197 - How AI Legal Battles Are Reshaping the Tech Landscape with Elizabeth Knopf

0:00There was that big announcement with Johnny Ive and Sam Altman. They had fun little pictures together. And then things went dark. Websites started getting scrubbed and everyone was like, what is going on? And it came out that they have been sued. People have sort of been distrustful of Sam Altman. Then about a week later, they announced this marriage between Johnny Ive and OpenAI. It doesn't smell so good on the surface.

0:28Liz, full docket today. We're going to talk about potentially the marriage of Johnny Ive and Sam Altman already going through a divorce. We've got some updates on Google's DeepMind, as well as potentially some fair use rulings for Claude and Meta. And then in the second half of the conversation, you're going to walk me, an idiot, through the process of artifacts in Cloud. You know those things where when you're like working within a Cloud chat and you actually create a document or a dashboard or, and then you have to go back and think like, oh, where was that thing? Oh, did I save that to my desktop?

1:02Okay. Well, Cloud has now made that way easier for you to go back and look at the artifacts that you created and generated. And you are going to walk us through how to navigate the new system because I'm stupid and I need to actually see how it's done. So I'm excited. That's all going to happen today. But first, what's going on with IO? So all of a sudden, about a little bit over a week ago, there was obviously that big announcement with Johnny Ive and Sam Altman. They had fun little pictures together. And then things went dark. The website started getting scrubbed and everyone was like, what is going on?

1:34And it came out that they have been sued by a company called I.O., not I.O., but I.Y.O. and apparently it's over the trademark of it, but it sounds like there's a lot more underneath that's a bit fishier and sketchy. So people have sort of been distrustful of Sam Altman. I don't know. Everyone have your opinion of it. But basically what happened was apparently this company IYO that was a spin out of Google's X Lab or something, they basically have developed these really interesting wearable AirBud technology, and were trying to get acquired by OpenAI. And then about a week later, they announced this marriage between Johnny Ive and OpenAI.

2:24So apparently there's been data that was shared. Who knows? We'll probably find out more in the coming weeks. But it doesn't smell so good on the surface. And IO, was this what they were calling the new product or this new division within open AI? Yes. Yeah, it does not look good. Looking through this, IYO, I-Y-O, like you said, was spun out from Google X. It was one of their projects that they were working on. They've raised over$60 million. They met with Johnny Ive and Sam Altman twice potentially to acquire them. I don't know under what circumstances. And then they decided to get together, obviously, Sam Altman and Johnny Ive.

3:07They launched, it's called IYO, letters IO as opposed to IYO. And now video has been scrubbed. A lot of information has been scrubbed online. It doesn't look good. What do you think the odds are, the chances are this gets unwound? I mean, I think it'll be fine, honestly. I don't know how much cash they have in the bank because they've been spending from what I understand. But I mean, they're doing fine from at least a revenue standpoint. So I think they'll figure out and navigate how to manage. It seems to be a trademark issue. You know, there are plenty of, I'm not a lawyer, but I think they'll figure out a way to navigate it so that they can create a wearable.

3:44Now, again, I think it's like if there was some interesting IP they were going to leverage that in actuality, there you go. But again, it seems like they'll either pay them off in a settlement or they'll figure out like how to create a hole so that it doesn't actually conflict. There will be some kind of payoff on this, but... It does feel like Sam Altman is just continuing to craft this image of ruthless Machiavellian leadership. Yeah. Again, these are just facts right now. Elon Musk is disputing the relationship that he had with Sam Altman and the understanding they had of what OpenAI was going to be.

4:27He's now turning it from a nonprofit into a for-profit. but there's lots of questions around that. He's also gotten into a tiff with IO, right? It's like he goes and gets Johnny Ive and now there's this weird relationship where it's like, wait a second, did he steal their IP? What's happening here? I think Mark Andreessen was the venture capitalist who said basically if Sam Altman was left on an island with everybody else in Silicon Valley, he would become the Lord of the Flies within a year. He would be the leader of that island because he understands how to work the room and the politics behind it.

