Transcription provided by Huntsville AI Transcribe
Yep. Okay, good deal. So welcome. First off, we’re thankful for Hudson Alpha for letting us use the space. It’s seriously appreciated along with their nice technology that lets us connect over Zoom. So tonight we’ve got Sam Landrood from Wimbledon, Dixon, Dickinson. I keep messing that name up in case you didn’t know. US LLC. No, not LLC.
It’s LLP. LLP Professions.
Oh, Professions.
Okay.
So, this discussion tonight is about the impact of law on AI.
I think we’ll look at doing one in the future with that opposite.
What is the impact of AI on law?
However, you know, but for now we’re looking at… Yes, it does.
It’s a big topic. It’s scary. So that, I’m going to hand it over to Andrew and then I will push buttons on the thing, the slides that you’ve got. We’ll cover that. Cool. Sounds like a good point. We appreciate you all letting us crash your party today. This is a cool group. I’m glad everybody has stayed together and kept meeting. Appreciate the chance to come talk about some of the legal aspects, I guess, of what goes on. I am Andrew Toggle and that most of you have made the first time tonight, so that’s great. I’m an attorney in town at Wimbledon, Dickinson.
I do a lot of local property, a lot of international trade, some cybersecurity issues that we handle. But I was not always that.
I started my professional life as an engineer for Western Digital, hard drive manufacturer. I’m doing modeling and simulation of nanospill. And so I try to stay involved in some of the technology and this is one of the reasons that I came to Homesville because there’s a lot of tech here.
A lot of smart people building cool and so I like to just be involved and see how that goes. My colleague Sam, I’ll let him introduce himself. I’m Sam and I also work for Wombledon. Dickinson, I haven’t had any meetings so I purchased now. So it’s very nice to meet everybody. I’m a patent attorney.
So that’s mostly all I’ve done in my legal career before that.
I was an electrical engineer at Hewlett Packard and then after the split, Hewlett Packard Enterprise. I worked in a server store, networking group and mostly doing software for more delivery. And so I’m a software.
And so now as an attorney, I decided to do patent wall because it’s a good opportunity to look about what are two new technology because people come to me with ideas and I’m able to translate that into a document that protects those ideas.
So it’s really, really interesting to get to learn a lot.
I’ve been doing a lot of work on AI lately.
It’s really big.
Now, AI has been around for a long time, but it’s picked up a lot of steam recently.
I’ve been on the admin of ChatGPT and it’s more mainstream now, I guess so to speak, but AI has been around for a while. Well, AI is a property.
It’s a much bigger deal all the time. It’s always important to you protect your ideas. Yeah, there’s a lot there too. But yeah, nice to meet everybody. Cool. So this is why we’re here tonight.
I just kind of gave you a briefing of some things that we’re going to talk about but look, y’all really in a way are the experts on the AI part and all we can do is come and offer some legal perspective. So we’re here as much to learn from you and kind of get to understand the way you think about these issues as a practitioner. We’ll start just kind of with a whirlwind tour of a few news stories that you probably have seen that we’re going to put some legal things together all that you can understand how lawyers think about it and how that might be different from how practitioners think about some of these things. We’ll talk a little bit about some of the new regulations or say legal regimes that people are using to try to get their hands around artificial intelligence. And now I’m going to turn over to Sam to talk more in depth about some specific IP issues that he is saying and working closely with related to patents related to licensing issues or something like that. So that’s kind of the idea. I’ll talk in a general way and I’m going to let the expert talk about the hard-eyed stuff. So if you go down probably two slides now. Here we are.
So this is just a news story that is very recent and something that I had been paying attention to because it involves a judge that I used to work with and he was called to the bench. Ken Ellusam is a circuit court judge and a village judge here in Alabama and he recently made headlines because he wrote an opinion in which he described how he had used chat GPT and other large language models to decide a case. And this is kind of an interesting claim. So let me tell you the facts of the case.
It’s a very boring run on the old case that happens every day. Have you all seen these trampolines that are in-ground?
Absolutely.
