking about what AI really is.
ACS boss Anthony Wong suggested AI comes in four different forms.
- The most basic form of AI is process robotics where systems mimic human action.
- Intelligent automation is where systems mimic or augment human judgement.
- Cognitive automation systems that augment human intelligence
- Artificial general intelligence (AGI) is the pinnacle. These are systems that can mimic human intelligence.
Robots and automation
Automation has been around for decades. Machines have been doing the work of humans and displacing them from the workforce since the industrial revolution, when steam-driven machines were used to remove water from mines. We’ve seen industrial robots in factories for decades. Other than a small number of exceptions, cars have been built on highly automated production lines for decades.
Whenever automation has been introduced into a process, it has resulted in people being displaced from their current work. I’m being deliberately careful in my words, using the word “displaced”. While many people have lost their jobs through automation, the number of people working continues to grow. But the human impact of automation is widespread – something that Donald Trump exploited during the US election when he promised to bring back manufacturing jobs. What many of his constituents failed to understand is that, while many of those jobs went overseas to countries were there were large and inexpensive workforces, the manufacturing jobs that people used to do can often be done faster and better by a machine.
I used the word displacement because we are seeing jobs change. Over time, we are seeing more and more repetitive tasks handed away from the hands of people to machines. And while simple “mechanical” tasks like applying a spot weld in the same place over and over, or attaching a door to a car have now been given to machines, we are seeing more complex tasks making start to move.
Where AI is today
Our world has always produced more data than our brains can handle. While the way this happens with computers is a relatively new phenomenon our brains have been doing it for millennia. We have always filtered out the “noise” in order to simply survive.
The world of big data – defined by volume, velocity and variety – means that we no longer have the capability to process the volume of data we have available to us. That means we need systems that can understand what is important and what is noise to help us make better decisions.
This is where the middle two categories of AI, described by Wong, come into play and where most of the current focus on AI application is focussed. We are looking for systems that help people.
We see this in security applications that collate log data and find patterns that would otherwise be invisible to us. That processed view of the data is then passed to humans who can use their skills to react to the data.
In human terms, the people who previously would have manually trolled through the logs, looking for patterns have been displaced by processors and algorithms. But, if provided with the opportunity, they could be engaged in higher order analysis skills, problem solving and issue resolution. Or, they could help define and build the algorithms used by the AI systems. As the algorithms “learn” more about what is important, they “adapt” and do better analysis of what matters most.
Isn’t that machine learning?
This is where things get a little hazy.
Human intelligence is about adaptability. From an evolutionary sense, being able to adapt to the environment is what allows us to survive, procreate and further spread our genes. So, we react to what’s going on and use the results of those reactions to inform our future actions.
We are now creating computer systems that can do the same. Except that the connectedness of systems means one system, like an autonomous car, can learn from the experience of others in near real-time.
Many systems today, whether you’re talking about cars, drones, security systems or industrial robots use machine learning to enhance the algorithms that drive their operation to, at least, mimic elements of what humans consider to be intelligence.
Skynet is not here
The fourth category Wong presented in his maturity model for AI was computer systems that can mimic human intelligence.
The fundamental difference between human intelligence and AI is adaptability. AI and machine learning systems are constrained by the limitations of their programmers. If we take an AI-based threat management system, it is limited by the data sources it is programmed to use. Those sources were selected by developers who decided, based on their knowledge, intuition and biases which data sources would be the best for the system to use and make its decisions with.
But what if there’s a new information source that you hear about? How does the system know about it?
Autonomous vehicle systems can do a pretty good job of predicting when a car is going to crash based on what it knows about vehicle movement. But I think back to the scene at the end of Back to the Future 3, when Marty is challenged to a drag race. While the loud revving of engines might have given a computer a clue that some dangerous driving was potentially about to happen, the real cues came from the words and facial expressions of the driver and his passenger – something computer systems are still not great at understanding.
In other words, getting systems that can mimic human actions, at a level that is more consistently accurate than people is easy. But faced with unexpected or new situations, today’s AI struggles.
During one of the many AI vs human Go matches, the computer was handed a rare defeat when the human player made an unexpected move and Google’s AlphaGo defeated a champion player, the system was defeated in one game.
The system was only as good as the programmers who created it. In that case, they programmed the system to learn what do in almost every situation when playing a champion. It did not anticipate an error.
Tech always displaces old jobs, but new ones are created
When the motor car became ubiquitous, thousands of horse-shoe makers found themselves out of work. But bad news for blacksmiths was good news for the mechanically minded.
Many of the jobs of last century have disappeared but they have been replaced. We expect autonomous cars and trucks to take over from truck and taxi drivers. And that will be hard for people in those industries. But new jobs in creating the software and hardware that make and operate those vehicles are being created. And those machines will need to maintained. Perhaps not by traditional motor mechanics but by electrical and software engineers.
Services will change. Chatbots will deal with mundane tasks with humans freed to work on higher value tasks. For example, in banks, counter staff carry out far fewer over-the-counter transactions. Instead, they have become advisors in everything from creating an investment portfolio to important life events like buying a house.
People will always be important. We are still decades from creating systems that can do everything people can do. But people will to need to adapt. Autonomous systems will impact our world and how we interact. Our cities will change – why build parking garages when autonomous car sharing services will deliver a right-sized vehicle to us when we need it? And then send it to the next person instead of leaving it idle in a car park for hours at a time.
There’s little doubt AI will change the world. I’m an optimist and don’t subscribe to Elon Musk’s view that AI could spark a global war. I suspect he went to a Matthew Broderick film festival and got a little caught up in the moment.
Will AI disrupt how we do things? It already has disrupted the world. But I think there are great opportunities it offers for us professionally and socially.