Want to listen to the full episode and all our other episodes?
Hearsay allows you to fulfill your legal CPD requirements every year.
Our yearly subscription is only $299/year.
With a yearly subscription, you can access all of our episodes AND every episode we release over the next year.
Episode 98
Buy Episode
Substantive Law
The Patent Predicament: Recognise or Relegate AI Patent Inventors?
Law as stated: 1 September 2023
What is this?
This episode was published and is accurate as at this date.
Patent Attorney Alana Hannah of FB Rice joins David Turner in the Curiosity Recording Room to digest the debate around AI inventors in patents, touching on the importance of patent attorneys as trusted technical advisors, and multi jurisdiction patent applications.
Substantive Law| What area(s) of law does this episode consider? | AI and patents. |
| Why is this topic relevant? | In the Thaler decision of the Federal Court, the eponymous provocateur sought to have his creative machine, DABUS, listed as the inventor on an Australian patent application. The Australian push was part of a broader project called The Artificial Inventor Project: “a series of pro bono legal test cases seeking intellectual property rights for AI-generated output in the absence of a traditional human inventor or author”. In Australia, that push failed. On appeal, a Full Court of the Federal Court held that an AI could not be considered an inventor for the purposes of patent law – an inventor must be a natural person. But the proliferation of creative machines over the last year has brought the concept of artificial inventors ever closer. |
| What are the main points? |
|
| What are the practical takeaways? |
|
| Show notes | The Artificial Inventor Project |
David Turner = DT; Alana Hannah = AH; Ross Davis = RD
| 00:00:00 | DT: | Hello and welcome to Hearsay The Legal Podcast, a CPD podcast that allows Australian lawyers to earn their CPD points on the go and at a time that suits them. I’m your host David Turner. Hearsay The Legal Podcast is proudly supported by Lext Australia. Lext’s mission is to improve user experiences in the law and legal services and Hearsay The Legal Podcast is how we’re improving the experience of CPD. Last year on our sister podcast Sidebar, we looked at an interesting patent decision in Thaler. Now at the time it was a bit of a sleeper story coming out before all of the hype around ChatGPT and generative AI. But today the idea behind the Thaler decision is more relevant than ever with the proliferation of advanced creative machines. Now in Thaler, the eponymous provocateur sought to have his creative machine – his generative AI in a sense – DABUS, listed as the inventor on an Australian patent application. Now, this Australian push was part of a broader project by Thaler called the Artificial Inventor Project, a series of pro bono legal test cases seeking intellectual property rights for AI generated output in the absence of a traditional human inventor or author. Now, in Australia, that push failed. The Full Court of the Federal Court, overturning the decision of the Federal Court at first instance, held that an AI could not be considered an inventor under patent law – an inventor had to be a natural person. But the proliferation of creative machines over the last year has brought the concept of artificial inventors ever closer and ever more relevant in the public conscious. Now our guest, Alana Hannah, is a highly experienced patent attorney at FB Rice. Alana’s background is in computer programming and data technologies and she brings a unique practical perspective to the discussion around inventorship and creative machines. Alana, thanks so much for joining me today on Hearsay. |
| AH: | Thank you for having me. | |
| DT: | Now I’d love to hear a bit about your career before we get started. There’s not enough lawyer programmers or programmer patent attorneys in the profession, I don’t think. We had one recently on the show in Ray Sun, but tell us a bit about how you got to where you are. | |
| AH: | Yeah. So my undergraduate degree is actually in astrophysics. So I studied… I studied that at uni, really, really loved it. I loved studying science. I loved learning. I did a lot of programming whilst I was doing that. And ended up working, doing a research internship for the Gemini telescope in Hawaii, where I was doing computer programming and processing data for them. And basically I was convinced I wanted to go on to do my master’s, my PhD. And then I did practical work in research and I went; “oh, I don’t want to do this at all”. | |
| DT: | Why not? You won’t offend any researchers. | |
| AH: | Yeah, no, I think it’s just very long-term projects. So people who had been working on some research projects for sort of 20 years. And because it’s all in space, there’s still no definitive answer on something. And I was like; “could I do that? Do I want to do that?”. | |
| DT: | That sense of kind of tangible feedback from your work or tangible impact. | |
| 00:03:14 | AH: | Yeah, and I think I was looking for something that was a little bit, I don’t know, maybe had some sort of smaller projects where you can see them to completion and just get a bit more fulfillment out of that. So I panicked and went on the uni job board and was like; “I just need to get a job out of uni”. I just needed to do something. And I came across an ad for a trainee patent attorney. And I was like; “well, okay, this is kind of weird because I don’t have an undergraduate law degree”. And so I started looking into it and I found out that there was a job called patent attorney, which wasn’t actually a lawyer, but it was basically a really niche area where you have the technical expertise to be able to work in helping people patent their inventions – because you need to be able to understand how an invention works, what it’s doing. So I ended up getting a job with a firm and undertaking my masters of intellectual property to register as a patent attorney. And then I kept studying because I just really loved the technical side of things. So I ended up doing a grad certificate in cybersecurity recently as well, just to flesh out that side of programming. It’s a bit of a journey because I so often have to basically clarify with people; “I’m not a lawyer, I’m a patent attorney” and a lot of people don’t know the difference. |
| DT: | I suppose because of the term “attorney” from American legal media. | |
| AH: | Yeah, you’re an attorney at law. And I think a lot of people hear patent attorney and they go; “lawyer. Yeah, you’re a lawyer”. | |
| DT: | It’s such an interesting intersection between these two fields though, isn’t it? | |
