Welcome back to AI Insider, the newsletter that owes its existence to my Facebook friends who asked for it. And if you’re new or not here via my FB page, welcome to you, too. Be sure to visit the archival website for an intro to who I am and an explanation of why chatbots have a byline here. Like what you see? Donate to support this newsletter. OK, let’s get on with it!
Someday, We’ll LOL at Today’s AI
By Michelle Johnson, AI Insider
Sometime in the not-too-distant future, we will look back on this era of AI and laugh. WRITE a text prompt to tell a bot what to generate? How quaint!
Seriously. We’ll look at AI tools from today like they’re a VHS tape or a Walkman. Imagine a future where telling a bot what to do by typing a text prompt seems as outdated as rewinding a cassette tape. Sounds crazy, right? But the pace of change is so rapid that it’s not entirely out of the question.
So, what will we be using in the future? I have no idea. If I knew, I’d be busy buying stock.
What does seem clear to me is that some trends emerging now may still be around in the future. For one thing, there’s clearly a push by companies to embed AI into all or most of their products. It’s not too hard to envision this “AI Everywhere” concept hanging around.
So, yes, there will come a time when it will be second nature to chat with any application, device, or personal gear that’s in your orbit. Or not. Maybe things will become so smart that they perform tasks without us having to ask at all.
Your coffee maker might know how strong you want your brew, or your self-driving car would know to take your favorite scenic route—all without you having to lift a finger.
While today’s AI might seem cutting-edge, don’t get too attached. What’s coming is bound to make today’s tech look quaint.
Meanwhile, if you have any stock tips, lemme know.
AI Takes Center Stage at Adobe MAX 2023
[Ed note: Like many other big tech companies have done at this year’s annual conferences, Adobe announced a slew of AI-related updates this past week. The focus (see below) is on embedding AI across their products. Below are highlights. If you want to dig into a fuller list of what’s new, watch this video.]
By Bard and Claude for AI Insider
Adobe Max, the company’s huge annual conference, just wrapped up in LA. Not surprisingly, this year’s big focus was on artificial intelligence (AI) and how it can help creative types work smarter and faster.
Over 100 new AI-powered features were introduced across Creative Cloud applications, including Illustrator, Photoshop, Lightroom, Premiere Pro, After Effects, and Adobe Stock. These features aim to automate mundane tasks and provide assistance to users as they create.
Adobe teased a bunch of new AI-powered features coming to apps like Photoshop, Premiere Pro, Illustrator, and Audition. One that got some buzz was “Context-Aware Editing” for Photoshop. This tool uses AI to analyze images and suggest tweaks and fixes based on what’s in the picture. So if you have a person in the foreground, it’ll know to enhance them rather than the background. Adobe says this tech is still in development but has major potential to change editing workflows.
A Few Highlights:
Text-to-Vector Graphics: Powered by the Firefly Vector Model, Adobe Illustrator now generates editable vector graphics from simple text prompts. (See item below for more.)
Text to Template: Adobe Express can auto-design templates from text descriptions, making it easier to create professional-looking content. It pulls from Adobe stock images and fonts which are cleared for commercial use. Express is free.
Video editing by Text: Adobe’s Premiere Pro editing application will auto-generate a text script that can be edited simply by selecting text in the transcript and then dragging the edit as a clip onto the timeline. A filter will automatically find and remove “ums” and “ahs”, and a new audio tool will clean up bad audio. (See below for more.)
And finally, Firefly, the standalone AI image editor, has been upgraded. Firefly 2, purports to generate more realistic images and has new camera adjustment settings such as aperture and depth of field. It will also automatically suggest prompts as you type.
Hands-on With New Adobe AI Tools
By Michelle Johnson, AI Insider
Full disclosure: I don’t use Adobe Illustrator. I’ve always found this high-end, fancy-schmancy drawing tool kind of intimidating. And since I’m not exactly a “creative” type who designs logos or book covers, there’s really been no need for me to dip a toe in and draw something from scratch.
But when I saw Adobe’s announcement that you could generate something called a “vector graphic” just by typing in a text prompt, I decided to take it for a spin.
Following the Adobe Max presenter, I prompted Illustrator to give me a “puppy” graphic. Here’s the result.
Not bad! Could I draw this myself? Nope.
Once again, I have to point out that for people with the knowledge and skills, these AI workflow shortcuts must be a godsend. In the right hands, lots of great work will get done faster.
