Adobe Firefly
What is Adobe Firefly? Adobe Firefly is a generative AI model, a set of models actually. Adobe's been working on for years. They've been partnered with Nvidia from about the time Bitcoin and crypto mining started to decline. Nvidia realized that the next shift that was going to happen using their products was going to be something like people are no longer buying GPUs to buy Bitcoin. They're going to start buying a 100s to train AI models and that turned out to be exactly right. The interesting thing is, most of these large language models and generative AI image generation models don't have any particular secret sauce. It's just a large training set of images paired with text or large language models just text with classification and some human feedback training data. But there's no secret technique. We all know the techniques. You just need a boatload of money, a boatload of people and a boatload of training data to produce something like this, that's why not very long after Dolly was released publicly stable diffusion and mid-journey pop up. They don't pop up out of nowhere. They pop up because anybody who can read open AI's research papers and understand them and get the time and resources and data together can replicate that on a few million dollars worth of GPUs. So Nvidia, Adobe, and about a dozen other companies have been in this partnership for a couple of years now to figure out how to generate AI services revenue basically for Nvidia on top of the hardware they build as they transition to become a services company along some hardware company. Now this is happening in secret. If you dig into it, you can find it. They have press releases. They have information on their website about all this. But anyway, how it affects Adobe Firefly. This isn't so much as a competing model for Dolly or stable diffusion. They're not going to sell API access to this. This is a transition that every company is going to go through. New space is going to be created, new market. It's not competing for market share with stable diffusion or open AI, they're not competitors. Every company in tech that exists is going to have to figure out how to either build or buy and integrate with AI technology because just like when mobile came out just like when the cloud became a thing, this is a competitive advantage that's going to wash over the entire market and you either get out competed by a competitor who employs that technology first or you figure out how to build it into your own products if you're an incumbent. This is exactly what Adobe is doing. They have a lot of smart product managers who can read the writing on the wall and see where the competitive vulnerabilities are in their product. Most products that Adobe make are for creative professionals and so the first applications of the Sensei AI are going to be things that empower creat ives to do things they couldn't before. This isn't replacing creatives. This is augmenting creatives. Think Adobe stock. This is thinking they have a library of millions of images. If you need an image that doesn't currently exist, you can either hire somebody to take that photo or a sufficiently advanced AI model can create it on the fly and that's opening up a new line of business, Adobe effectively competing against itself to build a better product before somebody else does. What does that mean later, though? Could you imagine a version of Photoshop paired with a large language model and a more advanced version of something like Stable Diffusion that could do in-painting and photo editing on the fly without a person that really understands how to drive it or control it? Imagine being able to speak or type with a Photoshop-like product that you didn't have to learn, you didn't have to watch tutorials for, you didn't have to remember where 80 different commands were in a complicated command palette or file edit menu type of situation. That's basically what's becoming possible and if Adobe doesn't build it, somebody will build it and compete for market share with Adobe, so from Adobe's point of view, if you can figure out how to build elements of that, you become that competing product before it even has a chance to exist, which is a pretty logical thing to do. [BLANK_AUDIO]
Stay updated
Get notified when James Futhey publishes new posts.