They scare me. But I guess some people like them. And I guess they’re better than having their trampoline up in the air maybe. Well the guy was hired to install one of these. Now in Birmingham he installed this exactly one that you’re looking at in the in-ground trampoline in some people’s backyard. And he was a landscaper. So they also hired him to do you know the rose bushes and thread wall etc. But part of his project was to install one of these. You do that in kind of the way that you would guess. You dig a big hole, you put kind of a retain wall and then you put in the trampoline and you see this wooden decorative cover around the edge on the side. Well it’s not a shock to me. But maybe someone was surprised. This girl was jumping on the trampoline. She fell. She hit her face on the wooden part and was pretty badly injured. So they soon got put in the trampoline. This happens like I said all the time. Somebody does some work. Somebody gets her suit and the person who did the work. Usually not a problem because almost all of these landscapers are insured. And so things happened and their insurance picks up the bill. He sent the wall suit to the insurer and they sat on it for a few weeks and then they sent back to him. We’re not going to cover this. Because installing this trampoline is not landscaping. Well he was pretty mad. So he suited the insurer. So the question before the court boils down to is the installation of an underground trampoline landscaping? Okay. Anybody guess what the trial law court decided?
I’m going to go with yes. I think it is landscaping. The insurer should have to pay. That’s probably what I would say to do. In particular, he signed as part of his application that he would not install any recreational equipment. And so the insurer said well that means that this is not part of landscaping we don’t have to pay. And that’s how the trial court decided it. Which basically… I think it would be exercise not recreation. I think it’s a similar kind of question. Is it recreational?
Anyway this got this decorative elements etc.
Well anyway they filled it up to Judge Nussim who is the basis here. Judge Nussim is a really, really smart guy. I never forget the first time I met him in Birmingham in the store.
He was an attorney. He carried around an iPad everywhere. He just constantly carried it around. He also wanted to show people new cool stuff that he could do on his iPad. He was a big believer that lawyers shouldn’t use paper and boards. Sort of like a tech forward. When he got this question is the installation of trampoline landscape he decided that he was going to ask chatGbt. Because he thought this is a question about what is language usual to me and who has read more language than chatGbt. So you scroll down to the next slide and you’ll see this kind of excerpt from his opinion that he wrote. He mentions that a lot of people will think this is heresy but that we should consider whether AI powered LLMS might inform interpretive analysis. Having thought the unthinkable I’ve said the unsayable. I’m going to say not landscape. Not landscape. I’m going to say not landscape. I’m going to say landscape because it’s told it in the land. I think that’s very smart. Is that the way LLMS thinks though?
They will get it in the context that it’s typically used.
That’s the decision. That’s true.
I would have said landscape because they built the wall.
Just to say it, how could you look at that design and not see a problem coming?
I don’t know.
That’s fine.
I’m sorry. I gave this thing that you jubbed on and it’s surrounded by a portal. It’s coming down at a rapid velocity.
I agree 100%.
Judge Newsom asked which agreed that yes this is landscape.
He gave us some explanation. He did a smart thing. At the end of his opinion which was a concurring opinion. He added at the end of his opinion. Everyone should do this. The output from all the prompts he had put in.
The output he had gotten and the different LLMS he had posed these prompts to.
This is one of his key points.
Judges should use these things but should use them in very transparent ways. I thought that was interesting. I’m going to share that.
It’s pretty deep. I think this is the kind of thing where an LLMS can help us.
The question is what is language?
We’ve seen different pieces.
Especially with different models trained on different datasets.
Traditionally the one on Reddit.
I just found out that there was a few of harassment clips. Oh my gosh.
I was going to ask if he had tried too different.
He did. There was a third girl. He’s got them up there.
A lot of these are from major players. I would expect most of those to give you similar answers.
They did.
They all agreed. I thought again that gave us some confidence. That’s the most confidence. I’m not trying to be dumb as an after-gift here. That.
What did it spin out as an answer?
What did it say?
One step past that is answering what chronic overboard. I think that’s important. It would be easy. It would be easy as a judge to write a few of these opinions where you’re just using this as an assistant. The last ten times the AI has agreed with what you think he would have come to anyway. It would be easy to become mentally lazy.
There are a lot of stories right now in ethics and AI. It’s not necessarily like deep learning language. What’s it called when you get out of jail? Redis-Ress is something. They’re models on that. Judges have been using to make decisions on do I parole somebody or do I not? The question is you can only make a model of that on people that have already been in prison. There’s no way to model that on people that haven’t been. One of the things that you can get into, if you cover this on IP and data sets or anything like this, each of these models is trained on a certain data set of data that most of these won’t release what data sets they used.
The typical thing that we used to go back to is initially a lot of large language models started off with Wikipedia.
What percentage of Wikipedia did you write it from then?
A large percentage.
What percentage of those are white men?
If you want a white male perspective of language, go look at Wikipedia. It’s just a thing. You need to be transparent and understand where it’s going. Most of these have been basically scraping everything on the internet at this point. They had to save room with face recognition. I’ve got one for you.
Face recognition. That’s how they determine your feed. You’re being fed a fly. This happened at the exact same time.
