2016 has been a big year for Artificial Intelligence. Taking over from the Internet of Things as “Most Talked About Tech of The Year”, this represents a set of technologies that will clearly have significant long-term implications for how we work, buy products and consume services. This explosion is made possible by the convergence of three trends: The massive growth in data availability in every field of human endeavour, a renewed interest in the algorithms underpinning artificial intelligence, and the plummeting cost of computing. There are few fields that will not be profoundly impacted, so it is not realistic to cover all major advances in a single blog post. Instead, here I touch on three very different aspects of AI that grabbed some attention in 2016.
In one of the many articles on trends for 2017, Fjord, a digital consultancy company, predicts that the next step in AI development will be on emotional intelligence, creating more engaging and effective interfaces to systems powered by algorithms. We have already seen how Chatbots are becoming an increasingly common means of how companies engage with their customers. From fast food companies to banks, taking in dating companies and toys, chatbots are becoming hard to avoid. Whatever the application, the more human-like and empathetic the bot, the better the experience will be.
One company that addresses this challenge is IBM in its Watson Blue Mix cloud service, which already offers a Tone Analyser service to detect emotions such as anger and fear. It can also detect personality traits such as openness and conscientiousness. This can be used by an automated chatbot to react to the emotional response of its customers. Similarly, Koko, an MIT Media Lab spin-off company that provides chat bots for supporting people who feel stress or need advice. Acting in an augmented-reality mode, 90% of chat activity is moderated by Koko’s AI systems, which is trained to identify people with malicious behaviour as well as focus the response to the type of problem being faced. Another MIT Media Lab spin-off in a similar space is Cogito, which provides real-time behavioural feedback to call centre agents as to the emotional state of the customer. This is based on traits such as volume, speed and pauses and the system helps guide the conversation by providing real-time guidance to smooth and improve the rapport between caller and agent. What all these companies have in common is to create interaction methods that both capture and harness the emotional state of the user in order to improve the relevance and effectiveness of the service being offered.
Making Money Automatically – AI based trading and investing
The finance and investment industries are waking up to the potential of learning algorithms to improve returns. The world’s largest hedge fund, Bridewater Associates, is building an AI-based algorithm that can implement the philosophy and investment principles of its founder Ray Dalio. The aim is to automate investment decisions, with humans no longer being responsible for individual decisions or transactions. Instead humans will curate and tailor the criteria used by the ‘machine’ and intervene should something go wrong. As the finance industry is already highly data-driven, it is seen as fertile ground for AI, particularly in detaching the emotions from investment decisions.
Of course using computers to make trading and investment decisions is nothing new, with algorithmic and high-frequency trading being responsible for around 50% of transactions in the North American and European equity markets. What is changing is the reach and accessibility of automated trading platforms as well as the sophistication and ambition of investment companies. For example, there are a number of platforms which effectively provide croudsourcing of trading algorithms, connecting people who create the algorithms to capital from investors. The main advantage of investment engines built on AI or machine learning over the previous generation of quantative algorithms or ‘quants’ is that they AI systems are adaptive and can learn. The hope is that such systems can automatically recognise changes and adapt in ways before quants can. The aim is not to make a trade before someone else does, but simply to make the best long-term bet.
This technology is rapidly being adopted by hedge funds across the world. Bloomberg reports on a Japanese hedge fund manager whose self-learning algorithm anticipated the impact of the Brexit vote on Japanese share prices, resulting in the best single-day gains that year. This company is not alone, as reports from Eurekahedge indicate that AI hedge funds are outperforming other funds.
Assessing the Economic Impact of AI
While all the world’s media was focused as to how the Obama administration would react to Russian hacking, in the week before Christmas, the White House released a report on the the impact Artificial Intelligence will have on the economy (see here for full report). The report compares it to the advent of mass production techniques in the 19th century, which although heralding great improvements in productivity, also saw highly skilled artisans and craftsmen replaced by machines and lower-skilled labour. The report makes some fairly non-controversial and, to a certain extent, blindingly obvious recommendations, including investing in AI R&D due to the overall benefits to the economy, while providing the labour force with the skills and education relevant to the new types of jobs that will emerge, and assist workers who will lose their jobs or see their earning power reduce. While Donald Trump has been focusing on the impacts of globalisation on the American labour force, it is clear that workplace automation will have at least as big an impact on American jobs.
Where the report is perhaps a bit more insightful is in its assessments of what new jobs may arise out of the advent of AI. Here are examples proved:
- Engagement. Where AI is used as part of of a larger task, say in medical diagnosis, human professionals will still be required to complete the task and engage with customers or patients. In the healthcare example, healthcare professionals will be required to work with patients throughout the diagnosis and treatment process. This symbiosis of worker and AI is being to be referred to as “Augmented Intelligence”
- Development. Clearly development of AI systems will clearly need large numbers of highly-skilled software developers and engineers of various disciplines. More widely, industry specialists across all impacted domains will be required to create the services, deal with the data input and generated, as well as dealing with the ethics, regulations and social impacts of automated systems.
- Supervision. This includes all roles involved in monitoring, supervising and maintaining AI-capable systems. Even fully-automated systems will require some form of human oversight and maintenance, while learning systems will need moderation and guidance.
- Response to Paradigm Shifts. As AI significantly changes the dynamics of given industries, entirely new categories of companies, jobs and products will emerge as a result. One example is in vehicular transportation. As self-driving vehicles become more commonplace on our roads, this will have implications on how roads and road-side infrastructure is built, the type of vehicles on the roads, the services offered to occupants of cars as well as the type of hire car, taxi, and mass transit services that will become possible.
It is clear that Artificial Intelligence will have a major impact on every industry and most workplaces. Much like steam powered looms kick-started the industrial revolution in the late 18th century, wiping out the garment cottage industries across England in just over a generation, many automatable jobs will disappear over the next couple of decades. What we cannot predict is what new jobs will take their place. For example, who would have foreseen the legions of social and digital marketeers in the job market just ten years ago?