5:01So I don't know, just all these little data points, Sam very much plays this like, oh, you know, we're just, it's going to be world changing technology and kumbaya. And behind the scenes, he's a killer. I mean, he's, I don't mean like an actual killer, but he's like a, he's got the Mamba mentality. You know, he's going out and executing. Yeah, no, absolutely. I mean, and that brings up another challenge that they've had with another acquisition that they were trying to do with Winsurf, which a couple of months back, this was sort of the competitor to Cursor which is how they were basically coding with AI.

5:34And they were going to acquire it for over$3 billion. And then Claude basically cut off their data to Windsurf. And it's sort of, there are some rumors, it's unclear if that's going to continue. I'm sure the acquisition will continue, but it was basically in response to a lot of the stuff that Anthropic, Claude, I'll use those interchangeably, are doing because they are really going aggressive on code and starting to be more focused in that regard. And they have been sort of praised as a leader. I mean, some people will argue 03 is better, whatever. But even last week, a couple weeks ago, Claude came out with something called Claude Code and really sort of moved the needle.

6:12So it's basically like these wars that are going on between them around coding. And then so after Claude cut off Windsurf from their data, because it was basically seen as windsurf was going to be utilizing that for open ai and doing some training off of it and obviously anthropic doesn't trust open ai so it's unclear now how things are going to play out yeah so it's like it's worse basically right now between the we already don't trust people with the data and now it's like oh also you have this supercomputer yes yeah sure take my data and use the supercomputer to plot against me it's just freaking nuts in other supercomputer news actually, before we get to that story, something I'm noticing is I don't hear as much about AGI anymore.

6:58For those who don't know, AGI stands for artificial general intelligence, which is effectively intelligence that's on the same level or above that of humans. And they used to talk about it a lot, like, oh, it's around the corner. AGI is coming. And I just don't feel like I've heard it. Now it's sort of changed to this new term, supercomputing, or I can't remember what meta's new division is called, but - Superintelligence. New market. Super intelligence? Yes. That's more just an observation. I think it's interesting how it's kind of like shifted and they're admitting AGI is not around the corner.

7:28I think there's a, like, just looking at where we are now versus a year ago, I mean, the level of sort of agentic capability, meaning that AI can go and do tasks on its own. It's now incorporated into a lot of tools, but you still have to sort of guide it. And so, however, people, I don't even know if there was a clear definition to say like the yes or the no, it was created. But yeah, I mean, it seems like there's more optimism in that humans are somewhat important. But I think also the value of what was historically created with code and what AI can now sort of take over from a task perspective, like the value there is getting crushed.

8:05So, I mean, it's hard to predict. I would be foolish to say I have any sort of clarity on this because nobody does. And anyone that sort of like suggests that they do have that view, maybe some of the cool sci-fi people do. But I don't know where it's going to go. But yeah, they're, I think, just remarketing things, whether or not they just need a new term to get people excited or this sort of moving benchmark is just, you know, unclear of what it actually means. To your point, yeah, they're either realizing it's harder than it should be or than they said it would be or they're rebranding it. Yeah, it's one of the two things.

8:39That leads us, though, into the next discussion that we highlighted in the beginning, which is this fair use and what Anthropic is going through. Yep. So I'll highlight really quickly what fair use is and traditionally what it's been, and then we can talk about what it means for Anthropic. Hey, I don't know if you remember this, but when we started this podcast, we entered into a social contract. I would spend time, energy, and money producing this podcast, interviewing these individuals and giving you insights into how to build, buy, start, grow your business. And you would like, subscribe, and leave me five-star review.

9:13Now, out of that, we both get to talk to really cool people and hear really cool insights. We both get a ton of value. But I just want to help you keep your word. So would you do me a favor? Will you go leave a five-star review for me on Apple or Spotify? It would really help. and if you want, even share this with a friend. Fair use effectively means if I publish something and I have a copyright or a trademark or I've created a piece of content or media, somebody else can't just take it, repost it and do whatever they want with it and then generate dollars or make money off of it. That wouldn't be okay.