There was a historical case, one of the email I’ve heard about where an AI model made the racist accusation, the racist thing of comparing a white person to a gorilla. We’ve talked about that here. It’s like the inherent stupidity that comes out of these models.
I had Mike Tyson doing his workout videos, preparing for the Jake Paul fight, and gorillas wrestling at the same time as the white feed. Both of them popped up at the same time. There was one where I think it was the ACOU. It was used by certain police forces to look through photos of people and figure out which ones were on their list. It flied several members of Congress. Everyone that got flied had some things in common. They were in the same class. They were non-traditional minorities. They were trying to strike folks just because of what they were trying to do. I know you got a lot of slides. This is moving.
That was kind of the decision I thought.
Another aspect of AI that has some legal code to it is what does it mean to be an author or a creator? I’m going to pass some examples here.
Anything I would use?
Yes.
At least to say the first draft or so. In some context, we don’t allow it.
Because it’s probably what you may be getting into. I think this is interesting. The question has become artistic works. Or even not artist technical works that have been created using artificial intelligence tools like wet rights. Either AI have in it or even a human. These beautiful pictures that they make. Or poems that they write. I want to look at some very old cases.
Where the copyright office had to answer this question in a new way.
This is a photograph of Oscar Wilde. He’s an author.
Someone took a photograph of him and tried to register a copyright on that photograph. It was challenged because it was something that was made by O’Shaen. Someone challenged his ability to get a copyright on this photograph. The copyright office decided she can get a copyright on a photograph. That’s good. People do it all the time today. It’s all over the internet. The reason they said that he began it is because the photographer has to tell the camera what to do. The camera can’t do anything on its own. You have to position the scene, position the camera. At the time you had to do the development work. That was an early question about machine created art. Let’s look at the terms we used to think of. It’s interesting.
The same thing happened when you went from black and white to color.
The next thing happened from film to digital.
It’s interesting.
You can go back and read Ansel Adams’ biography. Kodak was sending him film facts.
He would go shoot something with Polaroid to give a quick look at it. Now I know if I take this shot from this angle because I did it on a quick thing.
That same thing is happening now.
Chris Wade is a local artist in town. His painting is a great artist. I’d love to look him up. He does a lot of things like landscape with rockets and parts.
The mixture of nature and machine. It looks really surreal.
The same kind of surreal that some of these generated things.
He’s running into parts problems with that kind of thing.
I can almost guarantee you that some people look at his actual art that paints. It’s something that got generated.
It’s a little different than it’s not realistic. This is a very different subject.
What is art?
If Ansel Williams takes a picture, that’s art. He’s focusing on a certain thing he wants. If I’m down with the Grand Canyon, pull out my self-portrait. Is that art? Sure it is. I don’t want you to make that picture.
This book was written a long time ago.
More than 100 years ago. This book was written by international beings that spoke the book to a woman who lived in America. That’s exactly how she filled out her copyright application.
Which was denied.
Because she had not identified a human author with the work. She didn’t change her application. She said no, I wrote it. She stuck with this idea that she had channeled us from these interdimensional beings. She never got a copyright on this book. One of the reasons I can show you the cover of today.
It would have lapsed by now. There’s another example where the copyright office did the silent.
You’ve got to have a human involved.
There’s no human involved in creating the work. There’s no copyright. Click down. You remember these tasks?
They used to be my profile picture.
I wouldn’t recommend that.
I don’t recommend that. I’m not going to show you a picture of myself.
This photographer was in Indonesia. The monkey stole his camera.
He took a bunch of pictures of them. He wanted to get a copyright on them. There were funny pictures. Did you remember where the copyright office was?
I don’t know.
This is very different from the Oscar Wilde flow.
The monkey is not a human.
The monkey could get it. The monkey could get it. You need to copyright as soon as you make the work. The question is about registration.
The copyright is something that you have to make a work. Whether you publish it or not. Whether you register it or not. I’m going to know where some nuance is. That’s the basic idea. This is very different from the Oscar Wilde flow.
Let’s talk about some AI things.
Did you recognize these pictures?
This one on the left has the qualities you were talking about.
It’s realistic. It was generated by mid-journey. Pretty sure.
The guy who prompted the journey.
He submitted this to a contest.
He also tried to register a copyright for it. He explained he had used mid-journey. He had not done it himself. The copyright office said… Does he actually do two other work? That’s what they said.
There’s some appeals.
Currently, there’s no copyright registration. I think it’s called entrance to paradise.
I knew the guy that availed it. He submitted. He’s still a lover.
