Unknown Speaker 0:00 All right. Good afternoon and welcome to today's webinar, artificial intelligence and intellectual property presented by Autumn. My name is Sammy Spiegel, autumns professional development manager and I will be your staff host for today. All lines have been muted to ensure high quality audio and today's session is being recorded. If you have questions for our presenter, we encourage you to use the q&a feature rather than the chat feature. If you have a technical question or comment, please feel free to use the chat. Closed captioning is available and turned on. But if it is distraction to you and you'd like to hide the captions, you can click on the closed caption live transcript button on your toolbar. I would like to take a moment to acknowledge and thank autumns 2021. Online Professional Development sponsor, we appreciate your ongoing support. Before I'd officially begin today's session, and I introduce you to our presenter, I would like to grab at get to know our audience a little bit better. So if you could please take a moment to complete the following quick poll that should display in just a second. We would appreciate it Unknown Speaker 1:16 keep this open for just a few more seconds. Unknown Speaker 1:41 All righty, thank you so much for helping us get a feel for who's joining us today. So with that, I would like to officially welcome today's presenter Ben Cullis kowski. Ben is a patent attorney and partner with Thrive IP intellectual property law firm. After graduating from the US Naval Academy and serving as a naval flight officer in the US Navy. He became an engineering center manager working with electro optics and laser radar systems for chips in our aircraft, and also worked as a program manager for computer systems company serving the US government and eventually became a patent attorney and serves a variety of clients in the software, AI, electromechanical medical device and other technical art areas. His practice includes preparation and prosecution of patent, trademark and copyright applications, opinion work and IP litigation in state and federal courts, as well as TTB and CBAP. Practice. Please join me in welcoming them. I will let you take over control. Unknown Speaker 2:47 Thank you. Thank you, Sammy. Appreciate autumn allowing me to talk today about artificial intelligence and intellectual property. And special thanks to Sammy for helping me with the logistics of today's webinar. And she already introduced me so I won't, I won't talk about my favorite subject me. But there's a little about me and Unknown Speaker 3:13 just concluded. I can't see your we can't see your slides yet. Oh, I'm sorry. That is my fault. Okay, can you see me? Okay, here we go. Well, Unknown Speaker 3:28 that's the glitch that I was worried about. And it's out of the way. So I just finished a case in the Eastern District of Virginia where we were litigating patent or copyright infringement and then a variety of other issues. So I do work both sides of the coin, if you will prosecute the patents and copyright and trademark applications and also litigate them either asserting the intellectual property rights for clients or defending them as the case may be. So my main goal today is I believe I'm speaking to from the poll looks like a lot of smarter people than I am and but I believe in most of the audience's technology transfer professionals. So what I hope to do is give you some food for thought and hopefully make your everyday profession a little easier as you deal with perhaps, invention disclosures from professors and universities or whatever your clientele might be. And also, as you deal with perhaps outside counsel, if you're in the university setting, you may, you may be sending venture disclosures to outside firms. And this will give you a little more background about maybe questions to be asking those folks as you as you deal with them on a day to day basis. I also understand that some of you may be intimately involved with artificial intelligence Since already, and then there may be others that have limited knowledge, or maybe just things you see in passing on the news or in a movie. So with that in mind, I want to have a baseline and ask the question or answer the question, what is artificial intelligence? Which may seem an easy answer, because I think most of us probably immediately think of Arnold Schwarzenegger and the Terminator when you think about artificial intelligence. But for a working definition for today's presentation, I'll try to address that question. And then once we have that working definition, we can talk about AI relative to patents, as well as to copyrights. And, and again, as I mentioned, some responsibilities that we might want to keep in mind as we delve into the subject matter. So what is artificial intelligence, I like to think of it as a synthetic entity. So any kind of human created robot computer, maybe the Terminator, but it could be as simple as a handheld calculator. So it's going to receive data. And it's going to solve problems and give us some kind of a conclusion or result. And in the advanced AI that we'll talk about, it may even seem as if it's acting like a human being. And one definition, as I kind of touched on already, is that AI is the universe of computing technology, which is too broad it that could be again, even the old pantile calculator, right, but two plus two, and I get four, that tells me four is the answer. And that's not really what we're going to be talking about today, what I want to focus on, I want to narrow down to two forms of artificial intelligence, one is machine learning. And the other is deep learning or deep neural network learning. And as an intro