The second Industrial Revolution of 2010-2060?
It seems that lately we are hearing a lot about artificial intelligence, big data, and machine learning. You might work in an industry that’s undergoing rapid change thanks to data, or you might notice how data-driven companies are starting to transform the services that you use, from ordering a takeaway, to planning a holiday, to buying a house.
It’s easy to understand change that comes in small steps. Cars have got safer and more efficient since the 1960s as the technology improved. We can build bigger and stronger bridges than before, deeper and wider tunnels. But for the first time in more than a century we are undergoing a revolution akin to the Industrial Revolution of the 1800s. Entire industries are appearing and disappearing, and the old ways of doing things can disappear in the space of a few years. Some are calling our era the second Industrial Revolution, or the Data Revolution.
Who can use AI?
You might be asking yourself, how can I benefit from AI? Like the manufacturing technology of the original Industrial Revolution, an individual person can’t normally use AI for themselves directly. Even a small business may struggle to find a use for the technology. However a corporation or government can begin to benefit from AI when they have data on a million citizens, customers or employees. In other words, AI often benefits the powerful, those already in possession of data.
Technology startup companies often hit a problem when trying to develop AI: to make a face recogniser, you need millions of face images. Yet without a user base it’s hard to get these images. And without a good product it’s hard to attract a user base. This is known as the Cold Start Problem, and it’s one reason why Facebook, Google, Microsoft and Amazon have some of the best face recognition models, map routing software, machine translators and product recommendation systems.
How industries are using AI
AI allows an industry to take all the information that has accumulated over the past, find patterns in it, learn from it, and apply the knowledge to make predictions in the future. This in itself is nothing new but with computers we can look at more data than a human could look through in a lifetime. A single AI can do the task of a hundred humans, or even to make more accurate predictions than any human could possibly do – just as a single steam engine was able to do the work of many manual workers back in the original Industrial Revolution.
Some of the well known examples include the automation of call centres with natural language understanding, or diagnosis of diabetic retinopathy by a computer vision system which has learnt from thousands of retina images. There have been landmark events that hit the headlines over the last few years, such as IBM Watson winning on Jeopardy, or AlphaGo beating Lee Sedol at Go. The new data driven companies such as Uber and Airbnb exemplify how a company designed around data can achieve meteoric success with a simple business model: Airbnb is constantly collecting data on your behaviour and using it to improve their future recommendations.
Across the board AI has allowed industries to find data that they have collected over the years, and squeeze the value out of it. There have been challenges, however. Often regulation stops AI from being applied or data from being collected. Traditional companies dominating a conservative industry may feel safe from the AI revolution and not feel the pressure to adapt – that is, until a nifty startup comes along and beats them to it.
Concerns about AI
There have also been concerns about the impartiality of AI systems. Are they capable of racial or other biases? This was put to the test when Eric Loomis challenged a parole decision made by an AI, which went to Wisconsin’s Supreme Court. Since most machine learning models involve gradually reducing an error rate, called a loss function, while learning from a population, this means that the model with the lowest loss function may well perform badly on minority groups. A recent study by the US National Institute of Standards and Technology found that most face recognition systems performed worse for Asian and African-American faces than for white faces.
Other commentators have expressed concern about the rapid development of AI resulting in an unregulated ‘Wild West’ where complex machine learning models could end up doing harm to humanity. Again, like the Industrial Revolution, some see AI as widening the inequalities in society. The 2002 film Minority Report, starring Tom Cruise, envisages a dystopian future where the police have a “pre-crime” division that punishes criminals before they commit crimes. With predictive analytics could this dystopia become a reality?
On the other hand AI technology such as genomic based machine learning is bringing leaps forward in medicine, and self driving cars are likely to be safer than human drivers, so nobody can deny the benefits that AI is bringing to humanity in some fields.
At the current time, we can see how the first Industrial Revolution played out in the past but we do not know what the world will look like after the AI Revolution. I think most people would agree that our quality of life has improved in the last 200 years, climate change and environmental damage notwithstanding. Will the AI Revolution bring the same kind of benefits?
Some people foresee increasing social inequalities and an inevitable civil unrest. Will societies need to adapt, for example by introducing a universal basic income for those rendered unemployed and unemployable by automation? Perhaps some jobs such as nursing, which require empathy, will never be automated. On the other hand the full automation of driving jobs seems inevitable over the next few decades.
I’m interested to hear also how AI has affected your livelihood or industry. Let me know in the comments below.