Not just copyright, but for patents. He’s really pushing the limits. He must have some backing from somewhere. Finally, those are not cheap. This is hotbed. This one on the left, Zarya of the Dawn. Was a graphic novel.
If you look at the top left, you see number one. You see a volume. Then you see Kastanova, the author.
Then you see Midjer.
The reason is because Midjer did all of the pictures. She wrote the text. She spent lots and lots of time giving prompts. She took a long time. This was one of the early versions of Midjer. She would get the picture and the mouth would be weird.
She would use Photoshop to fix all of that. She submitted it to the copyright office. She went in and physically. They did say yes. She didn’t say anything about Midjer. The office saw news reports. She talked through her old process. They revoked her copyright. Initially. They did something interesting.
Here’s what they came down to.
She wrote the other. She got a copyright on the selection and arrangement of the pictures. How they were on the page. Not on any of the images themselves. She as a human made the images. She made a few corrections. She should have made the images. She made a photo. That’s a good strategy. When you get ready to do that. I’m going to find a backer.
Exactly. There’s been several lawsuits about who wrote the song.
It might be the lyrics or the melody.
You guys have heard of a little lady called Taylor Swift. Yes. She recorded a whole bunch of songs. That were early in her career. Someone bought the rights to those songs. She got really nifty about that. She went back and recorded all of those songs. She started working on the song. She had the rights for the first song. She owns the song. She owns the song. She owns the recording. She owns the recording. That’s what Prince changed his name. He used to use the song. He used to use the song. All of those songs are the same. They were all different. It’s a different thing. I don’t know if I can record a live concert. I don’t know if I can keep up with the live concerts. I don’t know if I can keep up with the live concerts. If two people listen to one person, they can’t tell the difference. That’s what I’m trying to say. It’s related to another A.I. fan. She did it for two originals. She did it for two reasons. The company paid the studio. The company paid the producers. The company paid everything. That’s normal when you’re trying to get off the ground. You can’t fund your own thing. You go to a recording studio or record label and they own what you do. You can’t take a song. Frank Sinopter sings the song. Willow Nelson sings the song. It’s very different. There’s cases too.
You can use an impersonator.
You can record a radio jingle that sounds just like Tom Wood. There’s cases about that.
They used A.I. to have a voice that sounds just like him. I don’t think I’ve done these slides in here about that. That’s another aspect.
I said, tell us.
He used his style. He made it to sound. He put something he played next to like a hundred others. I did it a little bit. It’s got Harley Davis and Sue. They started to sound like a harmon. That was a big deal. It was a sound. It’s the 45 degrees separation of the crack shot. It’s the land on the crank that does it. The movement of the pulses.
It’s kind of rhythmic, but it’s not like a it’s more like a galloping sound. Harley Davis’ fire could say bang bang, bang bang, whereas most people bang, bang, bang, bang bang. That was some of the topics. I’m going to pass it over to Sam now. We can talk about any of this. I want him to be able to talk about some of the stuff.
He’ll be a real expert on you. Let me know when to stop. I’m going to be over it with data now. Data is generated no matter what you do. You can get a copyright on a database. It’s a lot harder to force. You can go that route. You can also keep it as a trade secret.
There’s particular laws and nuances keeping it secret.
You can monetize it. There’s a lot of things to look out for when you’re trying to license data.
We can talk about that briefly. Things to consider.
How do you want your licensing to be used? Are you using it for training machine learning model to recognize certain images? What about royalty payments? How long do you want the term of the contract?
How long do you want someone to use it?
How are they using it? Restricting users and even restrictive reverse. This is just some considerations. If you go to the next slide, you can think about how you’re going to access the data.
You can also download the data on your systems. Will the data be displayed in some way? Are you going to control that?
For example, via a dashboard.
How is it going to be distributed?
The big one is derivative works. That’s probably overlooked sometimes. If you decide to do a licensed deal, I’m trying to look out for derivative works. If they had extra data, they manipulated it in some way. That’s always a big question. Whether it’s going to be exclusive or not. Some brief questions. If you have any questions, we can get them. I know the two, not necessarily data licenses, but one of the licenses for using certain models.
One of the big ones was OpenAI.
Their models initially had a piece in the license that said you couldn’t use it to build weapon systems or anything to do with weapon systems. That kind of restricted a lot of what goes on around here. Microsoft got a little more involved in the board for OpenAI. I’m not sure if they forced it or what.
That restriction is no longer part of the OpenAI license.
Did they say what the reason for the initial restriction was?
