Vice President, AI and DX European Research Practices
Read full bio
It seems like everywhere I turn online, I see commentary and punditry about AI. And most of the time I encounter it, it’s accompanied by a picture of a white robot touching a screen, shaking hands with a human, sitting in the pose of Rodin’s ‘Thinker’ or looking vaguely menacing. Quite possibly, the robot is doing multiple of these things.
Of course stock photography isn’t something I should spend much time getting angry at – but I can’t help myself. Why? Because it’s a reflection of, and (to a small but insidious extent) a shaper of our discourse about AI.
Sophia the robot – Hanson Robotics’ global sensation of stage and screen of 2017 and 2018 – is this kind of stock photography made real (I would say “made flesh”, but… well, you know). Over the past year or so Sophia has been interviewed by CNBC, Good Morning Britain, Business Insider, and even Elle, Cosmopolitan and Stylist Magazine. It spoke at a UBS technology conference in Hong Kong. It was also made – in a stupefying move – a citizen of Saudi Arabia.
We’re getting massively ahead of ourselves with this stuff. The truth is that, as Benedict Evans, a Partner at renowned VC firm
All the time we focus our energy on worrying about (or glorifying) humanoid robots, or driverless cars, or autonomous weapons or drones, we’re avoiding the discussions we need to be having right now about today’s real-world AI opportunities and challenges in a business context.
And there’s lots to talk about.
Leading-edge vs. here-and-now
Ever since the idea of Artificial Intelligence became part of discourse in computing circles in the 1950s – through multiple computing eras and economic cycles – the focus of ‘AI’ conversation has been on the leading edge of what’s possible. Back in the 1960s and 70s all the work was on rules-based inference engines; since the 1980s and 1990s, as computing power and storage capabilities have continued to grow exponentially, there’s been more emphasis on prediction through statistical analysis of large data sets, as well as making sense of unstructured data, like speech and images.
As the economics of computing have continued to shift the focus has moved to more and more real-time analysis of rich multimedia data (such as facial recognition), as well as the generation of multimedia data – speech, written languages, and even artificially generating speech and video that mimic or fake – using techniques like Generative Adversarial Networks (GANs). If you’re not familiar with the concepts of deepfakes or deep voice research then I encourage you to check it out, and then see why debates around the ethical use of AI are so prevalent right now.
The key thing to bear in mind is that AI is not one singular thing – though, looking at all the coverage of “AIs” like Siri, Cortana and Alexa might make you think so. There aren’t “artificial intelligences” you can just plug into a business.
Instead, AI today is a whole collection of tools and technologies – which can be combined, extended and packaged into different kinds of applications. When we look at what’s being explored and implemented right now in enterprises, it’s the tools and technologies that represent past innovations that are being applied effectively today – it’s not the leading edge stuff. But the important thing to realise is – that’s OK, and it’s actually a good thing. Because this stuff works and we know how to make it add value to our businesses.
Right now, we see AI technologies and tools being explored and implemented in enterprises within three kinds of applications:
- Chatbots – leveraging some combination of natural language processing (NLP), sentiment analysis, and machine translation (as well as other non-AI elements).
- Intelligent automation – leveraging image classification and feature extraction to make sense of inbound documents and extract data from them, and combine these with RPA technologies ‘downstream’ to drive appropriate actions against existing IT systems.
- Recommendation engines – leveraging predictive analytics of various kinds to classify and segment data sets (perhaps concerning customers, their histories, products and services, events and so on) to make prediction-based recommendations to human workers in the flow of decision-making.
IDC’s research indicates that 34% of European companies have plans to adopt AI technologies in some form in 2019. What’s more, over 70% of those already implementing AI are expecting to achieve business benefits within 24 months, and to a large extent those expectations are being realised. By 2022, we expect around $11 billion to be spent on AI technologies and applications across Europe.
None of this requires white robots, electric muppets or autonomous killer drones.
Let’s keep it real, and focus on how to define and implement real-world AI strategies that can deliver business value today.