very quickly, machine learning is highly data dependent data, data data, and deep learning is going to be more in the order of the autonomous, self aware type of artificial intelligence that we might see in the science fiction movie. And just to break that down a little further, machine learning and deep learning have three layers. So deep learning, or excuse me, machine learning is going to have a data entry or input layer. And then there's going to be the middle layer, that secret sauce layer, I'd like to call it, which basically is going to be an algorithm. And then then there's going to be the output layer, what we see as the result of what the algorithm does with the data. Deep learning is pretty much the same thing except the middle layer, or it's actually multiple layers, multiple algorithms, if you will. And what happens in deep learning or deep neural network learning is the algorithms are going to be talking to each other and learning from each other. So you may the the AI may be programmed to receive data from a human being, or it may gather the data itself, maybe through an electro optic AI or whatever. And then it's going to take that data, and one algorithm is going to do some initial processing, and then it's going to go through more and more algorithms. And perhaps so the six algorithm comes up with an interim conclusion or result. And then that gets fed back in sort of a loop to maybe the second algorithm that does some further refining. And it may, it's going to catch its own errors and do some other processing. So there's a forward and backward learning that's going on. And as it says at the bottom of each of the slide here, the problem not the problem. That's just the nature of machine learning as it requires human supervision, and data, good data at whereas deep learning is going to be more of the autonomous type of artificial intelligence that we're we see in science fiction movies, machine learning is it sort of says there is going to be a garbage in garbage out type of artificial intelligence. So perhaps, let's say the artificial intelligence is supposed to be an economic forecasting model. And I give it economic data from Brazil in 1973. That's not going to the results of that is not going to be very applicable to the United States in 2021. So you can see it's, it's highly dependent on the data. And then as demographics change, perhaps, perhaps the machine learning AI has something to do with demographics. If I'm not updating the algorithm. Maybe I'm focused on something to do with the 20 Something generation and I and they're moving to the Midwest. majority of them are moving to the Midwest for some reason. My data needs to keep up with that in order for it to be accurate. So there's a lot of hand holding with machine learning. And keep those two in mind as we go along machine learning and deep learning. So Pretty, we're probably familiar most of us with some real world examples. And again, depending on the source, you look at different layers of AI. Some some sources would say that my first example or natural language processing is actually another form of AI separate from machine learning and deep learning but, but for our purposes, it's an example of, of machine learning AI. And that would be Siri, where's the best restaurant in town. And it tells me go to Tommy's pizza joint on Main Street. And once again, if, if the AI has not been updated with the fact that Tommy's is no longer in business or moved location, it's going to be the conclusion is not going to be very useful. And in this case, Siri has been trained in my example, with the English language. So the language the English language is part of the data input. And the GPS coordinates of the restaurant is another set of data and so on, so forth. Speech recognition or cousin to natural language processing. Again, it's, you know, we've been using voice enabled text for a few years now, so that we don't have to text and drive. It recognizes my English language, and can text my wife on the way home that I'm going to stop at the store. Do we need anything spam filters, as much as I enjoy getting those emails from the Eastern European prince with a billion dollars that he needs to meet a handle. The spam filter is another simple form of having some algorithms that recognize junk email and filters it out set out to deal with it. recommendation engines based on my browsing history, and movies, I like it can tell me what what movies suggest to me what movie I might like next, I always joke and say that's big brother watching me. But it is very useful to have those types of AI in my life. And my favorite example, today is my Roomba vacuum cleaner that I got for Christmas. This is a great example of machine learning AI. I have a hairy dog that sheds tremendously. And this little robot for those of you might have one, plug it in and we own it first takes off when it's first run, it's going to bump into obstacles, and it's learning the footprint of the house. It's going to find out where the furniture is and the doorways are and it's going to record that data through trial and error to learn the most efficient route. And just for fun, sometimes I'll put a chair in the middle of the room and see what see what it'll do. And just fun with AI. So those are practical examples today. And what I want to show here is some more cutting edge technology, more cutting edge AI, if you will, and then I'm gonna turn it over to Sammy to show us Mr. Einstein briefly. Unknown Speaker 13:13 The common language of science to work language, rush to link acoustical acoustically Otterbox Commuter behind to sense impression. Unknown Speaker 13:31 And Sam, if you want to, I think maybe