ChatGPT Store is Here! How to Use and Make Money

Matt Wolfe breaks down OpenAI's new GPT Store, where you can browse and build custom GPTs. Creators will get paid based on user engagement starting in Q1, though most won't earn much unless they're proprietary or API-connected. ChatGPT for Teams costs $25 per user per month and includes GPT-4 with vision, DALL-E 3, and a 32k context window.
Questions this video answers
January 10, 2024
$25 per user per month (billed annually, minimum two users)
32,000 tokens (teams); presenter notes 120,000 tokens mentioned for GPT-4 Turbo at Dev Day
Planned for Q1 (year not explicitly restated in transcript)
Chapters
So I'm still here in Vegas at ces, but a flurry of news has been coming out of OpenAI around ChatGPT, and I wanted to make a super quick video to loop you in on what's been going on over at OpenAI. I'll get back to my CES updates probably in tomorrow or Friday's video. But here's what's going on with OpenAI and ChatGPT right now, starting with the big news that the GPT Store is now open. Now they are rolling it out. So it may not be in your account just yet, but by the end of the week I imagine you'll have access as well, assuming that you're a ChatGPT plus subscriber. Now, in order to access the GPT Store, you're going to simply log into your ChatGPT account. Over on the left, click on Explore GPTs, and then in here we see what the new GPT Store looks like. We've got our featured GPTs. Each week, people over at OpenAI are going to select specific GPTs and share them with you. You've got your trending GPTs here. I'm assuming these are the most different, downloaded, most popular ones, including Grimoire here, which is one that I've referenced in multiple videos. They're in the top four trending GPTs right now. We can see some of the GPTs created by OpenAI here. We can see some specific Dall E GPTs, some writing GPTs, productivity, research and analysis, Programming, education, lifestyle. So they have a whole bunch of categories here of GPTs that you can use. They've got a quick access menu to get to any of these categories and then we can search public GPTs. So for example, let's go ahead and search YouTube and see what sort of GPTs we can find related to YouTube. So we've got a video summarizer, YouTube summarizer, SEO GPT Generator, Alpha notes, Transcript Thief, Video summarizer, lots of video summarizers. A Thumbnail Expert, more YouTube summarizers. One of my early criticisms of GPTs is that if you make a GPT that's a good idea, there's not much stopping other people from just going and creating their own similar GPT. They would just have to go up to create up here and essentially explain what the other GPT that they saw does and let this create one for you. So if I wanted to make a YouTube summarizer, I could simply give it a prompt of make me a GPT that summarizes YouTube videos and it will make me a GPT that does what all those other GPTs do. However, where I really think GPTs are going to be powerful, I are when companies use this create new action area and then sync it to their own APIs. There's a lot of software companies out there that have APIs where you can use ChatGPT and essentially talk to other services. You know, the Ubers and the Open Tables, Google Calendar, Zapier tools like that. They'll be able to create custom GPTs connected to their APIs to communicate with those services. I think that's really powerful. Also, with GPTs, you have the ability to upload files and to create your own knowledge base that people can have conversations with. And I think companies that have proprietary knowledge and data that they can share in their uploaded knowledge base is going to be really valuable as well, because other people can't just duplicate those GPTs. So some other details about the GPT Store. They're going to feature new GPTs every single week. You can include your own GPTs in the GPT Store by saving your GPT for everyone. Now, now there is a review process. It is both automated and human reviewed. So if they think your GPT Store doesn't meet their terms of service, they will decline it and it won't make it into the store. Now, the question on everybody's mind, how do I make money with GPTs? Well, OpenAI hasn't specifically announced that yet. What they did say here is in Q1, we will launch a GPT Builder revenue program. As a first step, us builders will be paid based on user engagement with their GPTs. We'll provide details on the criteria for payment as we get closer. So, based on what they're saying here, it sounds like the way you can monetize GPTs is essentially by being one of the top GPTs. They will do a revenue share. It doesn't sound like it's going to be like the Apple App Store model, where you just put a price tag on your GPT, people purchase that GPT and then you split the revenue with OpenAI. It doesn't sound like that's what they're going for. It sounds like it's going to be more similar to what you get out of Twitter or out of Spotify, where OpenAI is collecting the user fees. If they use your GPTs at a high level, they will give you some of the revenue. Unless you have a GPT that does something really interesting, really proprietary, or is connected to one of your APIs, it's unlikely you're gonna make a ton of money off of GPTs. There will be outliers for sure, but the majority of GPTs, it doesn't sound like, are gonna make a lot of money because you really need yours to stand out and also be some somewhat proprietary so that other people can't just knock it off and make clones of it. But that's not the only interesting thing to come out of OpenAI this week. That's probably the most exciting for most, but there are some other cool things that they just released, including