| AH: | Yes. And it’s very much an intersection of science and technology and law because you can’t be a patent attorney if you don’t know the legal side. You can be very technically proficient by the way that you draft patent applications that you prosecute them. You need to understand how they might be litigated, the kinds of objections that you might come up against and that kind of thing. | |
| DT: | Yeah. You can really only understand the factual matrix if you have that technical background. You can’t understand the inventive step. I remember I was in uni around the time of the Apple Samsung case, that first generation of iPads, right? So it was all about the alternative kind of approaches to haptic feedback and finding the coordinates of a tap on the screen. And even just reading a judgment, that was quite difficult to interpret. So you really do need that technical background. What sort of people are you working with as a patent attorney? Who are your clients? | |
| AH: | Really broad range of clients. So I work with anyone from a solo inventor who’s created something in his garage and thinks; “yep, I can sell this to research organisations, universities, small startups, and then big global companies” because all sorts of people are filing patents in Australia. And then there’s a lot of local Australian organisations that are filing patents globally as well. So we also help them facilitate that. So just a very big mix of clients at the moment. I mainly work in the physics and computing spaces because that’s where my technical background is. So most of my clients are dealing with computing technologies. | |
| 00:06:21 | DT: | Makes sense. And what’s the intersection with the broader legal profession, I suppose? Do you often have IP lawyers coming to instruct you on the technical aspect of the application? |
| AH: | So pretty much everything from a client coming in, preparing the application, filing the application, patent attorneys, we can do all that ourselves. So unless you’re a registered patent attorney, you can’t actually draft a patent specification. So anything to do with liaising with the patent office, so whether it’s filing or communicating with them, patent attorneys do all that. When it comes to litigation, so maybe there’s some infringement happening or there’s an opposition, so someone opposes a patent that’s going through the patent office, usually lawyers and sometimes barristers will come in and we will be the technical support for that. | |
| DT: | So it’s like the experts, not the witness, but the experts. | |
| AH: | Exactly. Yeah. And so I’ve been involved in a few litigation cases. There’s sometimes just little technical things that lawyers don’t necessarily pick up because they aren’t super familiar with the technology. And so that’s what we’re there for, we’re there to assist with that kind of thing. Often we’ll do a lot of technical research as well, just to assist the lawyers in forming a case and formulating arguments and things like that. | |
| DT: | In that kind of litigation, it makes a lot of sense to have the patent attorney as the – I hate these terms – the clean expert or the dirty expert, the shadow expert, the one who can’t give evidence because they’re partisan to you, but help you to understand the expert evidence sufficiently to properly cross-examine or to properly make submissions about it. It makes so much more sense to have a patent attorney do that because you understand in that litigation what are the legal issues as well as what are the factual issues or the technical issues. | |
| AH: | Yeah. I’ve been in situations where we’ve been interviewing an expert witness and the lawyer has been asking all the questions and I’ve been sitting there essentially writing all the notes from the expert witness so that I can translate it to the lawyer later so that he can understand. And even if you’ve got someone and you’re going to cross-examine them, it’s handy to have a patent attorney that can just be like; “ask them this term, they haven’t really clarified so ask them about this”. And you can add that support to those cases. | |
| DT: | You give that depth to the advocates. | |
| AH: | And it does help to have someone that understands the law but has the tech background to understand what’s being said as well. | |
| DT: | Absolutely. Now let’s jump into AI inventorship. As I said in the intro to this episode, we did an episode on this tail end of last year, September, October-ish. Not a lot of interest. I think we re-released it in February or March and it blew up because, you know, completely different level of interest in artificial intelligence at the start of this year compared to the mid to late months of last year. So I feel like we’ve had a couple of short conversations about this case and then I’m excited to really dive deep into it before we talk about the case itself. This concept of AI inventorship, why is it important? | |
| AH: | AI is very trendy at the moment, of course. And so a lot of people are interested in what AI is doing and what it means to have AI involved in our world. So our jobs, how we do them, what it means for people writing essays in uni and school and everything like that. And I guess there’s a lot of companies as well looking; “well, how can we use this technology to create products or free up resources” and that kind of thing. And so I think everyone can see that there is a really vast array of applications to use it for, but there’s still a little bit of uncertainty around what it actually is. I think there’s a general perception that it’s just some magic genie that you can ask it to do something or you can ask it to give you something and it will just come up with it and say; “here you go”. But yeah, it’s definitely something that a lot of people want to utilise at the moment. | |
| 00:10:25 | DT: | It reminds me of that Arthur C. Clarke quote; “any sufficiently advanced technology is indistinguishable from magic”. And it feels like we’re in that stage. And with that comes a lot of misconceptions. You know, I was speaking to a big advocate for using ChatGPT in their own work and they were saying; “oh, it’s great. I use it to summarise websites. I just give it the URL and ask it to write some content based on what’s at that URL”. I’m like; “you know it can’t read the content of the URL, right? It’s not looking at that when it’s performing the task that you’re asking. It’s just the next token classifier”. But for them, it seems like that’s what’s happening and they’re developing this relationship with the technology. I suppose it’s also a tool where the output is so far removed from the input in the sense that we have a lot of creative tools at our disposal. We don’t tend to have the same quandaries about attributing copyright for an artist who uses Illustrator – Adobe Illustrator – to generate an image because the connection between the artist’s input and the output of the program is so proximate. But with generative AI, that distance is much greater, isn’t it? |