just go ahead and show the person does not exist slide. Unknown Speaker 13:45 Okay, and I believe we'll see a pop up. Down at the bottom right of the screen, you see this generative adversarial network explanation, you're welcome to click on that link later, it's your own. But you can go and click on it, if you want to. So that tells you a little more about what this deep neural network is doing. And let's go back to there and get back to that slide. So this person does not exist. And at this point, let's go ahead and say me back to my presentation. And I'll talk a little more about both of these things that we've just seen. Do I need to share my screen? Yes, please. Okay. Unknown Speaker 14:40 All right. Do you see me Okay. Yep, you're good. All right. So the, you might say, what does this have to do? What's practical about this? What? What we want to take away from this slide is that this is more like the IBM Blue chess, artificial intelligence. So we probably grew up hearing about, it may not be a practical application for it today, like the Roomba vacuum cleaner. But what this is showing is the advances in artificial intelligence. So for instance, with Mr. Einstein, what we saw in that little snippet was a two dimensional photograph that the artificial intelligence had taken as a piece of data. And then using the English language, it was able to mimic the movements of the of the mouth, the eyes, the cheekbones, such as the human face speaks. And in that case, like I said, it was a two dimensional photograph, I'm not sure if the audio was Mr. Einstein's, I believe he passed away in the 50s, I think there were there was audio maybe video of him. But that was a, that was an AI construct that we saw. Similarly, with a generative adversarial network. What we saw there was a what's called resolution convergence. And I don't want to go too deep into the mechanics of artificial intelligence. But essentially, what it's doing in that case is it's it's taking a generator, kind of that first layer we talked about, and a discriminator, which is a middle layer, and it's going back and forth, literally billions of pieces of data about what a human being looks like, billions of 1000s and 1000s of typical human being photographs and what the AI believes, a human would look like. And it starts out, in, for instance, on a generator and with a blank canvas, perhaps four tones of colors for shading. And it tells the discriminator Hey, what is uh, what would the area of right cheekbone look like? And it goes back and forth. And it, it basically is carving a sculpture from a block of stone, I think Michelangelo said, how do you how do you do that, and he says, I just chipped away, all of a sudden, to show what's underneath. And that's what's going on generative adversarial network. Artificial Intelligence is it's going back and forth, carving away, and it's presenting a person that looks like a person, but these people do not exist. They're they're an avatar for what to say, I think so human being would look like. Again, the point of this is, perhaps we can be used for this AI to further develop to ultimately reach that autonomous thinking that we talked about. It there may be practical aspects here, maybe surgical reconstruction aspects, and so on, and so forth. And then finally, I didn't open it. But the bottom of slide is meant to show that artificial intelligence can tie together, what we call the Internet of Things, multiple devices doing different things. And in this case, it's an example of artificial intelligence, moving the point of sale to where you are, especially in these days of zoom and such an ordering from Amazon, you can probably relate to this one. But there's a lot going on behind this with the artificial intelligence taking various disparate sources of data. And in the example I use in this case is, I could literally be on a run out in the country. And if I'm using my Fitbit watch, or some other smart device, they can tell me, Hey, Ben, it's time to order a new pair of running shoes, and I can literally order it out there next to the cow pasture. And it can be waiting for mail and doorstep when they get home. So that's just another practical application of artificial intelligence. So that leads to again, working definition of AI, machine learning and deep learning. Can artificial intelligence be an inventor? So we're going to enter the patent world now and hopefully, for those dealing with inventors, professors and like bringing you disclosures, you can look at these from a perspective of is artificial intelligence involved? Or behind the invention? So the question is, Can AI itself be an inventor? And I know this is a little bit of a busy slide. This comes from a question and answer Request for Comments from the patent office from the US Patent Office. It's actually coming up on about two years old, but the report itself was from last October. And the question essentially was what I just said, can I be an inventor in the patent office? Obviously, this is just a snippet from about a two page answer, but they cited the US Code the patent law 35 USC 101, a one. And they said, essentially based on code, and based on the Federal Circuit case law, you've got to be a natural person to be an n Venter. And people have tried to test the waters, just within the last year or so. A company tried to file a patent application with its AI called dabbas device for the autonomous bootstrapping of unified sentience. They tried to list Babus as the inventor on an application, the title of it is at the top of the slide there. And it's kind of funny, they are funny from a pattern point of view, they listed dabbas as the given name. And they listed the family