ChatGPT for Teams. Now, ChatGPT for Teams gives you access to the larger 32,000 context window. Now they did mention at Dev Day that they have like a 120,000 context window for GPT4 Turbo, so I wonder when we're going to start to get access to that. But teams do get access to a larger 32k context window. Now my light just died, so the lighting in the room just got more horrible. But I'm going to keep on rolling here. Chat GPT teams also comes with higher message caps on Dall E3GPT4 with vision browsing and advanced data analysis, previously known as code interpreter. They won't train on your business data or conversations. And there's a secure workspace for your team so you can actually share chats and easily collaborate with Chat GPT. Now, if you want to set up a team plan, pretty easy. Log into your Chat GPT account, come down to the bott left where it says upgrade plan and you can see there is now the plus plan and the team plan, which costs $25 per user per month. But it does say down here price build annually, minimum two users. So at 25 bucks a month per user per month, that's 50 bucks a month. And it's billed annually. So you're looking at 600 bucks to upgrade to teams Now, I also saw this tweet this week from my buddy Linus. This says Chat GPT now has memory. Now I can't actually find this in my own OpenAI account, but but it looks like this is a sign of things to come. We're about to get better memory inside of ChatGPT. He shared this screenshot here that says your GPT can now learn from your chats. Keep the conversation going. Your GPT will carry what it learns between chats, allowing it to provide more relevant responses. And it improves over time. As you chat, your GPT will become more helpful. Remembering details and preferences. Manage what it remembers. Your GPT has been designed to follow Your instructions in chats, you can reset your GPT's memory or turn this feature off in Settings. Now, when I go to my own settings here and click through all of the options, I don't seem to have any options around memory. But this is exciting because there's been a lot of times I've been having long conversations with ChatGPT, but over time it forgets the earlier conversations in the thread. They seem to be solving this problem with a new rollout that they're apparently doing. And one more quick thing about the settings here. If you are creating a custom GPT and you want to actually be credited for that custom GPT in the GPT Store, if you come to your settings down here, click on Builder Profile. You can have it share your name on the GPT and you can also add a domain name so that people can find your website. So even if you're not making money off of your custom GPT, it could still potentially drive a good bit of traffic to your website. If people use your GPT, they see your URL, it could be a decent little traffic driver to your business. So just something to keep in mind there, even if you're not going to make a ton of money off of your custom GPT. And then finally, since we are on the topic of OpenAI and ChatGPT news here, OpenAI did come out and make a statement about the New York Times lawsuit that's been going on. Basically saying that the training that they used was all fair use. Regurgitation is a rare bug that they're working to drive to zero. So the fact that it was able to repost the content verbatim is a bug that they're trying to fix. And they're saying here that the New York Times is not telling the full story. This is something that I reported on on a couple different New York Times related news videos. Is that the way New York Times prompted this is not how anybody would ever prompt ChatGPT. They gave it the URL of the article. They gave it like the first three paragraphs of the article. And then since ChatGPT is a next word prediction machine, since it had so much of the existing article and the URL fed into it, it predicted the next word based on the article. Pretty dang. Well, so ChatGPT technically did what it was supposed to do, but nobody prompts it that way to try to get articles verbatim. You can see here they say, interestingly, the regurgitations the New York Times induced appear to be from years old articles that have proliferated on multiple third party websites. It seems they intentionally manipulated prompts, often including lengthy excerpts of articles, in order to get our model to regurgitate. Even when using such prompts, our models don't typically behave the way the New York Times insinuates, which suggests they either instructed the model to regurgitate or cherry pick their examples from many attempts. My thoughts I don't really think the New York Times lawsuit holds weight. It probably holds more weight than most of the other lawsuits that have been brought up so far. But still, it seems the training was fair use and it seems the way they prompted it was a little sketchy to try to get it to repeat it verbatim so that they can turn around and sue. That's that news A lot happening over at OpenAI again. I am still here in Las Vegas for the CES conference. Hopefully I can get my lighting to work again for my next video. Because in my next video I'm going to be breaking down all of the coolest stuff I've personally come across here at cs. So if you're not subscribed to this channel already, make sure you subscribe like this video if you want to see more AI news in your YouTube feed and I'll make sure that that kind of stuff keeps showing up for you. So thank you so much for tuning in. Check out FutureTools IO. If you haven't already. It's a really cool website. I think you'll dig it and really, really appreciate you spending the time watching my videos. Thank you so much. I'll see you in the next video. Bye bye.










