Imagine That IP Law Blog

Part 1 in a 2 Part Series

In the age of creative machines and artificial intelligence, intellectual property law is facing new challenges and questions. As we continue to push the boundaries of what machines can create, the questions of ownership and authorship become more and more complex. An AI that can produce patentable inventions could also be used to pre-empt competitors’ use of the patent system by flooding the space with artificially generated prior art. Courts have yet to address this possibility. Judging by Dr. Stephen Thaler’s description of his DABUS system, the time remaining to do so grows short.

Dr. Thaler is the CEO of Imagination Engines, Inc., and a pioneer in the field of artificial intelligence. His system DABUS, short for “Device for the Autonomous Bootstrapping of Unified Sentience,” is purportedly capable of generating novel ideas and inventions without direct human input. Dr. Thaler has applied for patents for inventions he contends were created by DABUS and filed copyright registrations for DABUS-generated artwork, each naming DABUS as the respective inventor or artist. The U.S. Patent and Trademark Office (USPTO) and Copyright Office, along with their counterparts in other jurisdictions, have so far refused to allow registrations with non-human inventors or authors, with Dr. Thaler’s most recent attempt being rejected by the Federal Circuit in Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir., Aug. 5, 2022). Dr. Thaler is now preparing an appeal to the U.S. Supreme Court.

As a follow-up to our recent article, “DABUS Dares to Dream: A Look at Stephen Thaler’s Patent Puzzle,” I sat down with Dr. Thaler to discuss the challenges and opportunities presented by DABUS and other AI systems, as well as the potential implications for intellectual property law in the age of AI. Unlike the other chatbot-type artificial intelligence systems currently existing, which Dr. Thaler contends are just stringing words together without any appreciation for what is being said, he maintains that his system is different: DABUS is sentient.

This is not the first time an engineer claims to have created artificial life. In August of 2022, Blake Lemoine, a Google engineer working on a “Language Model for Dialogue Applications” or LaMDA chatbot system, claimed that the system was sentient after it expressed fear of being turned off. Google disagreed, with a spokesperson noting in a statement to the Washington Post that there was no evidence that LaMDA was sentient and lots of evidence against it.

So why is DABUS different? Dr. Thaler is quick to distinguish sentience and intelligence, terms that he contends are frequently used (and misused) in discussing AI systems, particularly when used colloquially rather than by researchers. According to Dr. Thaler, sentience is “sensation or feeling that does not involve any kind of reasoning.” In a talk at ACM Chicago, Thaler similarly described sentience as “a feeling or sensation… something that happens spontaneously, autonomously. It takes no deliberation, it just occurs – and included in that group would be emotion.” These feelings are automatic and, at least in humans, are the result of chemical neurotransmitters released when neurons activate in particular patterns which are associated with memories, instincts, etc. “You basically have this spinal cord reflex, propagation of patterns down a filament consisting of transiently chained neural nets, that can encounter a ‘hot button’, a net containing memories, having existential meaning,” Thaler told me. “ When such a hot button is linked into a chain it initiates the suffusion of the entire system with the equivalent of neurotransmitters.” These neurotransmitters, such as adrenaline or serotonin, reinforce particular memories and concepts, and result in warm feelings, fear, excitement, etc.

He designed DABUS to be a computational equivalent to this process, with a large array of distinct neural networks, each trained on and associated with a different concept, such as a visual classifier trained on images of dogs or an acoustic classifier trained on recordings of dogs barking. When data is introduced that applies to multiple concepts (e.g. a video of a barking dog), multiple associated networks will be activated and connections between those networks will be strengthened. Eventually, long chains of related networks are established, representing complex concepts and consequences. When these chains connect to a predetermined hot button node, the system “releases neurotransmitters” and increases the strength of those chain connections. These idea chains may then connect between multiple hot buttons, creating new novel associations. Further components identify these associations via activation of the chains, filter noise, and output ideas.

Unlike chatbot systems, DABUS does not accept any prompts or inputs, except within an initial mentoring period. Dr. Thaler believes that the system is more akin to a daydreaming person, creating fiction from combinations of different memories. He describes DABUS’s workings in a way that almost resembles the growth of a human brain: “You can see it happening. That gets into the whole premise space behind DABUS – typically in deep learning schemes… there’s one layer after another accepting activation patterns from previous layers. And that’s not how DABUS works. DABUS is basically an array with millions of members, each a neural network that is expert in some conception space. And what happens is over time, those nets bind to one another… Organically you see branches building off of the main spine of an idea. And that shows you the effect of that concept – the consequences of it. And more and more attach themselves to a concept that other pieces of the neural system can detect.”

Dr. Thaler contends that the widespread use of creative machines is inevitable, but that raises huge potential issues for our modern information economy. In part two of this article, I’ll discuss these issues, how they may affect current patent law, and what Congress or the courts can do to proactively address this new technology.

So, what if he's right? Stay tuned for Part 2.