name as an invention generated by artificial intelligence. And the patent office kicked that back and said, You can't list dabbas as an inventor, because it's not a natural person. And it went back and forth. This was this was really not so much on the merits of the invention. It was really the the, the administrative or technical requirement that you need to be a natural person and the people that had filed the application, got into a fight with the patent office, if you will. And the patent office got a little snarky, and basically, the argument they made essentially was dabbas has been, has been patented. And therefore it ought to be an inventor, I didn't quite understand the logical connection there. But the patent office came back and said, Look, just because just because the machine has been patented, doesn't mean it can be an inventor, any more than getting a patent on the camera allows the camera to hold the copyright. And anyway, that application got kicked out. The other question is Will Okay, so AI cannot be an inventor, but can it self be patented? And I sort of already gave you the answer with dabbas. The dabbas example is that, yes, AI can be patented. And I want to give you a little bit of an intro, I don't want to spend too much time on this. But those of you that have dealt with computer related inventions as I have in the last few years, the part of the patent code says you can get 35 USC 101 that you can get a patent, or at least it's patentable subject matter on a process, some machine an article of manufacture, a composition of matter, or an improvement on any of those things. However, judicial exceptions, things that are not patentable subject matter on it on their face, include this list, and you see the second bullet there is mere algorithms wherever we heard that before. And we saw that in the deep learning and machine learning. That's the That's the secret sauce. That's the guts of AI or algorithms. So that that's sort of foreboding, when we talk about can AI be patented? And so the quick answer, there's various ways to address this. And I should say that this area of patent law has become very complex, difficult, in some ways, because the Supreme Court who are not generally enough, I don't think there is a patent attorney on the court. They they kind of muddied the water up quite a bit about six years ago with a case called Alice. And then the progeny of cases that have come out of that, since then, have the courts have been wrestling with that judicial exception, whether an algorithm whether an abstract idea can be patented, and it's basically been all over the place. But what you might ask for an event or your professor that comes to you with an invention on artificial intelligence is, hey, does your is your algorithm doing something significantly more? Is it? Is it solving a technological problem, as opposed to just being an algorithm or a mathematical formula? You're going to be looking for a solution to a technological problem. And you're going to want to see significantly more than just the mathematical equation. And what do we mean by that? Here's an example. Again, answering the question, can it be patented? And the answer is yes, that this was an example. And here's another specific example, just from April, just a couple of months ago, we see at&t received a patent for adaptive beam sweeping for 5g or other next generation network. And in this patent, they talk about and describe and claim artificial intelligence and artificial intelligent component 208. Unknown Speaker 24:45 And in the patent application, they they define the artificial intelligence as machine learning. And we've heard that here just earlier today, and what is happening in their application. Well, it's now Patent is they're doing this adaptive beam sweeping, which is basically signal collection so that we can use our cell phones better and perhaps radar at airports and that sort of thing. And But importantly, they say that machine learning and AI are used interchangeably. So keep that in mind. And then they go on to explain that there are radio access network intelligent controller for Rick, which they give an element number 202, includes the AI component 208 that we saw on the drawing. And they basically say that, that the ric 200 can perform its operations using the AI component. Now, what does all this have to do with what I said about significantly more, so they're not claiming the algorithm per se, what they claim and here's an example claim, they say that using an about middle way down there, they use a, they generate a machine learn distribution model. And remember, they defined machine learning as AI. So you might read that as they generate an AI distribution model. And they use that AI distribution model to result in a what they call a beam sweeping control parameter. So the question is, is that allowable and patentable? And the answer is, yes, there's the patent number. But again, the way it was technically described and claimed, they showed that they were using the AI and in a significant way to help do this beam sweeping. What about AI as a patent tool? So again, we just talked about patenting the AI, now we're talking about, can we use the AI to help with the patent process, if you will? And you recall that report from the patent office in October? And they answered this question as well. And they basically said, the use of AI by a natural person to develop an invention is okay. And that makes sense. If you think about it, I mean, an inventor might be using a welding tool to bend a piece of metal, or I might be using a word processing program to type up the patent application. So these are all kinds of tools. And so it