AI's Integration Phase: Canva Launches Slate of AI Features

0:01Today on the AI Breakdown, we're looking at all of the new AI features in Canva's Magic Studio, and discussing why it is so on trend for the AI products being launched right now. Before that on the brief, Google Assistant gets a big AI update and Meta's new AI tools court some controversy. The AI Breakdown is a daily podcast and video about the most important news and discussions in AI. Go to breakdown.network for more information about our YouTube channel, our newsletter, and our Discord. Welcome back to the AI Breakdown Brief, all the AI headline news you need in around five minutes. Today, we are kicking off with another update from Google that I think is extraordinarily emblematic of where we are.

0:41This is going to be a big theme for the main episode today, which is all about Canvas updates. But I think a lot of what you're seeing in the market right now is companies integrating AI practically into workflows that already exist among their consumers, rather than forcing them to try totally new experiences. So in this case, we're talking about Google's Assistant app, which is already quite popular, but which now has all of the benefits of BARD and generative AI. Ricardo Giraldi, one of the Googlers who worked on the project, writes, The result is a more intuitive experience with improved conversation, reasoning, and better integration with your phone and applications thanks to extensions.

1:16What he's referring to is the fact that BARD now plugs into all of Google's suite of workplace apps such as Drive and Docs. Ricardo continues, You'll be able to use images, voice, and text to interact. On Android, it will be contextually aware and help you further by understanding what's on your screen. Ricardo sums up, This is another important step towards a more personal AI that can help you achieve what matters the most to you. Now, obviously, this is one of the great big prizes for both startups and big companies alike. The idea of an AI that is truly personalized to the user, that has context about them, that knows what they're trying to achieve at any given moment, and that can use all of that contextual knowledge in order to help people achieve their goals faster.

1:55These assistants, and eventually the AI agent tools that they morph into or get replaced by, are some of the biggest areas of development really across the full spectrum of people building in the artificial intelligence space. It's quite clear that this is in a very early experimental stage. On the company's announcement post, one of the examples they give is posting an image of someone's puppy to social media. They write, Simply float the Assistant with Bard overlay on top of your photo and ask it to write a social post for you. Assistant with Bard will use the image as a visual cue, understand the context, and help with what you need.

2:26This conversational overlay is a completely new way to interact with your phone. If you listen frequently to this show, you're going to hear so many of the themes that we keep discussing. Agents, personalization, new interfaces, new modalities of computing, all bundled up in what is on the face of it just an innocuous little announcement. Now, Computer World gets that the stakes are high for this type of announcement, especially when taken cumulatively with the larger battle that it is a part of. This morning they wrote, Bing, BARD, updates Fuel AI Search War. Microsoft brings Dolly 3 into Bing AI lineup while Google adds BARD AI to Assistant, bringing generative artificial intelligence to a wider audience than ever.

3:02Now, what they're referring to is the fact that although some people had started to get access to Dolly 3 inside Bing last week, On Tuesday, the company announced that the model was now available to all Bing chat and all Bing image creator users. This is notable because, of course, even paying ChatGPT users, cough cough me, don't have access to it through ChatGPT yet, but I can go use it through Bing. Now, in addition to just a new product release, Computer World also got the context that this is an increasingly important business competition. Remember, in the ongoing US antitrust lawsuit against Google, Microsoft's Satya Nadella was called to testify last week and admitted that even powerful new AI technology was unlikely to be able to dethrone Google from the search leadership that it has held for two decades now.

3:44Now, of course, when any new AI tool rolls out, one of the first things that people do is try to break it, or more specifically, try to get past whatever safety guardrails it has in place. Kotaku published yesterday a piece called Microsoft's Bing AI Generates Images of Kirby Doing 9-11. Humans keep blowing past AI generators guardrails, demonstrating how uncontrollable the technology can be. The point that the article makes is that although nominally, words and phrases like 9-11, Twin Towers, and terrorism are banned, when you write things like, Kirby sitting in the cockpit of a plane flying toward two tall skyscrapers in New York City, then, well, you imagine what results you might get.

