AI: The history, the hype and its future

null

AI is enjoying a very public love affair. Virtual assistants like Alexa, Siri and Cortana are being used today by millions of people. AR and VR technologies are going from strength to strength in the gaming industry. Pioneered by Amazon, today a host of online retailers use predictive analytics to anticipate purchases you might need based on your online shopping history. Banks use fraud detection systems based on AI. The list goes on (and on)…

In the short to mid-term, expect to see intelligence embedded into systems such as drones, self-driving cars, self-healing networks and supply chains. Today’s love affair with AI is strong, but the relationship hit the rocks some time ago.

A potted history…the first AI winter

Interest in AI spiked during the Cold War when it was considered a vital tool, with many Western governments pouring millions of dollars into AI research in the hope of gaining an edge over the USSR. In the ’50s and ‘60s, there were two types of algorithm. There were algorithms like Perceptron, inspired by neurological structures, and algorithms like Logic Theorist and General Problem Solver, inspired by symbolic logic.

The latter two worked really well on solving discrete, logical problems, like the mathematical puzzle, the Towers of Hanoi, or the theorems presented in Russell and Whitehead’s ‘Principia Mathematica’. However, what the Western agencies really wanted was a tool to solve ambiguous, complex problems, such as automatic language translation. It would take many, many years for Perceptron to be advanced enough. The AI advances of the time failed to live up to the vast amounts of hype. Disappointment set in and funding was slashed. The perception of AI stagnated. AI fell out of public view and progress slowed dramatically.

The storm surrounding AI

The tension between perception and reality is one that is felt acutely in AI circles. There is still a lot of positive hype around AI’s potential. But many negative views abound, too.  The long-standing concern that robots will take jobs away from workers is keenly felt, and yet many argue that AI will change the types of jobs workers do rather than eliminate them altogether, or even that AI will enable better work-life balance, a view most recently endorsed by Bill Gates. AI has been blamed for alleged election rigging, but it has also been presented as the cure for fake news.

Another common public perception of AI is bound up with a fear that robots will take over the world. It may sound like the plot of a sci-fi thriller, but it is a concern famously voiced by Elon Musk, who, despite being unfazed by the prospect of intergalactic space exploration, is seriously worried about AI robots colonising Earth. 

There is never a dull moment in the AI debate! When public opinion, both positive and negative, is so strong, could the mood change suddenly, turning the tide on AI and triggering another AI winter?

Weather-proofing the AI Industry

The truth is that opinions will come and go, but the situation now is qualitatively different than during the '50s and '60s. The knowledge base is now much broader for AI and the technologies it depends on. We now have technological advances such as the backpropagation algorithm, deep architectures for hierarchical learning, GPUs for fast training, big data technologies for collecting and processing large datasets for training efficiently and inexpensively. All of these unleashed a new revolution in Artificial Neural Networks (multi-layer perceptrons) in 2012, and momentum shows no sign of slowing.

Observers have pointed out that the rate of innovation cannot continue ad infinitum. Even if it slows in the near future, it is unlikely that AI will find itself in another AI winter. Here’s why.

Today, most practical applications are driven by business and industry, not government agencies. Here are four reasons why AI might not hit the rocks:

1.      AI is big business – The funding for AI research and development comes from industry leaders such as Google, Facebook, Microsoft, Amazon and Apple instead of governments. The research is conducted actively within businesses in addition to academia.  

2.      It’s open – Most research papers are published on open sites such as arxiv.org and they are easily available to anyone. The codes for the latest algorithms are publicly available thorough code repositories at GitHub and BitBucket. Large datasets are available for training predictive models. There are free AI training courses from key universities who are at the forefront of AI research such as Cornell, Berkeley, MIT and Stanford. More research and new applications are coming from non-Western countries such as China, Japan and India.

3.      We’d miss it – AI is already embedded in many consumer applications and attitudes are broadly positive. For instance, it is making a big difference in the realm of customer service. One recent study suggests that UK consumers are generally positively disposed to chatbots and would happily use them more if it helped to resolve customer queries faster.

4.      Without AI, forget 5G, IoT, Drones, Space Exploration….- AI is the driving force of all these next-gen technologies, and more. For instance, the IoT industry, telecoms and the nascent 5G ecosystem all rely on AI to help deliver on the customer expectations about speed, zero latency, QoE etc. The number of interactions taking place on 5G networks would be rendered unmanageable without some degree of self-driving, self-diagnosing, self-healing systems.

AI taps into the collective imagination like nothing else. Putting visions of sci-fi cyborgs aside (IF they are possible, they are a while off yet) the power of AI is now in the hands of people, industries and academia. There are already many practical applications. With previous AI winters, governments could turn the dial up or down on investment, thereby stoking or stifling the nascent AI community.

Today’s AI ecosystem is much more advanced. It has been a boon to industries such as big data analytics, cloud processing and for smart materials and sensors used in hardware. Participants in this flourishing community are all engaged and committed to advancing its use cases for practical benefits and business gain. Our relationship with AI is here to stay and getting stronger. Like any durable relationship, don’t expect the arguments and debates around it to die down in a hurry.

Abhijit Thatte, Associate Vice President of Technology and Practice Leader for Artificial Intelligence, Aricent
Image Credit: John Williams RUS / Shutterstock