| AH: | Yeah. And I think it does often seem like the AI is understanding and knowing what it’s outputting and it’s telling you the truth and it’s not going to give you something that might infringe someone’s copyright or anything like that. But it’s important to remember that it doesn’t actually know what it’s outputting. Like it’s not sentient. It’s not cognitive. It doesn’t know. And so the idea that we can just ask for something and it’ll pop out like a paragraph; “oh, here’s a summary of what’s on this website”. You go; “great, I’ll just copy paste and use that”. You’re not necessarily thinking; “maybe it’s taken this from somewhere. Maybe it’s come up with it on its own”. | |
| DT: | Yeah, that’s right. And one of Lext’s other companies; we have a legal information tool powered by generative AI and the innovation there is building that knowledge retrieval architecture over the top of a large language model to provide that external source of information. But when you’re not doing that and you’re just relying on this parametric model that is really a black box in terms of explainability of output, it’s very difficult to see what information in the trillions of tokens that comprise the training data set have contributed to the answer you’re receiving. | |
| AH: | Exactly. Yeah. These are large language models. The data sets that they train on are huge. It’s sometimes hard for us to even fathom how much information they’re pulling from. | |
| DT: | But also that the model doesn’t hold a representation of that information. | |
| AH: | Exactly. | |
| DT: | It’s comprised of weights drawn from that information. So there’s no sort of live queryable database of that training data. I think that’s the big misconception. And maybe you’re venturing into your cyber security studies, but I see people wringing their hands about; “well, if I put my personal information into a large language model query, that can be used for training in the future. And will someone end up with my personal information in their completion later?”. And of course, that’s not really how training data works. You know, we don’t query the database and then produce based on it. There’s a probabilistic process there based on, again, determining weights for the nodes in this very large network of nodes. So, yeah, a lot of misconceptions. But at the same time, tremendous interest and understandably so because it is a hugely transformative technology. It feels a little bit like being at the start of the smartphone or the start of the Internet, these big strides in computing and software. Now, this full court decision that we mentioned at the top of the episode that pretty firmly stated; “no, an artificial intelligence cannot be an inventor”. Of this concept of agent nouns, can the word inventor be the same as the word dishwasher? Could be a machine, could be a person? But no; must be a human being. That sort of accords with the position in copyright that we must have a human author. So someone else other than the AI owns the invention. Why is it so hard to identify who that is? | |
| AH: | It’s tricky. So this case was to do with AI being listed as the inventor. The owner of the invention may be distinct from the inventor, which is just something to clarify here. Because they weren’t claiming that the AI could own the invention, just that it was the inventor. | |
| DT: | Yeah. For some of our listeners who might not be as familiar with patent, but might have a passing familiarity with copyright. This is a bit like copyright and moral rights in the sense that the publisher typically owns the copyright. The author owns the moral right to have that copyright attributed to them. | |
| 00:14:59 | AH: | Yeah. So identifying the inventor for AI generated or AI assisted inventions, it can be difficult because it’s hard to know how much inventiveness – I’m saying that in quotes – the AI should be attributed. So has the AI come up with this thing all on its own? Or has somebody designed a specific AI algorithm, machine learning algorithm, given it the instructions, given it all these parameters for it to then come up with something? |
| DT: | Yeah. If there’s a kind of almost a deterministic relationship between the input and the output, where’s the inventive step taking place? | |
| AH: | Yeah. And actually, the person that invents the AI algorithm may be different to the person that’s using that and inputting parameters and information so that it can design a new coffee machine or something like that. And so then it gets really tricky. Or who owns there? Is it the person that designed the algorithm? Is it the AI algorithm? | |
| DT: | And in a way, the Thaler case is the simple example, right? Because we do have a situation where the person who seeks to be recognised as the owner happens to be the same as the person who would otherwise be the inventor, who happens to be the person who owns the algorithm. We’ve got this co-location of all the actors in one person. But typically, that won’t be the case. You know, we’ll have artificial intelligence developed by a large technology company – OpenAI or Google or Cohere or whoever – and that’ll be used by a consumer or a company. And you might have employees of that company who act to create those inputs. You have many more actors in reality than I suppose the Thaler case would indicate. | |
| AH: | Yeah. And the Thaler case is a very simple case. And it does make it easy just as an exemplary case to kind of look at; well, the person Thaler that created the AI, DABUS, is the owner. And DABUS is listed as the inventor who created the actual invention. | |
| DT: | Which was like a food container. | |
| AH: | There’s a few. One of them was a food container. But I think Thaler said that it wouldn’t have been right to list himself or a human as the inventor because they did not contribute in any material manner to the conception of the invention. That was DABUS, the AI. TIP: As we mentioned at the top of the episode, we’ve talked about this Thaler case before on our free show the Hearsay Sidebar. But for those of you who didn’t catch that one – and I recommend you go and check that out on your favourite podcast platform – here’s some background information to the case. The case citation is Commissioner of Patents v Thaler [2022] FCAFC 62. Dr Thaler is the owner and operator of the thinking machine, DABUS. DABUS stands for Device for the Autonomous Bootstrapping of Unified Sentience. Thaler used DABUS to design a type of food container and subsequently attempted to list DABUS as the inventor on the Australia patent application. Dr Thaler wanted to push the envelope and the conversation around AI inventorship, so he applied for patents in many different countries, in order to test each jurisdiction’s laws around AI and patent inventorship. In the Australian case, the parties agreed that:
So the question was merely whether DABUS could be listed as an inventor on the patent for the food container. The Full Court of the Federal Court’s answer? No. Inventors listed on a patent had to be a natural person. It is hard when we’re talking about AI inventorship; well, where do you draw the line? How do you know if the AI is the inventor or whether it’s a human? And I think that usually it comes down to, well, how much has it contributed or has it just been used as a tool? Has a team of scientists been using an algorithm to run different scenarios? But actually it hasn’t come up with anything. It’s just automated the process of running through things faster. So, yeah. | |
| DT: | Because, remind us; there are a few sort of critical elements to the patentability of an invention. It’s a manner of manufacture or a… that’s the right term? | |
| AH: | Yeah, yeah. Very good. | |
| DT: | Going back to undergraduate intellectual property here! And that manner of manufacture requires an inventive step, something that sort of takes the state of the art beyond its current state. Is that correct? | |
| 00:19:58 | AH: | So there’s usually three separate things. So manner of manufacture is actually slightly separate to inventive step. It basically means it has to be suitable subject matter to be patented. So for example, it can’t be an idea. It can’t be a business method or a method of running a business. It has to be something tangible, like a technology or a new way of doing something. Separate to that, you have the requirements of novelty, which means it has to be new over what already exists. And then the inventive step, which means it has to be… it has to have enough inventiveness in and how it’s differentiated over what already exists. So for example, changing the color of something is not necessarily inventive. |
| DT: | Although it might be new. | |
| AH: | Although it might be new. Yeah, exactly. | |
| DT: | Makes sense. And so the kind of when we’re talking about AI inventors, manner of manufacture, tick, novelty, presumably, tick. But it’s that inventive step where things get muddy because we’re not sure who to attribute that to. | |
| AH: | Exactly. Yeah. TIP: Just to summarise, for an invention to be patentable in Australia, it needs to satisfy four requirements:
On the point of inventiveness, an invention cannot have been publicly disclosed before you apply for the patent. This means that it cannot have been disclosed in an obscure magazine from 20 years ago in a different country. The publishing of the patent has to be the first public disclosure. Alana will touch on this a bit later in the episode. And when you’re looking at inventorship in the patent space and you look at the definition of inventor, usually it comes down to, well, who has contributed to the inventive concept? And it might be super clear that, well, the AI generated these three options and they were all inventive. And so we want to give the attribute that to the AI, but it might be that we’ve told the AI to look at these three options and run experiments to see if they’re valid. And then we go; “well, we came up with the option”. So us as humans should be attributed as inventors because we’ve basically come up with the inventive concept. And so it does get tricky when you’re using AI as a tool versus when you’re relying on it to come up with ideas, essentially where to put your inventorship, I guess. | |
| DT: | And in Australia, at least we have that clarity that the AI cannot be the inventor. It must be a human being. So, I mean, who is listed as the inventor on Thaler’s application? | |
| AH: | Yeah. And it is tricky because I think most companies or people would go: “okay, well, we can’t list the AI as an inventor. So we’ll just list the human inventor”. Like we’ll just list the person that was in charge of running the algorithm or running the project or whatever it is. But there are issues with naming somebody who is not an inventor as an inventor. So you can actually have your patent invalidated because you’ve incorrectly listed someone as an inventor or not listed someone as an inventor. | |
| DT: | You can of course have multiple inventors. | |
| AH: | You can have multiple inventors. But the issues arise where you’ve got someone listed that’s not actually entitled to be listed as an inventor. | |
| DT: | So, merely excluding the artificial intelligence from the kind of eligibility to be listed as an inventor doesn’t really solve the problem, does it? Because if the person that you have listed in lieu of the artificial intelligence cannot be considered the inventor themselves, I suppose it challenges even the patentability. | |
| AH: | Yeah. We’ve seen the test cases in Australia and the US and Europe about naming an artificial intelligence as an inventor, we haven’t really seen any test cases where someone’s challenging a granted patent, where the inventor was an AI and a human has been listed that shouldn’t have been listed. And I guess it will be interesting to see whether companies or organisations or people take that approach of just listing humans as inventors and almost try and hide the involvement of AI or whether they actually don’t file patents at all. So they go; “well we can’t list AI as an inventor, but we don’t want to list somebody else because if this patent gets challenged, this is really important technology, we’ll just keep this to ourselves”. That has other ramifications in terms of really not incentivising innovation, not getting technology out into the public, that kind of thing. | |
| DT: | Yeah. Very different strategy for protecting your monopoly on that technology. I suppose we might see these cases. I mean, one area that I think of that I suppose is already using artificial intelligence in product discovery or invention is drug discovery. Although even then that probably is more of the category that you are describing where you’re automating a laborious process, validating the combination of many factors that’s largely guided by a human inventor rather than that kind of broader generative AI creativity sort of step. But you raise a good point, which is that this is just one case in a whole range of test cases across the globe as part of the Artificial Inventor Project and those cases have had different outcomes, right? So tell us a bit about the project globally and where it’s had some success other than Australia. | |