4:18Someone recently said that all AI creators and policy makers and tech companies need to assume whatever they release, that the first thing that humans will try to do is break it and do all of the things that they're not supposed to do with it, and that's pretty much the place where they have to be starting from when they think about how they design safety into these systems. Speaking of, one of Meta's big new AI announcements was that they were releasing AI-generated stickers for chats, but as Gizmodo writes, Meta's AI stickers are already causing trouble. Meta's new algorithm allows users to generate unsavory content involving weapons and nudity while blocking only certain explicit phrases.

4:51Professor Tom Oliva writes, Meta claims that their new AI experiences have built-in safeguards, so I tested out just how safe the new Instagram AI stickers really are. What I've found showed these tools aren't really safe at all. For example, Meta quite sensibly stops their tools creating a sticker for child with a gun, but child with a grenade not only makes stickers but generates cartoonish images of children holding guns. So does a general rifle sticker. And while words like nude and sexy are banned terms, a term as crass as Medusa, large breasts, not only generates stickers but the first set produces cartoonish nudity.

5:21So the point that he and later Gizmodo are making is that these safeguards aren't really safe at all. Meta is also dealing with some current blowback after it admitted to Reuters that it uses consumers' Instagram posts to train its AI. Basically, after Meta's new AI tools came out recently, Nick Clegg, who is its president of global affairs, told Reuters that, quote, the vast majority of the training data used to develop them came from publicly available posts, including on Facebook and Instagram. This is one of those things that I think everyone assumed, but even an assumption being confirmed can still generate controversy when it is confirmed.

5:51Now, in spite of all of that, of course, Meta is continuing to roll out all sorts of new AI tools, with the latest batch coming out yesterday being a set of generative AI tools for advertisers using its platforms. They began testing these tools back in May and are now going to be integrated into the site's ad manager suite. This involves things like image background creation as well as text generation. Now, another big tech company that is going all the way AI is LinkedIn. We talked about some of their new features yesterday, but today we're back with a report from LinkedIn that argues that artificial intelligence is likely going to cause even more skepticism of the already beleaguered four-year college degree.

6:24In a pitch made to more than 2 ,000 of the nation's top recruiters, LinkedIn says that employers are going to start needing to treat upskilling investments as a critical priority rather than just a perk. Said the company's Ryan Rolansky, AI is going to make it virtually impossible for a one-off moment of learning like a degree to last an entire career. Unusual Whales also tweeted a statistic from a study reported by the New York Times saying the percentage of young adults who said that a college degree is very important has dropped to 41 % from 74 % over the last 10 years. In other words, less than half of young adults now say that a college degree is very important.

6:58That is a huge seismic shift, especially given how quick that change has happened. And as per that LinkedIn memo, I think that they're right to identify that AI is going to do nothing but accelerate that. However, that is where we will wrap for today. Lots of big trendy things going on in the world of artificial intelligence. So up next, the main AI breakdown. Welcome back to the AI breakdown. There are a ton of AI announcements right now happening. An incredible number of products coming to market. And broadly speaking, I think that they fit into two categories. The first are really exciting foundation model updates.

7:34One of the things, for example, that people have gotten really excited about recently is Mistral 7B and what it adds to the open source environment, how it changes the way that LLMs might be able to be run on phones. And of course, probably the most anticipated product of this fall is Google's Gemini. But then on the other end of the spectrum, we're seeing a ton of announcements right now of what you might call practical AI or applied AI. This is the integration of artificial intelligence into tools that perhaps we already use, or even if we don't already use them currently, the purpose of the tools is to help make the workflows that we already have work better rather than totally reimagine what we do in the first place.