makes sense that you could perhaps use artificial intelligence as a tool. One of the search one of the tools and again, this is maybe for your everyday work life, it might help you to understand this, perhaps when you're talking to your, your outside counsel, or if you're using a search firm yourself, you might ask the question, are they using AI as a search tool? Now, I, when I'm asked to do a search for a patent client, I use what's called the cooperative Patent Classification methodology at the patent office. It's a, what they call art units at the patent office. And it it's as the slide indicates, there's hundreds of 1000s of symbols, and art units 1000s of those. And it can be, it can be complex to do a search using the scheme, although it's fairly accurate. But what the AI can do is not only does it understand the CPC scheme, it's also being updated as the CPC schema changes, which it does, and it uses word relation methodologies. And what that means is, if anybody has ever most of us have looked at a patent application, sometimes we'll see coined terms that the inventor has created themselves. And so technology is changing. Words are changing new. New Words are evolving. And the patent office calls that the applicant as lexicographer that can be complicated doing a search because somebody may be using a new word that they've coined that might be related to something from the 1980s. The AI can relate those words. And that's as simple as that's the simple way to think of this. And then they the AI may be able to do the search better than yours truly, or perhaps a search company. And also it's using image searches. So that's also helpful because design patents are heavily reliant on the figures. And just because you're searching a keyword, you may not be seeing an image. And then the the key bullet is that it's going to make intelligent decisions. It's going to call results. So just real simple, let's say your patent your invent should say, is a new kind of bed. And so the AI is going to look for rectangular surfaces that you might lie down on. And it's but it's going to perhaps kick out irregular surfaces. And so it's going to make a an assumption that, you know, a bumpy cubicle surfaces Not a bit. And again, that's a garbage out garbage in thing because if it if it, if it kicks out the wrong kind of data, you might not have the best results. So anyway, AI can be used as a search tool. And I'll revisit this briefly. What about as a patent application drafting tool. And, of course, a lot of US Patent attorneys, you know, we can do our breath, and oh, my gosh, we're going to be replaced. In this example, I won't click on this one, you're welcome to look at this later at your leisure. But this company called Open AI has a language generator called GPT. Three, and it provided the title, the author's name, and just the word it and a six page document was generated. And if you read it, it actually tells a story. And it kind of makes some sense. So it's pretty remarkable that just based on the title and the word, it generated a six page document. So the question is, well, can I generate a patent application? And I have actually seen a result of, of an AI program generating a patent application? And if it can do that, can it generate an opinion tool? So for those of you in technology transfer arena, you might ask your outside counsel to do a patentability search and opinion or clearance search? Or maybe you're in a situation where you need a non infringement opinion, can I do that, and again, it's a data data reliant tool, if you provide it. Your invention disclosure, if you tell it the current case, law, and if you describe the prior art that you're aware of, perhaps from another AI source, then then theoretically, it perhaps could draft an opinion for you is the invention patentable. And you thought that kind of data could do that. For those of us that deal on the litigation side, anybody that's been involved in this even from from tangentially, you understand that discovery and patent litigation particularly can be overbearing, shall we say? 1000s and 1000s, of documents, I was climbing through storage sheds, when I first became an attorney as a as a young associate, and I was it was it was storage rooms and Alabama and summer digging through boxes and boxes of documents. Not fun. So what the what this company called logical, you gotta give them credit for a fanciful mark there. They, they use intuitive filters and keywords and, and as they say, and highlighted point there, they identify privileged documents automatically. So they take a large amount of data, and they can at least call it down to some relevant bite sized amount of data. And that's just one company there's, there's others out there. I'm not I don't know much about this company other than what you see there. But that can be very helpful to streamline the discovery process. So AI is being used in that case as well. Unknown Speaker 33:48 All right, so we've learned that AI can be patented. And it can also be used as as various types of patent process tools. Now the question is can it be protected can AI be protected by copyright and maybe he should have said this already, but very briefly, a patent protects I touched on already protects maybe an article a thing or a method or a process. Copyright is going to protect the expression of an idea so all of us are familiar with books or movies or songs that that probably have copyright a copyright registration attached to them. I'd like to give the example of painting I'm there's there are mountains around my area. If I'm standing there with a with a canvas on sketching and charcoal, the charcoal image of the camp of the mountain