The Patent Writing AI Revolution Is Here

In July 2021, the Federal Court of Australia ruled in Stephen L. Thaler [2021] APO 5, which allowed listing AI system DABUS as an inventor in a patent application. 

But what does that mean? Short term we are not looking at a research center filled with robots experimenting and building things and eventually filing a bunch of patents. So PHD’s and engineers, your jobs are safe, for now.

However, it might lead to machine learning algorithms that analyze patents and then comes up with just the right amount of differences required to file essentially a bogus patent. Which doesn’t sound very good at first glance, but who knows, maybe it will actually lead to at least a few useful “inventions”?

Stephen Thaler is the owner and developer of just such a patent-writing algorithm that he named DABUS which stands for “Device for the Autonomous Bootstrapping of Unified Sentience”.

An outtake from the patent for DABUS itself

DABUS has submitted patent applications in several countries. As a result of these applications, the government of South Africa recognized DABUS as the inventor on a patent.

It’s been more of a philosophical battle, convincing humanity that my creative neural architectures are compelling models of cognition, creativity, sentience, and consciousness. … The recently established fact that DABUS has created patent-worthy inventions is further evidence that the system ‘walks and talks’ just like a conscious human brain.”

Stephen Thaler

Initially, it was rejected by most patent offices around the world (including the US, UK, EPO, German and Australian patent offices). However, the Federal Court of Australia heard the appeal in the case and found that under the Australian Patent Act the AI could be listed as an inventor. The only thing is that a human (or specifically a “person”) has to submit the application.

The USPTO initially refused Thaler’s application because only a person can be an inventor. He of course appealed this and eventually it led to a compromise which I bet felt bittersweet for Dr Thaler “Thaler himself is unable to receive the patent on behalf of DABUS”. It does however, leave the possibility open that AI may be granted patents once they can be recognized as “persons”. Further down we will address how Thaler is tackling this issue.

As for Europe the case is still under appeal at the EPO, though at least one appeal has already been rejected. In the UK there was a similar process of rejection and appeals, but it ultimately lead to the conclusion that AI cannot be granted a patent under any circumstances.

But were the applications only rejected because Dr Thaler decided to name his system as the inventor? One could simply submit the application and claim that the algorithm that wrote it was simply a tool, like a computer, or pen and paper. And in that case there’s nothing in current legislation that outright prevents an application to be granted.

So what exactly did DABUS “invent”?

The process which DABUS uses to “invent”

With patents the invention itself can sometimes be a bit vague. It sounds counterintuitive, but for anyone who has ever read a patent application you’ll know what I’m talking about. The application has to cover as much ground as possible in order to be difficult to intrude upon by other products. What DABUS invented is “methods directed at an improved fractal container, which claims to be an improved food container for foods”. Specifically it is an interlocking food container that is meant to be easy for robots to grasp and stack.

DABUS’ invention, not a very human design.

So this particular invention is not that interesting in itself. What is interesting is that it’s an actual tangible product with real life applications. As opposed to a “bogus invention”.

Despite this I see a risk that software such as DABUS might be used primarily by patent trolls, a useless AI invention gets granted a patent which is only used to claim infringement. Then there’s the risk that AI inventions will flood the patent offices leaving little room for actual inventions.

Dr Stephen Thaler actually claims that DABUS is sentient. In my own opinion it sounds like a bunch of bogus. And as we addressed earlier, so does the USPTO. But this is what he says about that matter:

Arguably, DABUS may be considered “sentient” in that any chain-based concept launches a series of memories (i.e., affect chains) that sometimes terminate in critical recollections, thereby launching a tide of artificial molecules. It is these associated memory sequences, and the accompanying simulated neurotransmitter rush, that are considered equivalent to subjective feelings in humans (i.e., sentience). In this way, DABUS has an emotional appreciation for what it conceives.

Dr Stephen Thaler

Interestingly the website for DABUS mentions another AI inventor, the “Invention Machine” by John Koza which relies on genetic programming. Details about the “invention machine” can be found on Dr. Koza’s website.

Article as video: