The AI Daily Brief: Artificial Intelligence News and Analysis - The 9 AI Tools I Actually Use Every Day As A Content Creator

Episode Date: May 26, 2023

Lists of 1000s of AI tools abound, but here are the 9 I actually use every day (with a watchlist with 3 more I'm experimenting with). The list includes: 1. Feedly 2. ChatGPT 3. Perplexity 4. Descript 5. Midjourney 6. Photoshop 7. Deepfloyd 8. Canva 9. TubeBuddy 10. Vidyo.ai (watchlist) 11. Wondercraft (watchlist) 12. Elevenlabs (watchlist) The AI Breakdown helps you understand the most important news and discussions in AI.  Subscribe to The AI Breakdown newsletter: https://theaibreakdown.beehiiv.com/subscribe Subscribe to The AI Breakdown on YouTube: https://www.youtube.com/@TheAIBreakdown Join the community: bit.ly/aibreakdown Learn more: http://breakdown.network/

Today on the AI breakdown, I'm talking about the AI tools I actually use every day as a content creator. Before that on the brief, Microsoft joins the AI regulatory fight, TikTok is adding a chat bot, and some pretty amazing medical research using AI. The AI breakdown is a daily video and podcast about the most important news and discussions in AI. If you're enjoying it, please like, subscribe, and share, and find out more at breakdown.network. Welcome back to the AI breakdown brief, all the AI headline news you need in five minutes or less. AI Breakdown Brief, I'm building off of the AI breakdown newsletter, which comes out every morning and I call the first five. It's the five most important news events happening in AI that day. And if you're

enjoying the video brief, you might find it a useful way to see what's happening in AI in a very short, compact form. We kick things off with Microsoft's call for a new AI agency. Yes, Microsoft has joined other big tech companies, including Google and Open AI in advocating for more regulation of AI. Microsoft President Brad Smith gave a speech in D.C. yesterday that outlined a five-part plan for regulating AI, including joining Open AI and calls for a new dedicated regulator. Now, when it comes to what Microsoft is concerned about, Smith said that his biggest concerns were deepfakes, AI representations of public figures that could wreak havoc in the democratic process. Of course, not every company agrees with this approach, and in an op-ed yesterday

in the Hill, IBM wrote why a single AI regulator just won't work. Instead, they recommend that every regulatory agency be retrofit for an AI world, in effect for every agency to become an AI agency. Next up, OpenAI has launched a million-dollar grant program, earmarking 10, $100,000 grants for people to help create frameworks for democratic processes around important questions of AI. Those questions include things like, how should AI treat public figures when there's a diversity of opinions about those figures, as well as issues of bias and representation. Now, not everyone is enthused about this approach. Jeffrey Miller responded to OpenAI's tweet, saying the most fundamental democratic process would be to run a large-scale global survey of humans on

whether AI should be developed at all or not. If you're not willing to ask that, you're just imposing AI on our species without our consent. Next up, a really fascinating story. AI has been used to help discover a new antibiotic for a very dangerous superbug. Ascentobacterbominy is a type of bacteria that is classified as holding the greatest threat for humanity by the World Health Organization because it can basically pick up genes from other compounds to build drug resistance for itself. In a new study, researchers exposed that bacteria to 7,500 or so different drugs and drug compounds, and then trained an AI model on the 480 or so that showed promise in stopping the development of this particular bacteria. From there, in just a few hours,

the AI model was able to test 6,000 plus different chemical compounds that showed promise potentially in treating the bug, of which about 200 were then tested in a lab. Ultimately, one was shown to not only deal with the superbug, but basically to only treat the bug. that use case. We're at the very beginning of AI being used in medical research and this shows just how promising it might be. Next on the brief today, this seems completely inevitable. TikTok has joined the chatbot frenzy. Now, TikTok is already almost entirely powered by AI. The 4U page, which recommends the next video for you, is in fact an AI tool. But this chatbot potentially adds a whole different dimensionality. Imagine, for example, seeing a video of a recipe and then being able to ask the chatbot,

which they call taco, to give you more information about how that recipe could be made at home or what the ingredients were. Currently, TikTok is only testing this in the Philippines, and it was only discovered by a firm that keeps track of new features in big social platforms. Finally, today on the brief, Google's new search generative experience is here. This is what we were expecting post-I.O. And it is the way that Google is integrating generative AI into their search experience. You can join a wait list now when users are just starting to get access to it and figure out what it means for the reconstruction of the internet around a very different type of search. I'll have more on that on the weekly recap tomorrow, but for now, that is the most important

headline news in AI. If you're enjoying the AI breakdown brief, please like, subscribe, and share, and I'll be back soon for the main AI breakdown. Today, we are talking about the AI tools that I actually use day in and day out as a content creator. If you've spent any time around AI, you've probably seen something like this. Zane here says 1,000 AI tools were released last month. The chat GPT is just the tip of the iceberg.

And I'm not trying to call out Zane specifically. This is a standard format for content creators on Twitter and in other places. And the idea here is to make you feel like there's so much that you're missing out on and that there's incredible tools that could be changing what you actually do. And on the one hand, that's not totally wrong. There are incredible tools out there. But on the other hand, there's no way for any human on the planet to keep up with everything that's being built.

And so what I'm doing today is to show. show you the actual AI stack that I use day in and day out to create videos, podcasts, and other types of content. For those of you who are not familiar with me, I create a lot of content. This is my AI channel on YouTube. This is my crypto, Bitcoin, and macro channel on YouTube. All of these things feed into not one, not two, but three podcasts that I release, two of which are daily.

So when I tell you that it's imperative that I use AI tools to increase my productivity, it is truly imperative. The content that I create about AI and other topics is all usually based around the most important news and discussions. So one of the things that I use AI for is to help me keep track of the news. What you're seeing right now is a service called Feedley, which has just introduced a new feature called Feedley AI. So if you go to Feedley AI, you can create a feed that is customized for exactly whatever it is that you're interested in tracking. So I'm just going to click their Artificial Intelligence Feed, and you can see here that there are lots of ways to cut. customize this. You could add, for example, that you wanted to see only articles that match

artificial intelligence as well as trends. I could go down here and say, I want to match it to the trend creator economy. So now we've gone from a list that's all the artificial intelligence articles to all the articles that match both artificial intelligence and the creator economy. So you can see that's 114 articles this week. Now, admittedly, I'm pretty voracious, so I just look for all of artificial intelligence and do my endless scrolling until I figure out exactly what I want to see. When you do find an article, you can either open it up right there or you can read later with the bookmark. The next tool that I use on a pretty regular basis, usually every day, is of course chat GPT. Now, there are tons of different things you can use chat GPT for, so I'm just going to

talk about what I tend to use it for. Imagine that I've gotten all these articles about things that have happened in AI from Feedly, and I've honed in on the ones that I want to feature on either the AI breakdown or the AI breakdown brief. Sometimes, especially in the context of the brief, all I really need is a very high-level summary, right? I'm not going deep into the particular topic. I'm just mentioning that it's happened. For that, chat GPT can be extremely helpful as a tool for summarization. For example, 2018 Turing award winner, Joshua Benjio recently wrote a blog post called How Rogue AIs may arise. And while it's definitely worthy of a whole read, I was just mentioning it in passing in the context of a larger piece. Taking advantage of ChatGPT's new

browsing features, which has made it vastly more useful. I had it summarize the piece in less than 10 bullets. Now, I also found a Twitter thread that disagreed with Benjio's post, and I asked chat GPT to summarize that as well. Obviously, a Twitter thread is much shorter, but it's still really helpful to see how chat GPT summarizes and synthesizes what's being said in that type of medium. The other type of summarization that I use chat GPT for is actually more like simplification. A lot of these advanced concepts in AI are extremely technical, and to the extent that my goal as a content creator is to help people understand them in simpler terms, I myself can use chat GPT to get that explain it like I'm five type of understanding. This week, for example, I did an AI breakdown all

about new research in which scientists reconstructed videos from thoughts, and I wasn't working off of some news report. I was working off the actual research paper. If you've ever read a technical research paper, you know they're extraordinarily dense. So I turned to chat GPT to help me summarize it and was even able to specify that I wanted it for an undergraduate level audience. I've found that specifying the age or education range that you want chat GPT to speak to to is an extremely effective way of getting the results that you want when it comes to this summarization and simplification. In addition to just summarizing information, I'll often ask chat GPT to explain a concept

that has a specific terminology around it, such as recently I asked it to explain generative adversarial networks when everyone was getting all excited about that draggan research that had people modifying photos simply by clicking and dragging them around the screen. I'm sure you've seen the exact thing that I'm talking about. Now, as part of this daily chat GPT use, you'll notice that there are a bunch of plugins that have become really important to me as well. One that I use all the time is XPapers, which can pull from the archive database.

Those summarization tasks that I was just showing you, for example, use the XPapers plugin. I've also been exploring a number of different finance-related plugins. And then, of course, just the browse plugin, giving ChatGPT the ability to connect to the internet is extremely essential as well. Now, one additional tool that's starting to make it into the rotation on this research phase is Perplexity. Perplexity is a chatbot just like ChatGPT,

and it also sits on top of the GPT model, so it's pulling from the same model the ChatGPT is. But what I think Perplexity shows is that the user interface makes a big difference when it comes to utility for different use cases. So let's take a look at how Perplexity handles this question that I was asking on ChatGPT just a moment ago. What are generative adversarial networks? And for this, I'm using Perplexity's new co-pilot tool,

which is their most advanced approach. One of the things that perplexity does is it gives you really precise information about where it's searching. So you can see here it's searching the web for things like GAN's basics, how to GANS work, and then it shows you the 32 results that it's drawing from. What's more, as it pulls up information, you can see that it's actually sourcing where that came from, so that if you need more detail around that particular piece or that's something that's interesting, you can click on it and directly go to that article where it came from. You can see here that they've listed with icons all of the sourcing,

and they also have this related questions feature, which is really cool as well. So, for example, they suggest how to GANS differ from other generative models. Given that a lot of the way that I'm using chat GPT is for this type of background research that I'm then bringing into the scripts for the videos that I'm doing, Proplexity really offers something a little bit different and I'm using it more and more. Okay, so now I have all the information. I'm ready to make a video. How do I actually make it?

Well, for me, the easiest tool from both a production and an editing standpoint is Descript. When it comes to the production side of things, Descript is pretty similar to a lot of other tools out there. You can have either a direct camera view or you can screen record while you're also narrating. But where descriptorily becomes powerful is how it does transcripting. When you edit into script, you're actually highlighting text and engaging with it as though it were a word editor. So I can use my cursor to highlight a sentence, click delete, and boom, it's gone. Now, they also have tools for more precision editing, but this radically reduces the time that it takes to edit video. They also have AI that identifies filler words like ums and

and you can get rid of all of those basically with a single click. A last AI feature of the script, which makes it incredibly powerful, is their overdub feature. Using overdub, you can create a text-to-speech model of your own voice, and then if you flub up or need to add something into the script, instead of shooting a new video, you can literally overdub with this AI, typing in the additional words or sentence that you want to say, and it comes up sounding really natural. Okay, so now we've created our video, we've edited it, we have to get it ready for

YouTube, and that means we're going to need imagery for the thumbnail. For that, my first stop is always Mid-Journey. Now, this is also true for podcast art, as well as for art for my newsletters. Just today, I did a video about how Elon said that there was a non-zero chance that AI could go Terminator, and so I asked Mid-Journey to create Elon Musk surrounded by Terminator robots. I was able to customize the aspect ratio to match what I needed for YouTube, and basically just let it rip. Now, with this use case of Mid-Journey, prompting really isn't about necessarily precision, so much as creativity. Other recent uses for this were artificial intelligence on Wall Street. For a piece about the commingling of funds, I said bankers dumping two bags of U.S. dollars into one big pile. For the

NVIDIA breakdown brief, I wanted a picture of a stock market frenzy. And you can see the mid-jurney went to illustrative styles, which is exactly what I was going for, even though I didn't specify illustration or painting. And then finally, I recently did an episode of the main breakdown show about how Ron DeSantis, in his launch event for his presidential campaign, had specifically called out how he would protect Bitcoin. Now, I got an okay result, although it doesn't look exactly like Ronda Santis, and that brings me to a second tool that I've only just started to start to use for the sake of thumbnails and imagery. That tool is the new generative fill AI in Photoshop beta. Now, you've probably seen videos about this. I did one myself just recently, but generative

Phil puts the power of generative AI and text to image directly into the Photoshop suite. So you can choose any selection tool to grab what you want. In this case, I'm trying to grab DeSantis's shirt because I want to turn it into something Bitcoiny. And then you can see down here you have the generative fill box. I'm going to say, add a large gold chain necklace with an oversized Bitcoin B logo on it. And let's see what it comes up with. So it gives you a few options. This is the first one. This is the second one.