9:47Let's say that you wrote The Grapes of Wrath and I ended up taking it, self-publishing it and then selling it on my own website, The Grapes of Wrath. It's like, well, Nick, you didn't write The Grapes of Wrath. Why are you selling it? I can't do that. What fair use does, though, is it allows you to take a work and make it your own, make it unique. So if Liz posts a video about AI and I take a 60 second clip of it and I show her talking and then I do a talking head response to it, that can be considered fair use. Or if I take a clip of a show and I maybe put myself on a green screen behind it, that might be fair use.

10:20Or if I buy a copyrighted thing and then give attribution, that might be fair use. So fair use has sort of evolved as the internet age has come about. But now there's a ton of other fair use questions because these models are using copyrighted materials to train. So how are the owners of that intellectual capital compensated? And there's a lot of questions. And so Anthropic right now is in the middle of a lawsuit regarding fair use. Yep. Over the last week, there's actually been two sort of pivotal results. So both Anthropic as well as Meta. So essentially, they were training on publicly accessible data sources.

10:57Meta was using books to transform their LLAMA model. And they were, again, trying to prove this fair use context. And then Anthropic also was using copyrighted books. But essentially, there was, I guess, a term called exceedingly transformative that was enabling them to sort of get around this. So as they say, good artists, what, copy and then steal. And so this is essentially what we're seeing is that by combining all these different data sources, whether it's the, I guess, Anthropoc, something called the LibGen Library, it has 7 million pirated books. And so they were sort of saying, well, these are pirated and this and that.

11:39But apparently, again, this just is enabling them to utilize publicly information. And so basically, as they say, you know, if you don't, well, I guess anything you put out there is being utilized now by the algorithms and AI. So if you want to keep something private, don't share it. It's a very weird thing because they can go and they can crawl and they can train all their models on this stuff. And then there's no attribution for it. I don't know. I'm not smart enough to have an opinion on it. But I do think that for us, as we're looking at it, there will be cascading effects down the road. I mean, Hollywood in the form of entertainment is going to change.

12:17I mean, just look at what it did to newspapers in the first iteration of the internet, right? It's probably going to do to entertainment in the second iteration of the internet. But anyways. Well, actually on that note, there was a statistic that came out this last week saying that even the traffic to news sites because of AI, not even because of social media, but because of AI has gone down by like 23 or 20 some percent over the last two years. So that is having a huge impact. Yeah. So. Let's get to the meat and potatoes because I'm excited to really look at this. I want to know about Claude's artifacts because I use Claude a lot.

12:57I use it for very specific things. I love the way that they have projects, but this issue of artifacts has been a consistent an issue for me. And for people who are listening, think of an artifact just effectively as something that you created. It could be a template. It could be a document. It could be a dashboard. It could be an Excel spreadsheet. Anything that you used AI to help you create in any time, we're going to just call that an artifact. We're not saying document or spreadsheet or whatever the term might be. It's artifact. And Claude's had this ability for a little while, but they just did something recently in the last week, at least this is what Liz tells me, that has made the ability to search, find, and utilize these artifacts much easier and accessible.

13:38Yes. So before over on this little sidebar over here, you would just have chat and projects and now artifacts. So again, before what you would do is you would go into one of your chats and let me pull up. Okay. I was looking at Riley Brown's, some of his, he's has a great podcast on stuff. And so when you go in and you create, let's say I took a transcript I wanted to summarize to give me a guide for growing my audience. And so it created a document like that. If you're just listening or watching, effectively what it spit out was a Word document of a step-by-step, hey, copying this person, here's how you could grow your audience.

14:15But it's in the Word document format. Exactly. And then in other instances, I would generate a cool little tool, but I'm like, okay, this is cool. I have it here. I was talking about last week, it was really hard to export this and I had to share a link and all that. And then in other cases, it's like you're building little you know, HTML dashboards or software tools that are interactive. And so it was one really hard to find, as you mentioned, once you created them. And then two, it was sort of was like, okay, great. This is cool. But now how do I make it useful? And so what this allows, so first off, if you create anything, you have now just an easy way to access them.

