Human and machine, let’s combine skills!
Most people are prepared to pay extra for the presence of a human pilot in the cockpit of their plane, as a recent study proves. Even if the automatic pilot performs as well as (or better than?) its ‘real life’-colleague. What exactly are the benefits of humans interacting with technology? Martijn Vos shares his insights on the opportunities of combining human-machine input: “After a tremendous improvement in technology in the past decades, in 2017 the combination is still extra powerful compared to machine-only.”
“When I started my career 20 years ago at Ortec, one of the first lessons I learned was that humans improve on optimal solutions delivered by technology”, says Vos. “As an example: a tool used for creating a work schedule for a large group of people with all kind of difficult (labor law) restrictions was a tremendous help to the person responsible for creating these schedules. However, after the tool proposed the ‘optimal’ solution, the planner improved further on this, knowing that Mr. X – if asked –was able to start 15 minutes earlier (than the hard restriction) and Mr. Y needed a flexible start as he had a busy commute to the office. Using this flexibility the optimal schedule was further optimized, or at least won in robustness, making the man-machine team the winner.”
After Deep Blue
A few years later Gary Kasparov lost as the human world champion in Chess to Deep Blue, a chess program running on IBM supercomputers (1). “In the aftermath of this historical event a combination of human and computer proved even better than the computer as the human added intuition, time-management et cetera. Fair to say that after another decade the machine-only was as good or better than the human-machine combination (2). In aviation we have seen a combination of the pilot (human) and autopilot (machine) for a long time as well. Also here discussion is reaching the point of questioning whether we should not rely on machine-only, although consumers still prefer the pilot even if the ticket fare is more expensive (based on a recent UBS study(3)).”
In general there has been a tremendous improvement in technology since these examples, but even in 2017 a combination of human and machine is still extra powerful. During the 2016 Singularity Summit in the Netherlands, Carlo van de Weijer showed an example of a self-driving car in Amsterdam traffic. The ‘machine’ was unable to cross a busy road, with human bicyclists, pedestrians and car-drivers not following traffic rules ‘exactly’, while the car was programmed to exactly follow these rules. Technological advancement has progressed so far that there can be other reasons for humans to interact with technology, as was shown early August when Facebook had shut down one of its AI systems after chatbots started speaking in their own language defying the codes provided. The social media giant had to pull the plug on the AI system that its researchers were working on "because things got out of hand" .
Mr. Draghi translated
Ortec Finance applies a combination of technology and human input to most of its technology and applications as well. Vos: “We combine state of the art modelling with human interaction and curation. A pure statistical model has no knowledge of the precise and delicate wording of Mr. Draghi, nor can it deduct the impact of e.g. climate change or structural moves to ESG investing. At the same time a human may find it difficult to translate his ideas on US growth and interest to impact on e.g. emerging market growth and European Equity indices. Technology using statistical tools or artificial intelligence can do this translation in a quite consistent way (4).”
Vos sees another advantage in using the combination over ‘humans only’. In the process of creating forward looking scenarios, every human intervention is recorded. This makes it possible to justify every single step in the investment decision process. This is of tremendous value in a time in which stakeholders increasingly focus on good governance and accountability. Being clear on the division between human and machine input helps in assessing the quality of both. “Developments in this department over time and careful performance attribution may give a general idea as of which moment the input of humans to this process can be further reduced. This could give a further boost to robo-advice and create even better possibilities for institutions and individuals to reach their financial goals.”
- See e.g.: https://www.theguardian.com/business/2017/aug/07/air-passengers-pilotless-planes-fares-ubs
- See: https://link.springer.com/article/10.1057/s41260-017-0042-y