and somebody's next to me doing watercolors, we've both expressed the idea in a different way. I did it with watercolor, and he did it with charcoal or whichever was so that but we don't have them out. itself, the mountain is the idea, we don't own the mountain, we own the expression that we created of that mountain. So just like a professor coming to you with a dissertation, or he wants to publish a paper, that would be an expression of, of the idea that he has written in his paper, and it potentially could be copyrightable subject matter. And that's what this says. So number one is a literary work, for instance, literary works also cover potentially the algorithm, or the code, the computer code, if you will. And as you might already be imagining, so the AI might be protectable by patent and copyright, depending on the aspects that you're trying to cover. So these are things you might want to think about in your day to day lives is when someone when a professor inventor comes to you with a work is a patentable, and or is subject to copyright. And kind of like the patent situation, okay, we can perhaps copyright the AI itself, but can AI create the copyright? Or can I create copyrighted material. And there's really no clear definition in the copyright code that I'm aware of that says, No AI cannot be cannot create copyright. But what the Copyright Office has done is it's looked to the code like the patent office has done. And it basically points to various section of the code. And it says, You got to be a natural person. And section 302 says, the wife of the author and seven years after the author's death, so AI does not have a life and it doesn't have a depth in human sense of the matter. So the copyright office says you got to be a natural person, much like in the patent office, saying you have to be a natural person. Some of you are familiar with a few years back, this is a very famous example Copyright Office dealt with this issue of, if you will, about AI, years before the patent office did. But in this case, those may remember this a macaque monkey grab that photographers camera, snapped a bunch of selfies, including this one. And the photographer tried to get a copyright registration on this picture, and he tried to get folks to not use it. And the Copyright Office basically said, sorry, the monkey took the picture, the monkey is the author, if you will, and created the photograph. And because it's the monkey is not a person, you can't it's an eligible for copyright registration. So the answer is AI cannot be the author of a work has to be a natural person. This is just interesting to express what's going on in other parts of the world. In the UK, as far back as 19, Ada, their copyright act does recognize the creation of computer generated work without a human author. And recent more recently, in 2017, the European Parliament was advocating for electronic persons, that is still not solidified in that in those areas. But it seems as if it might be moving in our direction. But currently the the controlling case in the United States anyways, this Feist case of the top which basically says, again, you have to have human intention to have creativity. And it's not in the hypothetical I give, there's, what if AI, what if I create AI, and train it to paint pictures, maybe that's the third layer output is supposed to create classical paintings based on learning what classical painting looks like, and it creates the new Mona Lisa, that in and of itself would not be copyrightable as we just explored. But what if I'm using what if I'm the artist and I'm using AI as a tool and again, that in that case, I wouldn't be the author and in using the AI as a tool would be okay. Kind of like using paint or paint program on the computer might be the best example of that. Unknown Speaker 39:35 So as your folks come to you, the professors and inventors, one thing to keep in mind, a practical takeaway that some are not aware of, is once they say the professor's dissertation is published. In we'll just say within three months of that publication beyond three months of that publication, you're probably going to if there ever As infringement, you're not going to be able to seek statutory damages. So the point of this slide is to say, get the work, have the copyright application filed, or have the patent application filed, if that's also applicable before any kind of publication because in order to assert damages or seek damages, I should say, statutory damages, that's that's going to be the limiting factor. I guess, to put this in more practical terms. If there's no real damages, for instance, the professors, paper gets published, and somebody copies the paper. But they don't really make any money off of copying the paper, it's the juice may not be worth the squeeze as far as suing that infringer, because there's no real damages. And if it's post three months publication, you won't be able to seek the statutory damages. So the statutory damages are built into the law to encourage early, early registration. Just check the clock here on time. So the attorney responsibilities I mentioned, I think this would be helpful for especially those that are involved in the tech transfer, to understand what their attorneys are doing so that you can help manage these things, or at least ask the right questions. And the duty of confidentiality. This in the reason I'm bringing this up is if you're going to use artificial intelligence, for instance, as a patent, patentability search tool, and you're farming out the information, or I should say, let's say your outside counsels farming that information out to a third party to do the search, which is using an AI search tool, the