| 00:25:53 | AH: | Yeah, so I guess the idea of the project, as you said, is to put through these test cases where AI is listed as an inventor, so they were filed in a number of jurisdictions around the world and basically get those countries to really consider this idea that AI can be listed as an inventor. It feels like there has been more rejection for those applications than acceptance. There is obviously in the US and Europe, patent offices, they have rejected again, similar to Australia, that it needs to be a natural person. But the case did have some success in South Africa. So the patent actually got granted and that is a bit of a quirk of how the South African patent system actually works. So there’s no examination in South Africa. The case, essentially, once it’s filed, it goes through a bit of a formalities check and then it proceeds to acceptance and then grant. |
| DT: | Now, so they rely on objectors to almost validate the patent. | |
| AH: | And so no one has objected to that yet. But it is an interesting case because obviously a lot of the main jurisdictions, just like in Australia, they essentially got stopped at the formality stage. So they didn’t get to being examined. They got stopped at; “well, you haven’t listed an inventor and as a requirement, in order to file the patent, you need to have listed an inventor”. And so a lot of those ones got stopped at that stage and rejected on that basis. But because it’s all gone all the way through in South Africa, this didn’t come up in the formalities. If that ever gets challenged, would it stand up? That would be when South Africa would have to essentially consider that in their law and what that means. | |
| DT: | But I suppose it is a formalities issue, right? If I’m right in thinking that the examination in Australia is really about those things we discussed before: is this a manner of manufacture? Is there an inventive step? Is it novel? Those things are independent of who the inventor is, right? That is a formality. So it is interesting that you’ve got that result, but again, it’s a bit… it’s not been the subject of a judicial determination. So it’s hard to look at that as something that is a persuasive weight globally. | |
| AH: | Yeah. There’s no real decision that’s been issued around that. So whereas in Australia, you have the Full Court decision and everything. It’s gone through the various stages. So yeah, it is tricky. I think there are a few jurisdictions that are still pending appeals and things like that. But the general consensus is that a lot of the countries have patent laws that basically say; “well, this is not designed for non-humans. It needs to be a natural person that is listed as the inventor”. | |
| DT: | Now you mentioned before that at FB Rice, you’re assisting inventors, not only with Australian patent applications, with global patent applications. And there is a system for the kind of mutual recognition or global recognition of patents, isn’t there? There’s a similar to the Madrid protocol for trademarks. There’s a protocol for the kind of global recognition of patents. Can you tell us a little bit about that? | |
| AH: | So there is a thing called the Patent Cooperation Treaty. And basically what happens is to start the patent process, you will file a provisional patent application in a jurisdiction. It can be any jurisdiction that is party to that Patent Cooperation Treaty. You then have 12 months to file what’s called a complete application. And if you filed in a country that is part of that Patent Cooperation Treaty, you can file something called a PCT application, which essentially reserves your right to file in 160 different countries that are party to that Patent Cooperation Treaty. So each jurisdiction has their own patent laws and they’re all slightly different. So we talked about that manner of manufacture, novelty, inventive step in Australia. There are similar laws in different jurisdictions, but everyone has their own case law and specific exemptions and things like that. And so the idea is that you can file this PCT application. And then once it gets to sort of 12 months later from that, you can file into any countries of your choosing that are part of that treaty. And it sometimes just gives you a little bit extra time to figure out; “well, where do I want to actually file?”. Because realistically you can’t file in 160 different countries. It would be insanely expensive. Each country then has to go through examination. You have to pay renewal fees. | |
| DT: | So, well, yeah, just managing the portfolio after the first year. | |
| AH: | Oh yeah, exactly. Yeah. So yeah, I think that it is a very useful tool when companies are like, maybe they’re not sure, or maybe they just want to streamline how they file in each country. Because once you’ve got that PCT application, it’s got all the information on there or it’s published. So it’s very easy just to send to whichever countries you want, and essentially file… | |
| DT: | Fast tracked. If you’re resident in one of those 160 countries and you have perhaps South Africa, you’ve got a patent registered that is attributed to an artificial intelligence and have that PCT fast track process into other jurisdictions. You can see how a difference or a disconnect in terms of how AI inventors are treated in different jurisdictions could really muddy the waters in terms of how those PCT applications are treated, right? Because I suppose am I right in thinking that kind of implicit in that process is the fact that your home jurisdiction has accepted the application at least, if not formally registered the patent? | |
| 00:31:54 | AH: | Well, I guess it’s a bit tricky because when you initially file in your home jurisdiction, a lot of countries like Australia, there’s no requirement to actually list an inventor when you file that provisional application. But then once you file the complete application, so if you file that PCT, it’s filed with the World Intellectual Property Office. So it’s considered, I guess, to use the term global application. It’s not necessarily party to like Australian laws. So if they list DABUS as the inventor on the PCT case, the Australian patent office is not objecting to that; it’s not their requirement at that stage. So it’s only once it’s filed nationally at the end of that PCT timeline, that’s when they check for the formalities, which is why you’ve got all these sort of decisions happening around the same time. All these things coming out. |