14:52So here are all the ones you created. And then if you publish it by going back into the chat, let's see. Okay. So So here you basically would go to, you publish, and then you can copy a link and share that with other people. Why would you publish it? This is, again, just making it shareable for other people. So you created this cool little dashboard, and I want to say, hey, Nick, check out my pitch deck here. And then you can share this link, and other people can access it. They can interact with it. This one was a PDF, so it's not really interactive, but you can have dashboards and things. And so within Claude, on the left-hand side, there's chats, projects, and now it says artifacts?

15:29Yes, that is correct. So the cool thing though, so again, now it's like, yes, it's helpful for search and accessibility organization. Okay. I get excited about that. But what's more interesting is they are now helping to guide people to actually produce these little tools. And so they have this tab called inspiration, which I just thought was pretty cool. So now they obviously - Where is this inspiration tab? I'm sorry. Literally just go to click on artifacts and it'll pop you right over there. And it's the first thing you see. So there's an upper thing that says inspiration. And then next to it is my artifacts.

16:06Can you search? It doesn't look like there's a search function within your artifacts. Let's see. Yeah, not yet. That's unfortunate. I'm sure they'll have that point here. So that's still a bit of a challenge. But anyhow. Oh, and so is this inspiration other artifacts that other people have published? Exactly. So it's like there are common use cases that I'm sure Anthropik sort of looked at their own code and did an analysis where people build regularly. So video is cool, but I have what scientists call a face for radio. And so what's even cooler is long form audio via my podcast and my newsletter, Nickonomicspod.com.

16:41Go there for free. Subscribe to my newsletter. It's one email per week. It's super tactical. And then go to my audio podcast. I do three to five episodes a week depending on how curious I am. And it's stuff like this. It's all free. No sleazy sales pitches. is nickonomicspod.com. Again, if you're listening, we're in Claude on the left-hand side. There's chats, projects, and now there's artifacts. We're under inspiration. And even under inspiration, there's like these sub tabs of learn something, life hacks, play a game, be creative, touch grass. And we're scrolling through what other people...

17:11Oh, a trivia game. Look at that. Look at that. Hold on. So click on that. I'm curious. Oh, trivia. Okay. Yep. yeah okay so she just clicked on the trivia game and so now you can do first it gives you the prompt so it helps guide you in learning better prompts it'll give you then the full chat of how this was generated because it wasn't like one-shotted wait oh wait wait i could you i could look at the chat that you used to create something yes and then we'll get into the customization so i just clicked on the chat and so as you can see here it goes all the way up to the top create a trivia game trivia app and react.

17:47I won't read the whole thing, but you can click on it. It provides that it analyzes data, gives you sort of the flow. And then it said, okay, I like it. The glow should, it had some iterations and stuff. And so I don't, yeah. That is so interesting. Chris and I were just talking about this, not even on a podcast. We were just like on the phone and he was like, I think there could be a cool community where people would share their chats to show how they got to an output. He's like, do you think people would pay for that? And I was like, I mean, I think people would be interested in it, but I don't know exactly how it would work.

18:21But that's exactly what this is. You can publish these artifacts and see exactly how they got to it. It's their chain of thought. And that actually brings up a whole other point too. It's like all of these little ideas of even building trivia tools or whatever and little apps. I just feel like the LLMs are just getting into the space and crushing all these little startups. Like I've seen, there was a startup that I even saw that built a couple of tools that are just like crushed. If you have, you know. Well, go back, go back to that trivia thing. Cause it said you could customize it. Yes. This is where it gets exciting.

18:54So now you can adapt it to your needs. It gives you the framework and now we can prompt if you want to change basically the functionality, you want to change the categories, what the goal is like, you can change all of that now. So because it's given you now the artifact as a starting point, as a template. Okay. So Liz, Liz clicked on this. And as I'm looking at it now on the left-hand side in the prompt window, Claude's basically telling me, Hey, transform any artifact into something uniquely yours. Here are the elements you can do. One, you could change the topic. You could adapt the content for a different subject.

