Introducing the BrXndscape // BrXnd Dispatch vol. 004
On generative AI, landscapes, and a new BrXnd experiment
Hi everyone. Welcome to the BrXnd Dispatch. If you’re wondering what this is or how you got signed up, it’s a roughly-weekly newsletter of interesting stuff at the intersection of brands, AI, and creativity. You likely opted-in on BrXnd.ai. If this is not interesting to you, feel free to unsubscribe. Also, another reminder, NYC spring 2023 Brand X AI conference planning is in full effect. If you are interested in speaking or sponsoring, please be in touch.
This week I’m excited to announce the BrXndscape, a landscape of generative AI companies working at the intersection of brands, creativity, and artificial intelligence. Right now, the list of companies stands at 112, though there are surely many I’m missing (if you want one added, you can submit it here).
A few of these landscapes have come out recently—so what makes this one different?
First and foremost, it’s focused on brands and marketing.
There are obviously many other use cases for AI across the rest of an enterprise, but the lens I’m most interested in is how this technology will change the way brands go to market. To that end, the taxonomy and company listings are intended to help answer questions marketers have about how to take advantage of these new technologies. I’ve designed the taxonomy to assist marketers in identifying the tasks they need to accomplish, and the category and company pages go into depth on just what these tools can do for brands.
This brings me to my second differentiator: the information I’ve gathered about these companies. Each company page includes a bunch of data about what the company does and how the product and billing model works. As an example, here is a set of features for the tool Writer:
Note the disclaimer at the bottom of the features section. This was built in part using AI. I’ve developed what I consider some pretty neat techniques for scraping and text extraction using GPT3, and I’m putting them on display here. Is it perfect? No. Is it amazing that I could stand up a landscape with this much info for 112 companies in a few weeks? I certainly think so.
I’ve also incorporated pricing info, which, like the features, is extracted with the help of AI. That becomes particularly interesting on the category pages, where you can quickly compare plans across multiple companies. (I’ve chosen one plan per product as a representative.) Here is the Generative AI Logo Design pricing summary:
Another thing that’s different about Brxndscape is the natural language use case search. This is probably my favorite part of the whole app. When you land on the homepage, you can simply type in the marketing use case you have, and it will crunch everything it knows about these categories and companies to suggest the best set of tools for you to accomplish your goals.
Here’s a quick example:
I think marketers will like the way this works. Every company and category in the landscape has a set of use cases (based on the website scrapes and generated with the help of AI). I then used GPT3 embeddings and Pinecone for vector storage and search. The amount you can do with this stuff right now is astonishing. If you’re not familiar with embeddings, it’s definitely worth getting acquainted with, particularly since OpenAI updated the model and brought down its pricing in December. As opposed to using GPT3 (or ChatGPT), which works on completing sentences and answering questions, embeddings give you the raw vector values for whatever text you ask for.
To understand what this means, let’s imagine it works in two-dimensional space (it doesn’t—it is working with many, many times that number of dimensions). If I tell it that one use case of WellSaid, a text-to-speech product, is to create a podcast from blog posts, it will give me a set of dimensions that locate that “idea” in vector space. Again, in our two-dimensional toy version, let’s pretend it looks something like this:
Each dot indicates a different company/use case. They are situated based on the vector values assigned. When you search for something, it gets its own location in space, and that can be compared to other clusters to show relative similarity and closest neighbors. This is a toy because the real version happens in 1500+ dimensional space, not two, but hopefully, it makes it a lot more comprehensible. It’s an incredibly powerful tool, and it works because of all the hard work that already went into training the model off the huge corpus Open AI works with.
There’s more to it, but that’s a good place to start. Go have a click around, and let me know if there are companies to add or if you have any feedback.
Other BrXnd Goings On
NYC 2023 conference planning is in full swing. We should have a venue locked down in the coming weeks and start to put together programming. I also expect to announce of some of the fun things we’ll be doing in the next few weeks. If you have ideas for speakers, sponsors, or attendees, please let me know. Happy to send over the sponsor one-sheeter if anyone’s interested.
Brands X AI Around the Web
I’m very much with Benedict Evans on this one:
For the most part, search already works this way. We ask questions and get answers. Often those answers are right there on the page, and we don’t even need to click. But even more than that, my own experience with these tools is that contextual help is much more powerful than going off and asking things. The real competitor for Github Copilot isn’t Google, which is losing a few searches I might have made for how to write some specific function, but rather, Stack Overflow, which is losing the traffic that would have come from that search. Not surprisingly, Stack Overflow banned GPT answers. They say it’s for quality, and I’m sure that’s part of it, but it’s more fundamental than that. Sites that are reliant on Google for discovery, based primarily on questions, seem likely to be the most at risk from what’s to come.
I missed this Heinz campaign from last summer, but I love it so much. Here’s the gist:
Heinz has proved that even computers prefer its ketchup with a marketing stunt that had OpenAI's Dall-E 2 generator create a series of sauce-inspired images. Apparently, when the team fed the software random ketchup-related phrases, the results were overwhelmingly plastered with elements of Heinz' signature branding. We have to say, this AI art is some of the least weird we've seen, even with the ketchup bottle floating in a swimming pool.
I have lots of thoughts about this, which I’m going to reserve for a future email. But it’s definitely been my experience that the stronger your aesthetic identity and visual assets are, the better your brand is represented in AI.
AlphaSignal is a weekly email with research from ML/AI generated with the help of AI.
Finally, this is a post worth reading on why generative AI “is mostly a bad VC bet.” I generally agree with the questions about defensibility here, but there was one part in particular that caught my attention:
When machines allow your team, or anyone on the internet, to easily create tens of thousands of images related to your brand, how do you monitor and protect that? How do you ensure brand consistency in your own workflows? How do you spot imposters and fakes when they are easier than ever to create?
There are a few dimensions to this question. Obviously, making sure your brand isn’t faked is important, particularly if products are being produced and distributed. But the other side is that the brands that get faked are the ones that are really good at being brands. You can fake a Louis Vuitton bag because it’s consistent and iconic. My experience doing a lot of experiments making branded images is that the best brands come out best—meaning that, to some extent, they need the least protection. Obviously, you don’t want people believing the brand made something it didn’t, but it also means there is a much lower likelihood that something particularly off-brand will come out of the AI.
That’s it for this week. Thanks for reading. Keep sending links and join the Discord, the conversation is just getting started!
Talk to you soon.
— Noah