| DT: | Because they’ve had that WIPO application and now it’s the domestic ones that are being determined. Yeah, that makes sense. And I suppose the WIPO, because they’re not finally determining the application, they can be a little more permissive in terms of the form of that application. | |
| AH: | And so, yeah, the PCT application, it will always be an application. It’s never granted. Essentially it’s a placeholder. So it’s just holding your place while you decide on the countries. So they don’t have to do a really rigorous report or investigation into, well, who is DABUS. Maybe someone’s changed their name to that. Like, but once you get into each individual country, each country will have their own formality requirements for patent specifications and patent applications. And they will check those before they essentially say; “yes, okay, this can be filed”. | |
| DT: | Now, I suppose you’re meeting with inventors, you know, every day you’re advising inventors on their patent applications here and across the globe, with artificial intelligence playing a greater role in that invention process. What are some of the benefits and some of the risks for people like your clients, the inventors and the company and the technology companies that they work for? | |
| AH: | Yeah, I guess the benefits are what we were talking about before. It is a great resource to have access to. Suddenly a lot of these open source generative AIs are coming out and you can automate sort of tedious tasks and free up your staff for being creative or coming up with new ideas or doing more substantive work. | |
| DT: | That’s a great point, actually. | |
| AH: | Yeah. | |
| DT: | We’ve been talking in this episode about artificial intelligence doing really creative, inventive work. But generally speaking, it’s the humans that are going to be doing that because they’ve got time to do it because all the menial stuff is being done by AI. This is the exception rather than the rule. | |
| AH: | And when we talk about AI creating they still require a level of instruction. So humans still have to put in, give them enough instruction for them to come up and generate something. You can give a ChatGPT a link and say; “can you summarise this webpage for me in a paragraph and it’ll pop something out for you”. Now you might then take that and tweak it and do what you want with it. But sometimes just having the summarisation is really useful because you don’t have to go through and read all that stuff. So it is a good resource if used correctly, which is where some of the risks come in, I guess. | |
| DT: | One of the sort of uses from a creative or an inventive perspective – and I’d be interested to get your perspective on this as a programmer – when I’m doing a programming task, we have this concept in programming of the rubber duck. That describing the problem to a rubber duck, it doesn’t have to talk back, but just the act of describing it can often help you to understand the problem better. And you find that by speaking through it step by step, you understand something that you thought you understood quite well, just by externalising it. It’s a rubber duck that talks back. I’ve found that attempting to use generative AI to help me solve programming problems often helps me find the solution, even if it’s not the solution the AI gave me. | |
| 00:36:03 | AH: | Yeah. Sometimes it’s useful even to just be like even to say; “well, what are some things that could cause this?”. And it’ll give you five things. Three of them might be rubbish, but two you may not have considered. And so you go; “okay, let me just use that and I’ll just go down this path”. But you may not have come up with those things on your own. So it is, yeah, definitely useful as a tool, especially in debugging code. |
| DT: | Or discovering what you shouldn’t do in the sense that I’ve sometimes asked; “okay, I want to make an edge function for my database to do all of these things based on when this query runs. So write that SQL function for me”. And it starts. I’m like; “oh man, I shouldn’t do it that way”. It’s capable of it, but you realise what an anti-pattern it is. I think it’s definitely that sort of differential multiplier for an inventor, it’s something that can really accelerate the creative process when it’s used as that kind of pair invention. This is probably venturing more into copyright, I suppose, but with computing software being an area of focus for you in your practice, the use of code generative AI like CoPilot, now OpenAI’s code interpreter, CodeStar I think is the open source model. Does the use of those tools to generate code or blocks of code in an application that might later come to be subject to a pattern application? How does that affect the patentability of that kind of overall invention? | |
| AH: | And this is where it can get a bit tricky. There’s a lot of intellectual property considerations when you’re dealing with coding and AI. | |
| DT: | Just software at all I think is a tricky area, right? | |
| AH: | And so if you’re working on something and you’re saying; “well, can you just write me a function that does this thing” and it pops out your function and you copy paste that into your code and work on it. It’s not necessarily going to affect the patentability of the software that you’re designing. It does depend though on what it is that you’re pulling out of it, what the instructions are, what you’re asking for. If you’re asking for something; “I want a function that does this. It needs to do these things and have this basic speed and this accuracy”. And it comes out, essentially. If that’s the inventive part of your invention, then it can get a bit tricky. Like, well, who’s actually invented that? | |
| DT: | Yeah, well it all comes down to what the inventive step is, right? And there’s two ways to do it, right? One is I need a function that produces this output, that kind of very; “I don’t know how I would do it, but please try and work it out”. And there’s the; “okay, I need you to loop through this iterator. And then at each step, I want you to do this”. That kind of “take my idea and put it into syntax”. There it’s really clear that the inventive step is with the person. | |