Because they didn’t want to be associated with it? They were worried about liability or something like that. I’m not quite sure why. They just didn’t want to be associated with it. I think that’s more online. Some of the things they initially did when they released GPT-2 and made a big splash in an article saying it’s too dangerous. We can’t release this. People were wondering if you had to get more publicity. You really don’t want to release it. You want to reproduce the same type of thing. By that time it’s out of the bag and they want it released anyway.
You’ve got other countries that don’t follow the same rules. Are they the same ethics or anything? Is there a balance point between the patent on a technical function versus the copyright of the code that the back ended? I do a lot of software applications.
I never put code in a software patent application.
You can’t protect it. That’s the route I would go. If anybody were to ask me what I would recommend, I would recommend copyrighted code.
There’s some utility there. That’s a good balance. Everybody is always concerned about submitting their code to the copyright office.
There’s some kind of way you can submit.
You can get a copyright registration without submitting all the code.
It’s not accessible.
I haven’t done a lot of copyright work. The first 15 lines and the last 15 lines have to be visible. With DEMOX, that change in the code refutes the point of having a copyrighted code.
Eventually you get to a release version and maybe you want to copyright that.
You don’t have to. There has to be security. You have to make sure not everybody has access. It’s got to be more confidential. That’s another route. Copyrights have to protect them. There’s just one way to do it. One of the other model license things we ran into probably last year, I’m not sure when this was. The the confusion or something along that lines had a license to use it. It said you will always take the available attempt to stay up to the latest version. It’s free at the time. They changed it to a paid model. You were using it under the license. It’s a quick to bus. It sounds like they tried to build in some opens where style language to you.
They got flame banting for that.
I think massive of people went away from it. I don’t know if they want to change it back or not. When the elastic did have some claim, they were trying to keep all their stuff from being used on the AWS server. They changed it in license terms. Everybody took it under the new license terms. They came and did it with all the windows. They changed it to a free license to a paid license if your business makes X number. We stopped using it. I can’t tell people to do that. You have your customers from bigger than you. That’s the problem. They used the tool. They took one of the different registry. They have something everybody was using. They had to change it underneath you. The question is if you on that line, if I download a model and get the license that went with the model at the time that I’ve been using it, and now this is the version I have, if they update their license, am I required to step up to their new one if I’m still using the version I had and with the license that came with it? It depends on what’s in the original license. It depends on what’s in the original license.
It depends on the terms.
It’s really dependent.
It depends. It depends. Sounds like an AI engine.
That’s data.
This was pretty brief.
There’s a lot more about this. There’s almost a field for data pedigree.
As far as what data sets went into the model, what license was used for those data sets, where did they get their stuff from?
Can the lomerations and the copulations of data set on data set on data set and one of the initial parts of using co-pilot with GitHub or Visual Studio Code when it first came out.
It’s one that really helps you write code a whole lot faster. I love the thing. But I can’t use it on my company work because I don’t know what code was used to train the model.
I don’t know what license was on the code that was used to train the model, and I can’t put that into my code and then have somebody else through me. Because the thing that happened to get spit out matches was that they had copyrighted it. I would hope that they would have some kind of warranty. And they would cover you maybe like limitation of liability or something to that degree. That’s what you got a little bit of work. If they do that, that’s fine. I think GitHub actually got sued recently. It was a while back. It was a case because it basically scraped everybody’s code and trained their models. That’s also why I love these models. They don’t tell me what age source they used because they would open themselves up to that kind of thing.
Open AI got sued too by New York Times. I think that case is still ongoing too. We’ll see where that goes because they used basically every new source that they could get. So they don’t want paywalls anyway.
Someone has to decide which is which.
What bothers me about judges is suppose they are really unfamiliar with the thing that you’re working with. How can they make any intelligent or correct decisions about them? Think about it. You have 12 tourists here.
That’s a good answer. Are you serving on your own, Juris? Yeah, well a lot of normal people. There are 300 beers, right? I’m going to get that here. I imagine it’s the same on the legal side as what I run into on the defense side. I imagine it’s the same on the legal side as what I run into on the defense side. I have new technology, new things. I’m trying to get approval to get this new thing put into a lab. I’m trying to get approval. It happens to me by people who have no idea what I’m talking about. We run into that all the time. If you really get taken involved, the best thing you can spend money on is an attorney that can make a judge feel like she understands what’s going on. Often they still don’t, but you’ve got to make them feel like they did. The other side of this from a cyber perspective is if I know what code base was used to train your model. Oh yeah. And then if I know what kind of vulnerabilities there may be in the code that was used to train the model that is now generating code.
