Businesses pursuing artificial intelligence and deep learning projects are ultimately responsible for the actions of their algorithms, the ACCC has warned.
Speaking at an event in Sydney yesterday, chairman Rod Sims predicted the time will come when pricing algorithms will be sophisticated enough to control markets - and even “collude” with other algorithms.
However, because algorithms are less prone to the “flights of fancy” of humans that may engage in collusive behaviour, instances of “e-collusion” could be much harder to detect.
Theoretically, machine learning algorithms could be smart enough to stay one step ahead of authorities.
“To further complicate matters, the development of deep learning and artificial intelligence may mean that companies will not necessarily know how, or why, a machine came to a particular conclusion,” Sims said.
“To this end, it is argued that if similar algorithms are deployed by competing companies, an anti-competitive equilibrium may be achieved without contravening competition laws.”
Sims said, however, that programmers would not be able to blame their algorithms.
“You cannot avoid liability by saying ‘my robot did it’,” he said.
Sims said the ACCC had set up an analytics unit to deal with what it believes will be an emerging avenue of problems.
It is also working with other regulators and “practitioners” to get across the full range of problems that advanced analytics may present.
Sims said the ACCC would, for example, examine merger and acquisition proposals more closely to identify whether the combination of the two parties’ big data holdings would pose any market issues.
The ACCC said it has not yet identified a case of collusion between algorithms, though it has previously taken online platform operators to court - such as in the travel booking industry - over data-related issues.
Sims was confident that changes to competition laws brought in as a result of the Harper Review would allow the ACCC to successfully prosecute algorithmic offenders.
“At this stage, the ACCC has not seen any anti-competitive algorithms which require an enforcement response beyond what is now available to the ACCC under Australian law,” he said.