legal rules for attorneys is that the client needs to give informed consent. So what I'm trying to say in this slide to you is that you need to know what your attorneys doing with your confidential invention information. And if they're, if they're sending out the information to a third party search firm you, you need to consent to that you need to be aware of that. So just something to think about. Similarly, the duty of supervision requires that the the lawyer that outside counsel you may be using or perhaps you have inside counsel, or you are the attorney within your university. What this is telling us is that the lawyer needs to be able to supervise that, for instance, suit the third party search firm, and you need to give it direction about what to search for a good example would be, let's say you the university client wants to do a search a worldwide search because you might want to get patent protection not only in the United States, but perhaps in Europe, and the attorney sons out there with the information to the third party search firm. And the third party search firms AI is only looking at South American patent databases. That's not going that's not going to give you the good results you want. You want to know what's going on in Europe and the United States and perhaps worldwide. So the attorney needs to be supervising what the third party search firms doing, and making sure they're using the proper data. competency. This requires expertise. And I'm talking to Sammy, before this presentation attorneys most of us in most states, we have to attend continuing legal education classes, and particularly in artificial intelligence world as things evolve quickly. And in very complex ways. We if we're going to write a patent application for this very highly technical engineering field, we need to keep abreast of changes as the as the slide talks about. So you need to ask your outside counsel, you know, how are they keeping up in the field, especially with artificial intelligence Are they are they understand the latest and greatest changes in the law and in the technical area? And finally, for anybody that is a patent attorney that had a little heart palpitation about patent patent applications being written by artificial intelligence, if you recall, the GPT three that wrote that six page paper based on a title and the word it if you read it, take a second read through it, you realize it's kind of gibberish. doesn't make a lot of sense. And so even if I use AI to generate a patent application or an opinion, I'm not going to put my name on that, send it out to a client, because I'd be sued for malpractice, I have to read it, I have to make sure that it is not producing gibberish, that it's used the right data that the algorithms are up to date, and it's doing what I'm expecting it to do. So independent judgment is something you also need your attorneys to be doing. You don't want them rubber stamping, artificial intelligence generated document, because it may be, like I said, the garbage in garbage out type of document. And this is really what that is expressing here is that if the algorithms are wrong, if my economic model is based on the 1973, Brazil economy, and I'm trying to apply to the 2021, US it, it's going to be useless. And so what this is saying is that, again, that duty of independent judgment competence, I need to make sure that the model the AI algorithms are weighted correctly and, and using the right and have been trained correctly. Using the right data. Here's an example of not necessarily would be a situation with your university being sued, but perhaps your law firm if you have outside counsel, this was an example of blockchain technology and blockchain technology, if you're not familiar with that is it's literally chains of data blocks of data. And the idea behind this, which is heavily AI dependent is the millions of users around the world, the data, and so you cannot change. Even if one person trust to go in and say that they own a million shares of 10 million shares. All the other chains of data owned by the people have to agree with that AI, various nodes obey, I have to agree that that that new data entry is correct. And if they don't have a majority that agrees around the world, and they kick out that that is perhaps fraudulent. So in this case, what happened was, there were five patent applications that this particular firm, and I'm not beating up on this firm, this is in the news. And I don't know the results in this case, but this happened just in March, they they were sued by the client, rocket fuel sued this firm, because they basically said that the firm did not, did not understand the technology and did not understand the AI behind the technology. And, and the five patent applications were worthless. So that is an example of failing to understand the AI ramifications and things going incorrectly. Hopefully, that won't happen in your university settings. So the takeaway it for me, and I think for you, as well, as you're dealing with AI related inventions are using AI to help with a patent application or copyright applications. Can you explain it? You understand the AI, at least on a general level? And do you understand if they is using up to date realistic data? Can it be validated? Do you understand the meaning of the results? If if they if your outside counsel presents to you that hey, we found no prior art on your professors invention. You know, do you understand what that means that they look in the right places? Or were they looking at the wrong technology area? Was the AI looking at the wrong technology? Unknown Speaker 48:58 And I think most importantly, I would just jump down to the last