I can guess that a vulnerability that was used in data set might exist in this code base because copy base almost.
It’s a huge issue.
Everybody wants to require S-bombs and software that will tear the track where all your components came from. And we can do it the same way.
When people come to us with open source licensing, we use this software to come in and find out what code is here and what licenses do we need to care about.
What you’re talking about makes all those tools much more difficult to use because you don’t know where these things came from.
So I’m going to turn off the microsoft case versus the patent squire.
This was before GPT. To me GPT is kind of a steam roll problem.
But before GPT came out you had a voice credit permission software similar to Amazon.
So it was Alexa.
Echo voice assistance.
It was an open source voice specifically for winners. You could write a lot of the skills like what you could write for Amazon’s Alexa. There was a patent squire that tried to get around them or pull their work into his work. I forget the details of the case. But it’s online if you want to see it.
I don’t think it was entirely a state to you. So international law coming to play with something like that. What is the water we’ve worked?
Just stick in the blender and get front panning it. There you go. We didn’t even touch on any of the A4 controllers.
That’s what comes for the day.
I don’t know the commerce part. I’m going to see what else you get. We can talk a little bit about patentability if you like. I think a more interesting question is can AI be an adventure?
That’s been a question that’s been floated by Thale.
I think it got answered. What if AI helps with an invention? What if they allow the patent application?
You can be live on it. You can go to jail. Somebody finds that. You sign a declaration.
I certify that this is an invention.
I wouldn’t recommend that. You get that kind of thing. You also got models like Alpha Hold. It’s actually in use trying to find new drug combinations.
They take kind of things because they can figure out I can do this this way and it matches these rules.
They’ve actually found some breakthrough things there. They’ve tightened those. If you want we can talk a little bit about patentability. AI and software are patent applications on software.
One of the cases that came out around 2012-2014 was Alice Beameo. They were basically on what is an abstract idea and what’s actually pathable. They really hit the software. The financial, say, patents, or when do I trade it? Stop, right? Those were the three areas that really took a hit after those cases came out. You can still get software patents. It’s a little bit more difficult than you can do it. You just have to have more details. You’ve got to make sure you build your details under your specification. Maybe you can tie in some physical devices that’s always helpful.
Maybe you talk about solving a specific technical problem.
You can get it even if it’s IBM and they’re developing this model that’s potential patent. I’m sure there’s art out there on that. It’s not the model.
The model would be patentable, I’m pretty sure, but the output of it.
That’s a problem, right?
That’s what I’m wondering. There’s a lot of money going on. I think it would be patentable. The answer would be no. If AI created the new drive, I think they would give it the case.
I think they would be the case. There’s just so much more money to go behind the truck. The reason the patent is something is that you have the rights to it, but also from the government perspective, it’s a way to get ideas and inventions in the open. Yeah, absolutely. You just don’t like it. You just don’t open it. How are you going to do that? You’ve got to submit to the FDA. You’ve got to open it. I don’t know if people would prescribe it if they didn’t know what it was. That’s a tough question. I don’t think they necessarily answered that. Maybe an interesting dive into… I’ve never told anybody. I just thought your roads are different.
I don’t know. I don’t know either, but for us, the problem is jump off patents. What if you fire half-air engineers, and they know what you’re doing double time? That’s not what you’re doing now. Maybe there are ideas. It’s the rest that you take as a corporation. I worry about something similar to what you’re talking about with music, where how do you create a new rhythm or a new theme that’s not already out there and not already done. Software is similar.
I’ve been doing it for the last 15 years.
I probably couldn’t find some other product somewhere. If somebody happened to have a patent for something I was doing, that could come after me for whatever, I didn’t know what to see if the hello world button or whatever is a patent somewhere. Because why don’t we think about it?
You can get patents, but maybe it’s not always smart to get a patent. It is easier to keep it as a trade secret sometimes. Build those walls around the idea to make sure that it doesn’t get out. A corporation hires people and people leave. It’s really tough to keep those walls up. I think most of what is, I can’t remember the last one where somebody had done something and they got sued for patent infringement. I thought that was common knowledge. That’s another thing I wanted to say. One thing that I always used to tell, when I was in house I worked with software engineers and they were in dirt with us, we’ve done this or somebody’s done this before. We don’t think this is really patentable.
We’re not in the right position to make that decision. We’ll make that decision. We’ll look at what’s out there. And we’ll let you know if it’s possible or not. So what does it hurt at that point?
Working for a giant corporation? The worst we would say is we’re not going to file it.
The other thing you can do is come back and say, I don’t want to do it exactly this way. So at least you know that maybe it worked around.