This is the third one. And there we go. We have one that's actually pretty good. It's ridiculous. It's over the top. But that's exactly what I was looking for, right? This is for a thumbnail for a YouTube video. You want it to capture attention.

Obviously, this tool is only a few days old, so it's just starting to get into the rotation. But I think I'm going to use it pretty frequently. The last tool that I'll mention around thumbnails and imagery that's more on my watch list than daily use is Deep Floyd. This is a stable diffusion-related project that's only in the sort of research stage. It's on Hugging Face, but it's not production ready exactly yet. But the promise of Deep Floyd is that it actually understands text and can put text in photos, which obviously mid-jurney so far cannot.

So you can see here, I said, punk-looking girl with a shirt that says Bitcoin. They give you four options and you pick your favorite to upscale. Let's do the one that looks like Avril Levine. So again, not necessarily super production ready. These eyes are still pretty wonky, but you can see the promise. And it does something different that's really important if you want Max creativity with your thumbnails and your imagery. Next up, a small but significant one, and this is the new AI tools inside Canva. So Canva has a whole suite of tools, but I just want to show you the one that is the most time saving and simple for me.

And that is their background remover tool. This seems so simple, but this used to take me so much longer. Now you just go over to edit photo. And under tools, there's background remover. Boom, you're ready to go. Now, for the sake of completeness, the way that you would finish this is you would go to share, go to download, select file type PNG, and then transparent background, select the page that you're actually on, then download, then bingo-bango, you have the thing that you need. Now, one use case that I didn't mention chat TPT for is writing copy. This is something that, of course, people use AI for all the time, and the only reason that I'm not mentioning it is that I've been writing my own titles, headlines, copy for so long that I have a sense of both what I like and what I think is going to be incentivized by the algorithm.

rhythms. But at the same time, even with that confidence slash hubris, a long-time tool that I've used in YouTube, which is TubeBuddy, has started to integrate AI tools in a way that makes it much simpler to at least gut check myself. So now, although I'm always coming with my own first start to the copy, I'll often go to its title generator tool and just generate ideas to see if I find anything that I like better. Most of the time I don't, but sometimes I find something that I think actually does a slightly better job of capturing the essence of what I'm trying to do. The other that I've started to play around with of theirs is their AI thumbnail analyzer. They say this is an experimental tool that's undergoing training, so basically don't put too much stock in it. In this case,

they're suggesting that the thumbnail I have now would be more effective than a version that has a bigger Elon on it. I'm not totally sure what that's attributable to, which is probably why you need to be careful with these things, but it seems like another really interesting thing that I could see myself using more in the future. Now let's wrap up with a few tools that have either just started to use or on my watch list. Video.AI allows you to take one video and make tons and tons of short videos from them. I am starting to use this tool and there really isn't any reason I'm not using it more other than timing. It is a massive time saver compared to former versions of this type of thing that weren't using AI. So I'd highly recommend you check out video or there's a couple other competitors

as well that take your YouTube videos and turn them into lots of short form videos optimized for YouTube shorts and TikTok very, very easily. Another new tool that I'm exploring is Wondercraft AI, which is basically a podcasting AI tool. And I'm not imagining it as a replacement for myself, but instead as a way to supplement and augment and extend myself. I also do newsletters on top of this audio and video content. Maybe I embed an audio version of that created with Wondercraft in the newsletter as well. And along those lines, I am also currently experimenting with training a voice AI of myself with 11 labs so that if I do decide to explore something like Wondercraft, I have a model of my own voice that I could use to apply to that reading.

I'll do another video all about how to find the right tools. But for now, I'll just shout out future tools as one great resource for how you can discover tools that fit your particular needs in the AI space. This is created by Matt Wolfe, who's also another great AI YouTuber. So go check out FutureTools.io if you're looking to just go through the full range of possible tools. So friends, that is it for the AI tools I actually use every day as a content creator. I hope you found this useful. I love being able to use these tools to bring you the AI breakdown and the AI breakdown

brief along with all my other content. And hopefully this helps you think about how you can make your own content more effective and productive with AI as well. All right, guys, that's it for today. Until next time, peace.