Can a computer 'own' an idea?
With major advances in computing power and availability of large datasets, the use of Artificial Intelligence (AI) has seen a massive boom in almost every sector. Simply put, AI programs are taught an algorithm which they can pass data through to generate an outcome, this outcome can then be the basis of a human decision. At base-level, an AI is nothing more than a programme which can perform the job of a human.
The hyperinflated persona of an all-powerful, all-knowing AI portrayed to the public by popular culture is a long way off; but AI’s are getting smarter and humans are entrusting them with more and more responsibilities.
Programming AI requires the creation of three cognitive centres: ones for learning, reasoning and self-correction. However, most AIs we encounter on a daily basis are known as ‘Weak AIs’, which fundamentally lack the ability to function outside their distinct purpose. However, as AIs are getting more complex their ability to identify their own weaknesses and make improvements is also improving. These AIs may not be yet able to think for themselves as we have been led to fear, but these ‘Strong AIs’ are making humans reconsider what is possible for an AI, including creative processes thought previously to be reserved for just humans.
Until recently, every single invention that has ever existed was the ultimate product of a human idea; and the premise of protecting ideas using a patent was somewhat simple:
The inventor has the idea, the inventor describes the idea to a regulatory body who then review the idea and only grant the patent if the idea is original and protectible. At which point the inventor would be granted exclusive ownership rights over the invention for a limited time.
When the idea of inventions and related patents were only limited to the imagination of the human mind, this process was relatively simple. However, the lines between human and computer “imagination” have become blurred with the existence of AI-assisted inventions, the integration of AI into creative software. Such software operates to allow AI to perform simple tasks involved in product creation at much greater speeds than humanly possible, however the resulting product ultimately still belongs to the human operator.
Now however, we are entering a new age of product design with AI-generated inventions. Where the AI is not simply a vehicle for processing data leading to an invention, but the creative process generating it. AI taught to “think” creatively, isolate an unmet need, and process how to practically and purposefully meet that need with a new invention; all of this with no input from any human operator.
An example of this is concept is a program called DABUS (Device Autonomously Bootstrapping uniform sensibility). This program has had a lot of media attention within the last year, as DABUS is an AI specifically created to simulate the human creativity in order to solve problems and generate inventions. DABUS caught media attention when it created two products that actually appeared to have some weight, at which point the people who own DABUS decided to patent its ideas. However, an unexpected issue then arose when it came time to fill out the patent application form. Who was the inventor?
Requirements of patent law is to name an inventor, which is loosely defined as a person ... who makes significant intellectual contribution to the claimed invention. The key word being person, which meant that subsequent US and EU patents that were filed under DABUS were rejected for not fulfilling the requirements of being an inventor, as DABUS was not a natural or legal person.
Image credit | BBC NEWS - food container designed by DABUS
One of DABUS' inventions is shown above; a design for a food container based on fractals to allow a more secure interlocking mechanism and also an easier gripping mechanism for transportation by robot. The second device is a lamp designed to flicker to mimic neural behaviour supposedly making it harder for humans to ignore.
As the first product specifically is designed for robots by a robot, it raises the question;
Can an AI ever fulfil the requirements of an inventor?
The answer to this is uncertain; as AI gets more complex this question will need to be continuously re-asked as to who actually earns the entitlement to any creation. Is it the programmer? Or is it the person who set it on its journey of creation? Or is it the AI itself?
This uncertainty has forced this seemingly closed issue of DABUS patents to resurface in recent months, in August 2020 an appeal was launched to reconsider both of the patents initially refused by the EPO (European Patent Office) and the UKIPO; which if they side in favour of the AI could change the way we see patents and artificial intelligence forever.
Lots of companies currently use to AI to make discoveries, such as in Biotech, the process of using AIs for high-throughput screening to find new drugs is common practice. These AIs screen millions of compounds against a target in order to find a ‘hit’ which can be the basis for pharmaceutical development, potentially generating huge revenues and bettering human health - a seemingly win-win scenario.
What would happen if the ownership of these potential drugs were taken away from the company and instead given to the AI that found them? Do AIs deserve recognition and therefore rights to what they contribute to society?
...we'll soon find out