| AH: | Yeah. And I think sometimes a lot of people they’ve got a concept. A really common one is: I want to design a mobile phone application. And so it’s like, they just say; “I want an application that does this and looks like this” and it just pops out, but there’s no real understanding of actually how it’s functioning, how it’s taking in inputs and giving outputs and processing information. And to get a patent application, you actually need to describe in detail how it works. So quite often when people are pulling stuff and they’re not understanding what it’s doing, because the AI has just designed some software code for them. That can be really tricky to: one, put in words how it works with enough technical detail that you can meet that patent eligibility requirement. But two, you also don’t know whether that’s already protected by a patent. So if you don’t have an understanding, if you’re just copying, pasting stuff out and typically, typically I see non-technical people do this. So maybe someone who doesn’t really have a background in programming, but knows that AI can generate something for them. I see them come with: “oh, we’ve got this software”. I ask; “how does it work?”. And they go; “oh, well it does these things”. And so it gets tricky; “well, did you check that there’s no copyrighted code that’s been used in that? Did you check that the technical processes that it’s using in the software are not covered by patents?”. Because the AI doesn’t know either. It doesn’t know if it’s giving you something copyrighted or not. It doesn’t know if it’s giving you something that’s covered by a patent or not. And so it’s just important to be aware of those sorts of things as well. Especially when you’re pulling together lots of different stuff to make sure you’re aware that there may be patents out there that are potentially infringed, but you have to just check. | |
| DT: | And I suppose even more broadly than software, just in terms of patent applications generally, although AI can be a great accelerator, it does, if there’s an over-reliance, make it very difficult for you. It’s not just those steps of, is it novel? Is there an inventive step? Is there a manner of manufacture? There’s a process that requires an in-depth understanding of your own invention. That will make it very difficult to prosecute that process unless you have that intimate involvement. | |
| AH: | And so often when we do draft patent specifications, we may be talking to the CEO of a company and then we say; “put us in touch with your tech team so they can actually explain to us what’s going on”. Because when we ask and the CEO says; “oh, we’re using AI, we’re using machine learning to do this”. And I say; “okay, can you explain that to me?” And he’s like; “oh no, we’re using AI”. And there’s just no understanding. It’s again, going back to that magic genie, there’s no understanding of how it works. It just does the thing. And so often we will sit down with tech teams and actually get the technical detail to figure out how it works and make sure that if we do any searching, that we know what we’re looking for. We’re looking for other patent applications or copyrighted software that may have used these processes and things like that. | |
| 00:41:53 | DT: | And how do you do those searches? Is it the tech team tells you; “oh, this part’s based on a base classifier”. And then, so you go and look for other patent applications that’s mentioned base classifiers. |
| AH: | So typically when we’re talking with the tech team, we first figure out what the inventive parts are. So there’s a lot of parts of software that may be just general functions and it’s not necessarily new to use them, but maybe the combination of these three things together is what’s inventive. Or actually they’ve just come up with a new way of doing this particular thing that improves the speed at which data is processed. And so we’ll go; “okay, well, that thing. That’s what we want to search for. We want to see if anyone else is doing this”. And so like at FB Rice, we have an in-house searching company that we work with, that basically we have access to a bunch of patent databases, publications, journals, all sorts of anything that might disclose or be publicly available. | |
| DT: | Of course, even if it’s not patented, if it’s a disclosed invention, it can’t be patented, right? | |
| AH: | So if you’re looking for something inventive, but also has to be new. So if someone in Russia has put out a paper on; “oh, I’ve got this new process and here’s how it works”. And you go; “well, is this what you guys have been doing?”. And even if you don’t know about it, unfortunately it’s already out in the public. So what you’re doing has to be inventive above what is already out there in the public domain. | |
| DT: | Yeah. Wow. That must be a real art, performing those searches, right? | |
| AH: | Yeah. And a lot of the people that we work with are just trained to understand, like we might give some keywords and some key phrases and then they take them and also expand on that. They know exactly what keywords are we looking for? What variations are we looking for? So you don’t want to make it too narrow that it misses stuff, but too broad and you get too many results. So it’s a real balance. | |
| DT: | Now it just feels inevitable that there will be adjustments, big or small, to our intellectual property legislative landscape as a consequence of artificial intelligence. I feel like the copyright sphere absolutely will require some clarification at the very least because we have so many copyright lawsuits against owners of foundational models at the moment. But I imagine that’s the case with patents as well. And IP Australia recently released a report on the intellectual property landscape and artificial intelligence for the kinds of intellectual property that they manage. So trademarks, designs, circuits, circuit layouts, plant breeders rights and patents. Tell us a bit about what’s happening in that space. What are some of the proposals around adapting the law to respond to these artificial intelligence changes? | |
| AH: | I think IP Australia knows that they have to change basically. And I think most countries know that they need to address this idea of AI inventorship and dealing with AI technologies. Because there’s just no laws around it at the moment. We were talking about how a lot of countries’ laws are designed for humans to own copyright, for humans to be listed as the inventors. And so at the moment, I know that there is a lot of requests for comments. The USPTO, the US Patent Office has done a number of call outs for requests. IP Australia recently released their report that did float a couple of options. Do we list AI as a joint inventor? Do we have a specific section that says; “was AI used in developing this?” And you check a box that says yes. But almost all the reports that I read that are coming out are like; “we need to do more research in this area because we need to develop policies around this”. It just feels like almost there’s a little bit of a divide between what it means to develop a policy around AI generated inventions. Is it AI assisted or AI generated? | |
