We share the view that excessive attention to the “Singularity”, super intelligence and humanoid robots is often more sensationalist than useful. But we’d shouldn’t make the opposite mistake of thinking that modern AI is a toy that will not impact our lives before 2050.
Our position is that artificial intelligence will significantly transform our lives between 2020 and 2030. Today we see the early signs: AIs surpass human performance on some tasks of the professional elite in the medical and legal fields.
Deep neural networks are a truly disruptive technology. They make it possible to automate tasks that are totally inaccessible to traditional software approaches. Deep neural networks also enable us to interact with computers in ways that are rapidly approaching human interactions (albeit in limited areas).
This new way of programming – instead of describing in detail how the system should work – we set an outcome or objective and the system alone seeks a way to achieve it – is, no exaggeration, a revolution. The openness of research, including that of large companies such as Google and Facebook, is unprecedented. The speed with which new research papers become usable open source code – often a matter of a few days – is gobsmacking. The shared rate of progress of the machine learning community since 2015 is an exceptional moment in the history of computing.
Many start-ups are working on innovative products that incorporate AI. These products will come on the market in the coming years. Even without fundamental new discoveries in research there is already enough to feed ten years of product innovation with the AI technology that exists today.
This wave of new AI products will be part of the current theme: digitalization. A McKinsey study published in 2017 claims that 45% of human professional tasks can be automated with today’s technologies. While AI plays a role in this 45%, most of it is “traditional” automation. But artificial intelligence often brings the missing link for total automation.
This leap forward will not necessarily be made by today’s dominant players. They will be hindered at various levels: by the resistance of managers to see their staff reduced, by the regulatory framework and even by the high quality of their current service. Indeed, it may be easier for a start-up to offer a rudimentary automated service, an EasyJet of the domain if you will.
Companies that automate early can enter a virtuous circle that quickly leads them to market dominance as they will capture data earlier than the competition, allowing them to improve their services while keeping prices low. This will bring them more customers and therefore more data. And so on and so forth.
Soon their services will be able to compete with — even surpass — human services in quality and at an incredible price.
We believe that beyond further technological progress, this type of mechanism will produce a rapid transformation of our economic fabric and that companies and individuals must prepare for it now.
Timothy O’Hear, President
Laura Tocmacov, Director