For the rest of us, this will open up a world that we may have shied away from and maybe spark a little creativity in us non-creatives.
So, what’s a “vector graphic?” See how when I select the graphic below it picks up each part of the image? That means that every piece is editable. It’s easy to recolor, resize, delete, and make other changes. Or to scale it up or down.
If you have Adobe Creative Cloud, check to see if your subscription includes Illustrator. Then, tell it what you want to see and watch what it generates.
I also took a look at the new features in Adobe’s high-end video editing application Premiere Pro. I had a video of my niece’s outdoor wedding that had a bunch of wind noise in the background, so I wanted to see if it could be fixed. Premiere’s audio “Enhance” tool made quick work of toning the wind noise down. (In the past, I’d be all over YouTube looking for tutorials on how to fix this.)
I also loved being able to edit video clips by selecting text in the auto-generated transcript that appears when you open a video. Once you select the text, you can just drag an edit onto the timeline. All in all, I was very impressed with the ease of use. Two thumbs up.
Finally, I hit the standalone version of Firefly, the AI image generator that I’ve been using since it was launched as a beta back in March. There’s a lot that I like about Firefly, including ease of use and the fact that Adobe trained this image model on their own stock photos. (No questionable pedigree here.)
I was shocked to open the text-to-image feature in Firefly 2 and spot a new section with settings that replicate actual camera settings, such as aperture and shutter speed. (See image above.)
Adobe swears the new image model generates more realistic images, but in a quick test, I found it still pushing out mangled faces. When it did hit the mark, the results were impressive, but there’s a lot of competition in this space. You could do just as good a job with Bing, powered by Dall-E 3 or the new Ideogram.
That said, Firefly is often my go-to image generator, and with new features such as sending images directly to Adobe Express for use in things like social media posts or print products and easy removal of backgrounds and other distractions, it will stay near the top of my list.
Free Stuff!
OK, now that we have your attention….
Remember a previous newsletter mentioned that ChatGPT now has “eyes,” meaning that you can upload images and ask the bot about ‘em?
Well, if you’re not willing to part with $20 a month for ChatGPT Plus to prop open ChatGPT’s eyes, here’s a FREE alternative, LLaVA: Large Language and Vision Assistant [Ed note: Judging by the URL, this may be some computer science major’s class project. If you’re ok with that, give it a try! I’d give them an A.]
How does it work? You just upload an image and then ask the bot questions about it. Here’s an example. I asked: “What’s unusual about this image?” — MJ
What the Actual Hell?
See, this is why we can’t have nice things. Stories like this were the talk of the interwebs this past week: Humans can’t resist breaking AI with boobs and 9/11 memes (TechCrunch).
Yes, that’s right. Meta and Microsoft discovered the hard way that people just can’t be trusted. Users got busy generating questionable stuff with their new AI tools in no time flat.
Meta (aka Facebook) is rolling out a sticker feature. I’m sure they envisioned people making butterflies and rainbows. Yeah, right. Things got sophomoric fast.
Microsoft upgraded its image generator with Dall-E 3, and “creative” types started generating memes showing fictional characters involved in 9/11.
And no, we’re not showing this high-end artwork here. Click the link to the TechCrunch story if you really want a look. — MJ
Random Shorts
Google’s Generative Search Experience has expanded its horizons to image generation and writing. This means when you pop a query into Google, not only can it provide text-based answers, but it can also create custom visuals and written content on the fly. Bing Chatbot has been doing this for months and is likely not impressed.
Dropbox dropped Dropbox Dash and Dropbox AI. You can now search across tools, content, and apps from Dropbox Dash in one place. With Dropbox AI, you can get summaries and use natural language queries (aka, plain English) to ferret out stuff in text, audio, and video files.
Have you tried Zoom’s AI Companion yet? Just what you need: something else on your Zoom call. In this case, “a smart digital assistant that can help you get more out of your meetings and emails.” Zoom AI Companion can generate questions, summaries, next steps, and suggested content based on your conversations and other content. And more.
Aht Gallery
One of the interesting things about AI image generators is their ability to mimic any style of photography. This antique style is called an Ambrotype. An ambrotype is a 19th-century photograph created by underexposing a glass negative and mounting it against a dark background, resulting in a positive image with a soft, ethereal quality, according to ChatGPT. Prompt: Ambrotype photo of an African American family of three and a dog. (Michelle Johnson via Midjourney, Leonardo AI, DreamStudio)