bullet. Well, second to the last is or am I using the AI this court my work product and conclusions? Or am I just using it to generate my work product without giving it any kind of supervision or even understanding it? And I already mentioned you know being being cognizant of public disclosure dates because of if nothing else, the statutory damages that might be lost. And for patent applications. I didn't mention this earlier, but those of you that deal in the field understand that there would be a disclosure in the United States. You still have a year from a public disclosure, but you might have forfeited your chance of getting that application abroad. And I know professors I've dealt with many of them they want to get their papers published and journals and such but we want to make sure that we you know, get get the protections in place. afford the for publication. And that is, that is my presentation, I hope you've taken at least some tools, understand that AI can be patented, can be copyrightable subject matter if it's appropriately prepared. And it can also be a very useful tool and to help you with those processes. And perhaps these are questions you can be aware of as you work with your inventors and professors. And with that, I'll turn it back to Sammy. Unknown Speaker 50:36 And thank you so much for such an informative discussion today. I know I took away a lot, I don't see any questions added into the q&a feature just yet. So we'll give our attendees a couple moments to add any questions in if you would like to do that for Ben. But Ben, and that last slide, really, I think, captured what I'm going to ask you. But any other like key takeaways that you want to make sure that our attendees, don't forget, as they're kind of walking back into their offices this afternoon, next week beyond that, that'll really help them shape what they're doing in the office day to day. Unknown Speaker 51:15 Um, I think the the main, the main thing I would say is that as long as they're working with their professors, or inventors, of course, as I mentioned, make sure they're not publicly disclosing these things beforehand. But But if, if the inventor comes to you with what appears to be an algorithm on its face, ask them, you know, what, what problem is that algorithm going to be helpful to solve? And there may be an invention there that, that maybe the professor doesn't even see. And so those are, that's the question I would ask is, if they've if they've come up with an algorithm or an AI tool, you know, is it patentable? Is it copyrightable because this is intellectual property and property has value? And perhaps if we can get it protected, then we can license it and do all those good things? That's, that's what I would do. That would be the takeaway. Unknown Speaker 52:18 Excellent, thank you. And it looks like we did have a question come in, says the data that is being utilized in the algorithm, there would be is there would there be an ownership issue there for the data being used in the algorithms? Unknown Speaker 52:32 Yeah, very good question. So it depends on the source. It depends on the data I hate to give the attorney answer of it depends. If it's, you know, if it's public information that's being collected? Probably not. That's a pretty vague answer. I understand. If I'm, if I'm taking the data from a company, and I'm using taking very broadly, there could be a trade secret issue, if the company that I'm getting the data from has taken steps to kind of keep it a secret, but yet I've somehow gathered that data to feed my algorithms that that might be a trade secret issue, which we didn't, we didn't touch on. But yeah, it depends. Unknown Speaker 53:17 I don't have to be a follow up session for that. Yeah, trade details, part two. Unknown Speaker 53:22 One other thought I meant to touch on it. But so back when we were talking about what's going on in Europe, for instance, there, there is the question of will US patent laws, will US copyright law changed to eventually allow artificial intelligence to be an inventor or an author about copyrightable work? And right now, I think the answer is no. I think the fear from most of the professional organizations that I'm aware of is that if AI starts becoming an inventor, for instance, and I'm not saying this is right or wrong, I'm just saying this is how it's viewed by some is eventually AI could crowd out the human mind and start becoming the painter and sculptor and the inventor and there won't be anything left for humans to protect. So right now, it's it's a great cocktail conversation, a question, but I don't think Congress has the I don't think there's a big groundswell to change the copyright and patent law at this time, if anybody was interested in that. Unknown Speaker 54:33 Excellent. Well, with that, it looks like we have all of our answer questions submitted. So it looks like we'll be able to give folks a couple minutes back today. On behalf of autumn I would like to thank you, Ben for such an informative discussion and thank all of our attendees for joining today. And as a reminder, a recording of the webinar will be available for viewing in the autumn Learning Center within a few days of this session and is included with your registration fee, and you can visit the auto website to sign into the Learning Center, view the recording or purchase any past webinars that you may have missed. And don't forget to complete the webinar evaluation which will pop up when you close out of this webinar to help us serve your needs in the future. Thank you again so much for joining us. And thank you again, Ben, for your insights and I hope everyone has a great rest of their day. Transcribed by https://otter.ai