AI News: AI Video Creation Explodes in 2024

So I'm currently out in Boston for HubSpot inbound and, well, I'm kind of in a tiny hotel room with no desk. So I'm improvising and using what I got because I want to share all the latest news with you in the world of AI. Right now we're sort of in the midst of conference season. This week alone we had HubSpot inbound. We had Salesforce's Dreamforce event. Amazon had an event where they announced some new features. YouTube held an event where they showed off some new features. Snapchat held the event where they showed off some new features. Next week we got Meta Connect, which I'm going to be flying out to that right after HubSpot inbound. Tons of conferences are going on, which means tons of announcements are happening. But I'm going to go ahead and jump in and start with an announcement that didn't come from one of these conferences. Let's start with talking about OpenAI. So this week OpenAI put out this blog post here, an update on our safety and security practices. Now this blog post talks about some of the new security measures they're implementing as well as being more transparent. But probably the biggest news to come out of of this specific article on the OpenAI blog here is this portion right here about establishing independent governance for safety and security. Now when it breaks down the board here in this message, we can see the board is going to be chaired by Zico Coulter and include Adam d', Angelo, Paul Nakasone and Nicole Seligman. One name that is absent from this board is Sam Altman. He is apparently stepping down from this safety and security committee and and completely putting it in the hands of others. Now the biggest news that's been going on in the AI world as of recently was last week's announcement of OpenAI's O1 model. It's a model that is much more effective at logic and reasoning and complex math problems and sort of STEM subjects. Well, this week OpenAI announced that they've increased the rate limits for plus and team users by seven times. Now, admittedly I have not been home almost all month this month, so I haven't really had the opportunity to deep dive and play with OpenAI01 myself. I'm super excited to get back to my house after medcon next week and really spend some time deep diving on this 01 model. That way I can share with you some videos and tutorials and different ways that I'm using it and my own findings from playing around with it. But honestly I just have not had the opportunity yet. However, I have been seeing some really, really cool stuff that people have been doing with OpenAI01. For instance, my Internet buddy here, Amar Reshi, actually managed to make a 3D version of the Snake game in under a minute. So check this out when I pull this up to full screen here, he managed to make this 3D version of Snake that we're seeing right here on the screen in under a minute and it actually looks really, really, really good. So I mean with coding tasks my main go to has been Claude Sonnet 3.5. Lately I thought that's done a really good job with coding, but I don't know man. OpenAI01 seems to possibly have it beat with coding, so I can't wait to play around with it some more. And since OpenAI's 01 has come out, pretty much all the tools that have integrated the previous versions of GPT4 have started integrating this new model. For instance, if you are a GitHub Copilot user this week, you can now use OpenAI's 01 model directly inside of GitHub Copilot. If you're a fan of Perplexity like I am, our event here announced that they've added a new reasoning focus on Perplexity for pro users. So if you're one of the paying pro members of Perplexity, you can now use the new OpenAI 01 mini model directly inside of Perplexity. Now, it doesn't search the web yet and integrate that with your responses and you only have 10 uses a day as we can see right here with it. But exciting to see that Perplexity, a model that I really enjoy using, is adding all of the latest state of the art models into their platform as well. Now, there's been some weird stuff happening at OpenAI around this 01 model as well. For instance, people are claiming that OpenAI is threatening to ban them if they try to essentially jailbreak this new model. According to this article here, someone claimed that they received a warning email if they use the term Reasoning Trace in conversation with 01. Others have claimed that the warning is triggered Simply by asking ChatGPT about the model's reasoning at all. Here's a screenshot of an email that one person received. Hello, we're reaching out to you as a user of OpenAI's ChatGPT because some of the requests associated with your email address have been flagged by our systems to be in violation of our policy against attempting to circumvent safeguards or safety mitigations in our services. Please halt this activity and ensure you are using ChatGPT in accordance with our Terms of Use and Usage Policies. Additional violations may result in loss of access to GPT, 4O, et cetera. Now, if you have used this O1 model at all, you'll notice that when you ask it a question that it sort of shows you the reasoning and the logic that it's thinking through. As it's thinking through it. However, it's showing you like a summarized version of that logic. We're not actually seeing the complete chain of thought that it's going through. It's just like little snippets of the chain of thought and people are trying to really reverse engineer that and figure out what is this model actually sort of telling itself and going back and forth on OpenAI is still trying to keep that hidden. I'm assuming it's because they're worried that other companies might sort of reverse engineer it. And apparently if you do try to reverse engineer it, you could potentially get banned from using OpenAI products. Here's another really weird one. I've only seen one claim of this, so I'm not sure if this is fake news or not. It totally well could be, but one person on Reddit claims that ChatGPT actually initiated a conversation with them before they ever sent a message. We can actually see in this screenshot that they shared here. ChatGPT shared the first message here. How was your first week at high school? Did you settle in well? This person Milo here responded, did you just message me first? Chatgpt says, yes, I did. I just wanted to check in and see how things went with your first week of high school. If you'd rather initiate the conversation yourself, just let me know. They ask if it's a new update and ChatGPT replied with yes. It's part of an update that allows me to check in or follow up on things we've talked about. Like your first day of school. I can still wait for you to start the conversation if you prefer. Again, I've only seen this one claim of this happening. I feel like if this was happening a lot, we'd see it all over X and we'd see other YouTube videos talking about this right here. This is the only one I've seen, so I don't totally know how legitimate it is. I also don't understand when this would happen. Would it happen like when you open up the ChatGPT window and start a new chat and then the new chat just starts messaging you I don't really quite understand how this would happen happen in the first place, which is sort of making my alarm bells trigger that this could be fake. Moving on to Google, there's been a handful of little Google AI updates this week, most of them actually on the YouTube side. But they did announce that they're going to start flagging AI generated images directly in search later this year. Inside of Search, Google Lens, and the Circle to Search feature on Android. If an image has metadata that states that the image was generated with AI, it's going to let people know as soon as that image pops up. Now it has to have that metadata. It can't actually tell what's AI and what's not unless this metadata is actually sort of baked into the image. Moving on to some of the YouTube announcements. There was a little YouTube event this week where they announced a bunch of new features rolling out. I'm going to focus on the AI features, including one that's coming out of Google DeepMind. Google's rolling out some new video generation features directly inside of YouTube shorts. They're going to be integrating Google's VEO model directly into Dream Screen. You can see here they have a little Create button at the top of the screen inside of the app here. When you click on Create, it asks you to describe what you want. It gives you some sort of presets like vintage, watercolor, digital, etc. And in this example, they type cinematic underwater reveal of the Golden Gate Bridge. And then they click on Create. And then what it does is it generates four different still images that it will use as the sort of starting frame of your video. So you can see it generated these four images here. You select one that you like and then it does its thing and converts it into a video version of that image that you just generated. Here's another example of a dreamlike secret garden with vivid colors, et cetera. And here's a whole bunch of different examples of what it generated from that. But that's not the only AI features they're rolling out for YouTubers. They're actually adding a inspiration feature here which can help curate suggestions that you can mold into fully fledged projects. So they can type, brainstorm video ideas, new types of art, and they click get ideas. And then it lets them explore video ideas and then it helps them create the outline here. And then it starts generating thumbnail ideas for that video for you. So it really kind of looks like it's going to help across the board with ideation, outlining thumbnail ideas. The entire gamut of the sort of creative process of ideation for YouTube. They also announced that they're bringing auto dubbing to YouTube, which is really cool because I can record a video like this in English and it will automatically dub it and localize it.