Machine learning algorithms are “psychopaths” without regard for human consequences, according to Timo Elliott, SAP VP and global innovation evangelist, who cautions organisations should be piloting ML programs in areas with less potential human consequences.
“[Machine learning] can do some amazingly good things but it requires human judgement,” Elliott told Which-50 following his presentation at the Gartner Data and Analytics Summit in Sydney.
He explained machine learning is creating great value for organisations but must be paired with human oversight.
“Algorithms are psychopaths. They don’t really know what they are doing or how that affects human beings. So it’s really important as organisations start implementing machine learning that they start in areas that don’t touch on human beings too much.”
A good early machine learning initiative, according to Elliott, would be something like finance or logistics — “areas where you’re just dealing about facts and you don’t have to worry about judgement”.
Starting in areas where potential consequences are high is fraught and requires extensive ethics considerations, he said.
“As soon as you start touching on human resources or customer interfaces then you really have to think long and hard about the ethical considerations. We believe that companies will be hiring ethicists when it comes to their machine learning technology in the future.”
Failing to consider the power and necessary oversight required is one of the biggest around machine learning pilots, Elliott said.
Any effective use of machine learning requires strong data foundations, Elliott says, another requirement that many organisations are struggling with.
“Machine learning algorithms are not that complicated. What is complicated is having a strong, large quantity of high quality data. If you don’t have that it’s much harder to get started.”
However, the data is often already available to organisations, Elliott says, and business applications can provide the best starting point. Because business applications like ERP provide vast amounts of high quality, often already refined data, he said.
About this author
Joseph Brookes is a writer for the Which-50 Digital Intelligence Unit, of which SAP is a corporate member. Our members provide their insights and expertise for the benefit of the Which-50 community. Membership fees apply.