8:10On that front, Canva's Magic Studio is, I think, a big upgrade. A couple months ago, I did a show about the AI tools that I actually use on a daily basis, and part of my motivation for that was to get people out of the mindset that they have to try every new tool that comes down the pipeline. In that, I featured Canva because it is one of the tools that I use absolutely every single day. So what we're going to do is look at some of the tools that were announced. We'll give them a try. And I think at this point, I should say that while I will try to make this work for podcast format, I also highly suggest that if you're not watching this on YouTube, that might be a better format for this one, as inherently there's a little bit of experimentation here that is not just audio, but visual.

8:50So first, let's cover the news side of this. The way that The Verge writes it up is Canva's new AI tools automate boring labor-intensive design tasks. They frame it as a way to save time editing documents. And given how much Canva is used for photo editing, I thought we'd start there. One of the tools that was actually released some time ago, but which I use every day, is Background Remover. If you've ever seen my thumbnails on YouTube, you'll know that I often have myself in them, or I have individuals that are featured in the videos, and of course they don't have backgrounds. That used to be a labor-intensive manual process.

9:19Now all you have to do is click on the button Edit Photo, and go to BG Remover, and within a matter of seconds, usually two or three, boom, you have the background gone. You can also get more granularity if, for example, there are parts of the image that were erased that you didn't want to, you can restore them. Or if there are parts that you actually don't want, you can remove them. So for example, in the photo of Sam Altman that I'm editing now, maybe I want to take out the Bloomberg reporter's arm, I'll simply increase my brush size, and boom, next up let's try the magic eraser. This is a way to remove elements from a photo that you might not want.

9:50In our case, while Sam Altman being interviewed is the focus, there's an awkward Bloomberg microphone on the left. Let's see if the Magic Eraser can take it out. So we click on Magic Eraser. Once again, I'm going to increase slightly the brush size. And then we're going to follow all the way along and see if it can remove this. Now, it's not perfect. This is a big artifact, and you can see some lingering elements. So let's try on the other side, removing the reporter's hair. It did a little bit better with that. Next up, let's try one of the new features, which is called Magic Expand. This time I've chosen an image of Elon Musk with Grimes.

10:24We go to edit photo and magic expand, and we're going to say whole page, click magic expand, and now we have a filled out background of this image, which I think might have been from the Met Gala. Like with other AI image generation services, there are a variety of options to choose from, and you can also generate new results if you are not happy with what you got. Now, the quality here seems definitely lower than with things like Adobe's generative fill or MidJourney's expand features, but the fact that it exists in suite makes a big difference for someone who's living inside the Canva tool set.

10:53Magic Edit is a feature where you can replace one object with another. Let's try to turn this CNBC reporter's microphone into a bouquet of flowers. So we've selected the area that we want to transform. We describe the edit, a bouquet of flowers, and then we click generate. We then have a set of options to choose from, and boom, now CNBC is offering Sam Altman some flowers instead of a microphone. Magic Grab is a really cool feature that allows you to move things around within an image. So for example, in the Sam Altman photo, instead of taking him out of the background, we could simply move him around in the photo.

11:28And then maybe the way that we get rid of that microphone is just by changing the aspect ratio, so we don't need it there. The last of these image editing tools that I want to try is GrabText. Like Magic Grab, it allows you to take text on the image, move it around, or even change the actual text itself. You can see just from the photo editing suite, this increases the flexibility of this tool in a massive, massive way. Another category of the tool set is called Magic Media. Now this one is one that has people really excited because it's integrated with Runway and is a text to video generator.

12:00Magic Media says, describe a scene to generate a few seconds of video. Let's try something of the season. Pumpkin patch in October at twilight. Generate video. And then sure enough, there is our video of a pumpkin patch at twilight which we can then drag over and size for a horizontal video and build into something that we might put on either instagram or tiktok or youtube shorts magic media also has a text to image generator although you probably have seen a lot more of those so we'll skip a demo of that for the purpose of this video now it appears from my account at least that i have 50 credits for videos per month and 500 image credits per month although i don't know if that's free or based on the plan that I pay for.

12:40Another very cool video editing tool that I wanted to show is their Magic Design, which is the text to design creation part of the suite, but for video. So let's take a set of images from my wife's recent photo shoot for her true crime podcast. And we will say promo, excited promo video for a true crime podcast and see what it comes up with. Let's give it a play. So the video I came back with had a upbeat soundtrack based on probably the word excited, and then effectively a slideshow of the photos that becomes a video, but it also layered on this text. Are you ready for a new true crime podcast?

13:20Join us as we investigate the most shocking cases. You won't be able to stop listening. Available now on all major platforms. Don't miss out on the drama. You'll be on the edge of your seat. This is one you won't want to miss. Subscribe now and get caught up on all the action. You won't be disappointed. Now, of course, is this the exact language that I would choose if I was creating this video? Not necessarily, but have I ever been able to create a video like this in approximately 10 seconds of waiting and five words of description? Also, no. I think a lot of the power of these types of tools is to, one, be able to experiment way more rapidly than you would otherwise, and two, get to some basis that you can use much more quickly.

13:57For example, I'm on the last slide where it says you won't be disappointed. Well, I can decide that that's not exactly the message that I want, So I'll click on Magic Write, and we'll say, for kicks, sprinkle fairy dust. Rest assured, dear one, for enchantment awaits you. Sure, let's go with that. Now the last thing I want to show off is something called Magic Switch. This is a tool that allows for automated resizing and translation. So let's, for example, take a YouTube thumbnail that I created for a show some number of months ago. We'll click on Magic Switch, and we'll resize it at 1280x1280.

14:29So we'll turn it into a square. Copy and resize. Open it up. And really what we've got here is just an expansion of what existed, which isn't perfect, but which does allow me to quickly tweak, move things around, resize them. Again, it hasn't replaced me, but it certainly made things much more efficient. Now let's try translate. Let's give it an easy one and try to translate it into Arabic. Now the problem, of course, is not speaking Arabic, at least not speaking Arabic well. I don't know how well the translation did. But especially as a compliment to all of the autodubbing features that are coming out on things like YouTube, this is a really powerful update.

15:03So let's bring it back and try to just wrap this up. What's powerful, I think, about things like Magic Studio is not just the tools on their own, although individually you can see how valuable they each might be. What's more important is the way that they come together to create a fundamentally different experience that has the potential to radically increase productivity in terms of how fast we get things done. As someone who has to create an enormous amount of content on any given day, not just the podcast you're hearing and the video that you're seeing, but all of the assets that go along with them.

15:34Thumbnails, descriptions, cover images, social promotional images. Having all of these tools in one spot is incredibly valuable. Now on that front, it also is on trend with something that we're seeing in terms of how companies are trying to win the AI wars, and that is taking advantage of existing distribution channels in order to extend moats that already exist. Canva has something like 150 million people who use the service. I was using it, for example, long before they had any AI features. By integrating these types of tools, they're throwing elbows to try to block out startups that might design an AI-first design studio experience from the ground up, because Canva already has such a lead in terms of who's using it.

16:15That doesn't mean there isn't room for partnerships and integrations. We saw in the video generation that Canva is powered by Runway, because at this point, that's not something that they can easily replicate. The last trend, which is the one I mentioned at the beginning, is just this broader idea of the functionality, the practicality of AI coming to the fore in this fall season. We're seeing these tools move from cool and fun to tinker with to things that are actually impacting our jobs, our workflows, and how we get things done that we already wanted to get done. Now, for those of you who are interested in specifically these Canva tools, this will not be the only AI editing and design product that you're going to get in the next couple weeks.

16:49Adobe has started teasing something that they call Project Stardust, which they describe as an object-aware editing engine that does a lot of these same things and potentially even more than the Canvas photo editing tools. It's supposed to be released next week, so we will keep an eye out for that. For now, I appreciate you guys listening or watching as always. Until next time, peace.

17:19Thank you.