19:23Two, you could update the style. Three, you can make a personal. Four, you could share your vision. And then it asks you, where do you want me to begin? So this is a trivia game right now that says test your knowledge with Claude generated trivia questions. And there's categories, et cetera. So I, for example, actually try this, Liz, say I'm a 39 year old parent. I have two boys ages 11 and eight. I would like to have a round robin trivia game with them where the questions difficulty is tailored to each person's age. I wonder if it will be able to do that it should be all right and she's just typing that out she hit enter and and i'm sorry liz if you're not 39 because now it's gonna be like wait a second what is this 30 oh it's transforming it yes i will transform this into a family round robin trivia game and you can see it's pumping out the code right now right in front of our eyes what the freak by the way you your kids are kind of my kids' ages.

20:26Have you ever played trivia games with them in ChatGPT? We did a family road trip. We had ChatGPT give us ideas for games. And then we did that like, here's a story, then have the next one add to the story. And it basically was able to remember the whole story and it added a little bit to it. So it was like really fun. We did it for about an hour and it worked really well. It was actually a really fun game. So we didn't do trivia, but we did that on our road trip. That's amazing. Can I tell you, My dad came last week. And one of the things I did is, you know how it is when grandparents come, they're just always talking.

20:59Well, at least my dad, he's always talking and telling stories. He'd be super interesting. He was born in Germany, raised in Brazil, came to the United States, professional soccer player. But I gave my son my pendant and I turned it on, you know, my little limitless AI. And they talked for hours while he was here over the course of time. And so it recorded all their conversations. And now I have his voice, obviously, but I just, I have the stories and I thought that's going to be a really, really cool memory for my kids. It's like, now I can take those stories and I wouldn't have happened if I hadn't had, you know, AI or the ability to record them.

21:31So I thought that was really cool. Yeah, no, that is so cool. And I know you and Chris have like sort of even talked about those ideas before on one of your pods about, you know, being a better memory catcher, but like, that's huge because we're all are losing family stories and we all want that history. I love that. All right. So this is, I don't know how close it is. There we go. Perfect timing. There it is. Oh, cool. Okay. So we'll say. Oh my gosh. Buddy will say, whoops, they put in so you can categorize like eight years old. Categories, general knowledge, sports history. Very cool. All right.

22:04Questions per player. All right. So let's just press go and see what happens. Let's see here. All right. So if you scroll up, there's no way to add. Oh, you could add another player if you wanted to. Yeah. Okay. Oops. Maybe we need to because for some reason it was struggling a bit there. No worries. But I just think that's amazing. Yeah. Sorry to like derail you. No, no, no. Holy crap. Exactly. I can take other people's artifacts and just tailor it to myself. That's amazing. Totally. Totally. So a couple other things I just wanted to share because I think there's some other use cases that might not be as obvious for people to consider.

22:37So a couple of weeks ago, we started sharing about MCP model context protocol and hooking up Claude to that. So one thing I've started playing around with, and I haven't perfected this workflow, so hopefully this works, but once you get some of these, you can sort of think as your MCP as the data retrieval, and then these little tools and artifacts with that nice UI that helps sort of guide you through your prompting or whatever you want to create as the visual output, that it helps populate the data. So you don't even have have to write things. It'll access some of the data, populate it into your tool, and then output it.

23:15So let me give an example. So I just ran this and I can redo it just for time's sake. What I asked this meeting notes summary tool to do was to use my Google Drive that it has access to to then find one of the most recent meetings I had and then to populate those raw meeting notes into this little box here and then I press summarize and it went ahead and summarized it. So we can just sort of recreate this here. Let me go back. Yeah. So that was one example. Let me just see if I can do it live to show something cool. Liz, you know what this is doing is like, this is taking on sort of the lowest level of lovable and replet.

23:57Yes. Yes. Like this is the wrapper that we all talk about. Now you can natively create a wrapper by creating these artifacts within Claude. Yep. And it's like, you know, there's the AI automation piece within AI agents, which essentially are doing tasks and they can do them more recurringly. Unfortunately, Claude doesn't really have that recurability yet. I've been trying to solve that here, but like a lot of stuff you still want to iterate through. But one of the challenges is, is that, you know, you, even though you might give your project knowledge, some general instructions of how you want an output to help with the recurring prompt, sometimes it's visual.