At the time it was digital reverb. I don’t, PV, put the first digital reverb chip into an amplifier. And reverb has been around for a long time. You know, they were trying to figure out if I do it this way. I mean there’s patents out there for how they do it. You know, is it a whole effect?
There’s all kind of stuff there. And that was like, it’s reverb. This is not a, I mean this is a physical thing in a room right now. There’s reverb right now with me talking. Is that somebody else? It just got out. But using some chip to create that effect? Maybe it was, yeah. Maybe it wasn’t the time. But yeah, maybe at the time you probably could have gotten that. They were worried more about it. I’m not going to get sued by somebody that already has a patent. And figure out what, you know, it’s pretty hot. And again, that’s where patent attorneys can help, right?
They can do landscape searches and let you know.
So this is just starting with the background.
And this is very legal. Focus, you may or may not be interested in this, but like I mentioned, software is basically patentable. What determines section 101 of the United States Code, right?
And basically tells you what can be patented, right?
And you are a useful process.
Machine or manufacturer, it doesn’t matter.
And so like I mentioned, Mayo and Alice, those cases really changed.
It really changed everything.
Diagnostic, medical diagnostics are still very difficult to get. And you have to have, you really have to set it up to where there’s some result and you’re doing something based on that.
To actually get a patent on that space now.
So just straight diagnostic claims.
So you’re not going to get through.
So first, you can get it. Business methods are very difficult as well. So if you include terms like cost and, you know, stocks or, you know, any kind of financial terms in an application, it’s going to, it’s going to raise that flag and you’re going to get it with a 101 rejection.
Okay.
Yeah.
One of the things that we were looking at on a different project I was working was we had found a new way to detect whether a kid had been left in a car seat in a hot car.
You know, that kind of thing.
We were actually working that direction. And then we went to look through a patent search and figure out, wait a minute, there’s already like 35 patents and stuff. Most of these are not held by manufacturers or anything else. They’re held by patent holding companies. There’s one reason that people do that.
And that turned that project from a technical project into a legal thing that’s not fun for, it’s fun for people, but not… It’s not for me. It’s not too much fun for you. Right. And it just turned into a, well, this is a useful activity. And if you wonder why that’s not really available on these things now, that might be part of it.
But that’s when we learned about the finding one that’s close to what you’re doing and actually extend the previous one, what they knew, you would change or a new twist or something and the patent, your new thing. I didn’t know that was a… Yeah, incremental also.
Patent, there’s tons of patents. There’s so many patents now. And people are filing up all the time. IBM falls tons of patents all the time. You know, these really large tech companies fall just so many patents. So a lot of them are just incremental changes.
They’ll just be, they’ll change something slightly and that will be the patent. But then you only get lots to the change. That’s right. Yeah. So you can’t necessarily use the thing that you changed, but no one else can use that change. Your little changes. Oh, that’s horrible. Pat, putting your zip code in, finding the location and control going around making money on it. Oh my gosh.
Well, and you know, and that’s where I think, you know, people push back I think maybe, maybe they can invalidate that patent honestly.
But sometimes it’s just easier just to pay the money and then make them go away.
Because that’s where you have to go for five years ago. That’s where my friends actually went into. Oh, okay.
To where there’s a clothing line out of Lawrence part of them.
That’s kind of the thing. But you can put your, they have changed with your zip code and they find it. Well, we found them and they wanted to pay for it and they just taken that system off. So they paid in and took their system down. Usually you pay and the wife’s going to keep doing what you were doing. He sued them for doing that. So they paid and they did that differently. Here’s our location. Don’t put your name on it.
But what it gave me was that’s how the Ocelot office was doing it. Well, the in-suffer is super post office. I’m sure the government would fight to some degree and imagine.
And it’s also dribbling now. Sure, drive, pee or drive or put your zip code in. Use your information to know the stuff where you’re at. So they buy the license to out them. They would have bought it. They wouldn’t have licensed it. But I think it’s very picky to where he has it, buddy. As you should say, he has donated a lot to the city. I’m glad he’s using his powers for good. I’ll say to your point, when you mentioned the child kind of detection, I think in one way to drive change it’s maybe somebody publicizing it’s like, hey, there’s all these patents. This kid was left in the car last week.
Just as a hypothetical, right? And maybe if this was made possible to the public, then maybe people would actually let that happen. They would have a free license to use that technology. So that doesn’t happen anymore. Well, that was one issue.
The other issue that we wound up stopping because of half the people we surveyed said they would never consider themselves as being somebody who would leave a kid in the car and would never buy this. So there’s four million kids born in the US every year. Half of these will never be purchased. So your market cap if you have the whole market is too many units of you.