| DT: | Because that policy objective of saying; “okay, well, the inventor has to be human”. That’s still achievable and feasible, but it’s that all of our heuristics around how the law works need to adapt. We need to have frameworks for thinking about what it means to invent something using artificial intelligence. | |
| AH: | Yeah. And is there a way that if you are using artificial intelligence and it has generated something, is there a way you can automatically transfer it to the company that’s using it or the person that’s using it? Is there a mechanism that we can do that to make sure that you are listing an inventor, but it’s not just an AI that’s listed. | |
| 00:46:36 | DT: | It’s an attribution. |
| AH: | And then one of the tricky aspects as well around developing policies in this area is where is the line between; “well, I’m using AI as a tool” to “oh, AI has actually come up with this whole thing itself”. Is that possible without me giving the algorithm instructions? I mean, countries will start getting policies in place just even as AI is used more and more, just in the general workplace. Where that line is I think will potentially differ in different countries. | |
| DT: | Yeah, absolutely. I mean, I feel like we’re already developing a bit of a consensus around generated text because of course that’s just such a common use case, right? That there is a consensus that no copyright vests in artificial intelligence generated text because it has no human author. So it’s just not subject to copyright. And I suppose we need to develop those norms, but the difference with patents unlike copyright is there is a source of truth. There’s a source of, kind of, verifiability in the form of the patent office and it’s up to our regulators to give us that clarity. | |
| AH: | Yeah, and I guess usually when something like this is happening there is request for comments. There’s roundtables that happen where you’ve got all sorts of stakeholders basically involved in, I guess, just making the policymakers aware of all the risks and the benefits and the different ramifications of what various policies will mean. So if you don’t let AI be listed as an inventor, will companies just keep it a secret? | |
| DT: | Just not register. Just rely on the confidentiality. | |
| AH: | Rely on the confidentiality. But then you’re not getting new technology being put out into the public. So because people can build on patent applications, a lot of scientists they’ll look at patent applications. But yeah, are you de-incentivising innovation? So I think that will be, that is one of IPA’s objectives is to keep incentivising human inventors, companies to keep inventing. | |
| DT: | Because that’s the bargain, isn’t it? It’s that by disclosing the invention to the world, you’re awarded the monopoly. Everyone else can stand on your shoulders and we move forward collectively. But – this is an episode all on its own – that when technology cycles are so much faster than seven years or however long your patent protection typically awards you. Keeping that as a proprietary piece of information might make good business sense, but it does result in the hoarding of knowledge. A few very innovative technology companies in the United States, but that’s probably for another episode. Look, we’re nearly out of time, Alana. But before you go, I wanted to ask you a question we usually ask our guests about, sort of, our student listeners. Sounds like you have a very exciting and interesting job working with these inventors at the cutting edge of artificial intelligence, programming, computing. If we’ve got a listener who’s listened to this episode and thinks; “geez, I want to do that”. What tips would you have for them to get into patent attorney work? | |
| AH: | Yeah, I mean, one of the things is that you need a technical degree in something. So if you’re studying law and computer science or law and biology or whatever it is, that is your entry point is to have a technical degree. But I think if you’ve got a passion for the technology, it’s going to be really easy to make that step into training as a patent attorney. Alternatively, if you want to get into IP law, know a lot of the lawyers that I work with, they have technical degrees. So not necessarily in a specific technical field. So as a patent attorney, you can only work within the technical fields that you’ve got experience in. So I can’t, so I’m not qualified to work in chemistry and biology, but a lot of the IP lawyers will have some technical degree and it just means that they can understand technical terms. They can read technical documents. And so I guess if you’re looking to specialise, you get into either IP law or being a patent attorney, I think just look for opportunities where you can really hone those skills communicating technical concepts. I used to love doing all the lab work when I was at uni and everyone else hated it. And I was like, this is great. You just get to muck around with stuff and then explain what’s happening. And I loved explaining stuff. And so I think if you’re honing those communication skills and how you can make really complex technical topics seem really simple, get people to understand them. That’s a really useful skill to get you into doing what I do. | |
| DT: | It sounds like for a patent attorney, that’s the secret sauce or the superpower, the passion for explaining the technical side, right? Because you need to be able to understand and then interpret and then explain. And I think that’s the piece that a lot of people, both on the technical side and the legal side as well. Alana, thanks so much for joining us today on Hearsay. | |
| 00:51:41 | RD | As always, you’ve been listening to Hearsay the Legal Podcast. I’d like to thank today’s guest, Alana Hannah, for being a part of it. As you well know, if you’re an Australian legal practitioner, you can claim one Continuing Professional Development point for listening to this episode. Whether an activity entitles you to claim a CPD unit is self-assessed, but we suggest this episode entitles you to claim a substantive law unit. More information on claiming and tracking your points on Hearsay can be found on our website. Hearsay the Legal Podcast is, as always, brought to you by Lext Australia, a legal innovation company that makes the law easier to access and easier to practice, and that includes your CPD. Hearsay is recorded on the lands of the Gadigal People of the Eora nation and we would like to pay our respects to elders past and present. Thanks for listening and see you all on the next episode of Hearsay! |
You must be a subscriber to access this content.