Over the past few months a number of patent applications have been filed in various countries counting AI as a developer. This has led, in the end, to the availability of legal resources for four countries to process this question – two (the United States and the UK) measuring this concept and two (Australia and South Africa) allowing AI as a developer.
If you read some of the findings, it shows that the judges, rightly in my opinion, base their decisions on their interpretation of the law as it is written. This makes sense. However – this question is probably the first of many around the AI space and new creations. In this article, we highlight some of the technical basics that work here and raise additional questions that will come up soon after – and why businesses will need answers.
Is it AI Invent?
Although AI cannot be counted as a designer in a legal sense – do we believe AI can invent? I think so. The Merriam-Webster dictionary defines inventory as “production (something, like a useful device or process) for the first time through thought or ingenuity and experimentation”
Hmm. However, inventions are often the result of experiments, and in that sense – AI is innovative. As examples, AI can create art by combining newly given patterns in a variety of ways when it sees pre-assembled patterns. AI can generate code, which means it creates a solution to the problem with that solution expressed as a code.
Deepfakes are also a form of creation (although they are often dangerous and harmful). Deep deepfakes in the future can create all the movies, artificial medical data, and more. DABUS, the AI that underpins many recent legal decisions, is specially designed to produce a combination of ideas to build new concepts.
Why does it matter?
Now – if we think AI can invent, why is it important that AI can be counted as a designer? As I hope we can imagine AI doesn’t matter (at least for now!), Why would a business care?
With this, a small domain in copyright can help. Companies protect intellectual property in a number of ways, patents are one big way. Copyright requires one or more authors, who make a legally binding statement that they “created” the invention. Copyright machines, at least in the US, require that the list of founders be accurate and complete – meaning that anyone who participates in that process must be listed, and no one has failed to contribute properly. This company becomes the supplier of its establishment. The company’s IP rights are then protected by this patent.
If AI creates something a business wants a copyright – who is the designer?
One can argue with the person (or people) who created AI the founders of AI and by adding anything constructive. However, according to the above definition, one could argue that people would be breaking the law by claiming to be the creator if they did not have knowledge of the AI software that made this creation. This process becomes even more difficult and very fast. AI may have been created years ago and recently learned something new with new details. In this case who is the right person composer? Someone who wrote the original AI or someone who restored it recently before it built its structure? What if they were both unemployed at the company when AI created this establishment? What if AI was created in company A and sold to company B where it did its creation while working on Company B computers under the supervision of Company B employees?
As AI is so crowded, a simple form such as helping a person explore options by doing experiments, will contribute to the establishment. Companies will want to protect the invention. There is little known evidence that companies have been barred from filing patents when AI was involved in the creation of this invention. This can easily deter businesses.
If the two AIs are the same – who is the founder?
A previous example has shown why preventing AIs from being listed as designers can be a problem for companies. Now let’s take a look at the problems that can occur if AI is the founder. Unlike the human brain and the learning experience, which is different and (so far) unanswerable in direct words, the same is not true of AI. The AI brain (called model) resides in a computer file. By simply making a copy of that file – one can create a second AI with the same predictive power. So how can one identify the precise AI that produced the invention? Does it matter?
What about reading through transmission?
AIs also have a very powerful strategy known as transfer learning – where secondary AI can build on the first and most advanced features by learning new things. Many sophisticated AIs use this technique to reduce training time and costs. When such AI develops, how should the first AI find the right adjective? Are you a co-inventor? How would we know if real AI was involved in the invention?
Is AI human in some respects?
If you look at the decisions that have been made so far in relation to AI as patent holders – much of it has come down to whether the judges in question considered AI to be the equivalent of a natural person. There are other consequences to such a decision. If we decide that AI should be a natural person for the purpose of creation, what else is human nature? Are other factors that we say to people, such as patents, obligations and accountability, also work?
In a nutshell
As AI is more efficient – one can expect the number of different things created for AI to grow, whether it be art, books, poetry, code, movies, etc. For businesses, more and more people will be relying on AI, or a combination of people and AIs, to produce new things. If these inventions are not legally protected – businesses will suffer. The question of who owns that creation, by any definition of the word ownership – is a sensitive business question, and one that the rules will need to answer in the end.
So – whether you agree with the US / UK or Australia / South Africa – it is fair to say that the question raised by these films is very important. The correct answer however may not be yes or no – in AI personal recognition can be appreciated. Over time, the best solution would be to realize that AIs can invent, and change patent law for the effective retention of AI involvement in patents.