And then you look at what you might get on a margin for that and you’re so far down.
That’s not the way to sell that product.
By lobbying the Department of Transportation to require these kinds of services. They’re probably going to cease the whole thing. That’s how you get it. Yeah, I agree. But that turned it into a not-fundable. Back to the whole… To worry you did the same thing that made it back to everyone after. Say that again? That the Lorian did the same thing. Back to the future? It wasn’t so much about his car. It was the way that he just wanted to make it. I’ve been touch with his daughter Kat and she’s getting her own version of an automotive company on Instagram. She’s got a lot of his whole documentation. She’s a chip off the old block. She was in a lot of the documentation and a lot of the press material. She was the voice of the old stage as a kid selling these cars because Lorian was autistic and she is too. But he had and I gave my opinion on it. It looked like the plan that he was trying to implement. It was almost like your first version of hacking or social engineering. Kind of like the model of Linux, an open source kind of thing where you could use different parts. That’s why you have some different parts from other manufacturers.
I’m a Bowerian.
Pinto Borex.
Pugio Transmission Dimension. I did my nose.
Trying to stay focused. He was like a little bit of a problem too. He invented airbags and he pushed for the requirements of seat belts. Seat belts were first fully implemented by manufacturer by fellow. He rarely went out of his way. He was the father of the muscle car. He wouldn’t have muscle cars without John DeWorrie. That was his latest thing in the 60s. He had so many ideas to try to get out there. He did try to go down the path of lobbying with the Barron Transportation and HTSA and all that stuff. But they just found another way to come after it. Let’s keep moving. There’s a few more slides. You can skip this one.
There’s just types of software that you can’t get patented on. This is AI.
You definitely can. There’s at least a few different aspects that you can patent.
For example, your training set.
When I say that I don’t necessarily need the data itself, but maybe how you’re pre-processing your data.
Maybe you’re doing it in a newer unique way. You want to do that. Maybe you’re generating your training set in a different way.
Maybe that’s something you want to protect.
These are areas you can protect on that end. The model itself, you can actually protect new models that you develop. We actually put in some examples of claims that we’ve worked on for other companies.
One of them was on a new model itself, which you can see on the left hand side.
It’s what they call a single index model tree.
It was a tree with different index models that did an edge tree node of the tree.
That’s pure software, by the way.
There’s nothing physical about this at all.
This was actually in the financial space too. You’re going to get a little one from every perspective here. We got this one through. I was actually really happy with this one.
Then on the right you can see you’ve got the diagnostics.
This is more on the outcome.
On the previous slide in the last box was the output of the model.
That’s the end result.
What you’re trying to get or do, that’s something that you can patent as well. For example, you have some health care system that takes S&P profiles and tells you based on that whether you’re at risk for a certain disease or whether you shouldn’t be prescribed some opioid. That’s the diagnostic part.
You’ve got to do something. In response to that, we don’t prescribe good opioid.
That’s how you get that one.
You have to have something right. Some next action. You can get that. Then IT threat detection.
This is a little bit longer of a claim.
They use AI to do that. These are all published, patented applications. They’re all publicly available. There’s many more examples that I can give you.
You can just search in Google Pads.
For example, you can search for machine learning. A lot of things will pop up.
If you ever decide you want to get a patent on any of this stuff, you want to do this. This is something the attorney who most likely would hope. Think about ways that you have success rate. That’s making sure that you’ve got a really high level of detail in your specification.
A big one to me is including what the problem is. I just solved that particular problem. Then pinpointing a technology is also helpful. It’s not always possible. Kisafir can apply a lot of different things.
If you’re solving a problem, I think that’s really big.
If you point to a specific technology, that’s also really helpful too. For AI, on the next slide, it’s basically the same. You’re looking at the same rules.
If you ever submit a disclosure for AI to an attorney, I would recommend finding one of the examples.
I think that’s going to be real key in the future too.
Examples are going to be really big. I’m just going to focus on sufficiency of description. We’re just going to use machine learning. They don’t really know what they’re doing. They’re just saying we’re going to use it. You need more than that. You can’t just say you can do this. That’s it. More detail together for any kind of disclosure. Again, that’s a strong. There’s some cases that you don’t have to go through this. I don’t know how much time we have.
We normally end about 15 minutes ago.
This is on AI.
There we go.
Only because of AI.
That’s fine.
There’s a lot you can unpack here for sure. You want me to jump back to? Let me get back over and check. The first thing I would do is stop recording.