In August 2019, a group of lawyers led by Ryan Abbot at the University of Surrey, filed two patent applications with the US, EU and UK patent agencies. The first application is for a food storage container, the second for a warning light. Both products were invented not by a human – but by an AI machine called Dabus.
The patent agencies reviewing the applications are now considering the same fundamental question – Can machines own intellectual property rights?
The short answer will be no. Yet as the global race to lead in AI gathers speed, policymakers in the US and Europe will be wondering if the law needs to be updated to make sure the IP incentives to create and innovate that have worked in the past remain effective in the future.
It is implicit in intellectual property law both in the US and Europe that in order to own IP rights an inventor in the case of a patent and an author/creator in the case of copyright must be a human being. Machines, however intelligent, are merely tools.
And there is nothing new about using machines as tools in the process of creation. In the 1990s, David Bowie generated lyrics for his Berlin trilogy of albums using a ‘verbasizer’ which randomized sentences to create unexpected word combinations. Visual artist Anna Ridler uses artificial intelligence as a tool to create art – recently generating thousands of ‘imagined’ tulip varieties with unique characteristics now on display at the AI exhibition at the Barbican, London.
In both cases, despite the involvement of a machine, the final output landed squarely within the boundaries of what the human artist intended to create. It did not matter whether the machine could own rights in the work – because there was no doubt that the human behind the machine did.
Yet as machines begin producing output that goes beyond the realms of the human mind, we are entering a different scenario.
In 2016, Google DeepMind’s ‘AlphaGo’ defeated Lee Sedol, reigning world champion at ‘Go’, an ancient Chinese game, with move 37 – a move that no human had ever contemplated before.
In March 2019, Warner Music signed the first record deal with an algorithm called Endell which is set to generate 20 albums this year.
Dabus, the AI machine behind last week’s patent applications, has created products in areas in which its programmer has no background.
In these cases, the machine cannot own rights because it is not a human. And the human behind the machine will fail to meet key criteria to claim IP rights, such as an ‘inventive step’ for a patent and ‘originality’ for copyright, because they did not create the work. As a result, AI generated inventions forge a loop hole in IP law and promptly fall straight through it.
Why should this matter to policymakers? Intellectual property rights exist to incentivize people to innovate and share their innovations. IP rights grant an inventor or creator exclusive rights over their output for a set period of time in exchange for them disseminating it. In this way, IP law strikes a balance between the needs of the individual to earn a living from their work and the needs of society to benefit from that innovation and be able to build upon it.
One could argue that because machines do not need to be incentivized to produce, nor do they need to earn a living, they should not be granted IP rights. But the same is not true for the people and companies behind them.
If there are no intellectual property rights in AI generated inventions, the people developing and investing in AI will only have rights in the underlying AI software or code – and in the data they input into the machine, but not in what it produces. Will this be sufficient if the real value of the AI machine resides in its output rather than in the AI software itself?
Will these individuals or companies be able license or assign rights in the AI machine’s output in order to bring it to market? And will they have sufficient recourse if their invention is copied and commercially exploited by others? Unless they can use other non-IP mechanisms to protect their investment and output – such as contract law, trade secrets or technological solutions, the AI loop hole in IP may act as a disincentive to investment in AI innovation.
This will be of concern to policymakers wishing to encourage the AI innovation not only in products and services but also in tackling humanity’s most intractable problems. They will need to ask themselves whether the current IP framework, coupled with other contractual, technological and trade secrets mechanisms, is sufficient to incentivize AI generated creations – or are new measures necessary to bridge the inventive gap opening up between man and machine?
New European Commission President, Ursula von der Leyen, has promised policy proposals on AI within the first 100 days of her term – and these questions could be among those under consideration. The debate on AI and IP no doubt will be lively and will force us to think about the fundamental purpose of IP. That is an essential discussion for all those whose livelihood depends on being rewarded for the efforts of the innovative and creative brain.
Photo credit: Anna Ridler ‘Myriad (Tulips)’ on display at ‘AI: More than Human’ exhibition at Barbican, London