039: This is just the beginning
[Before - Session 13] I've been exploring knowledge management in the age of AI
We’re in that 24 hour period before a presentation. You know the feeling. I know the feeling. I suspect my students definitely know it. I’m excited to see what they’ve developed; and I’ll share excerpts in tomorrow’s “During” post. While they’ve been busy working on Assignment 1, I’ve been exploring.
Key insight: Knowledge Management is integral to the future of generative creativity. We’re all about to get a lot smarter about ourselves and our organizations, thanks to AI.
But let’s start with something fun like the legalese of art…
Generative Art and Legal Rights
Portions of Adobe’s Firefly have opened up. I’ve been playing around with their AI text-to-image tool. The interface is definitely easier to use than Midjourney. Notice the ability to toggle (lower left in images) between a reference image and the prompt. Who actually knows what that toggle means? 😆 Regardless, we get to explore and that’s the point.
I’m less impressed with the output versus Midjourney. But as we’ve seen with AI, the training data and models improve rapidly. So too will the output.
Adobe’s key distinction in this space has been its approach to rights. Where Midjourney, Dall-e2, and Stable Diffusion have been opaque about their training sources and offering zero guidance re: commercial use within their interfaces — Adobe is definitely trying to appeal to agencies and marketers. Every business affairs, sourcing and legal role is asking “can we own [the thing you made generatively]?” Adobe is attempting to answer.
Note this interstitial when you first log in to Firefly:
And note the label Adobe adds to downloaded images (see my cover image above, note the lower left corner).
…related
We’ve all been crawled. I’m ranked a mere 6,976,997th in the long list of sites crawled for Google’s LLM training data. As the Washington Post explains, “the training data for OpenAI’s GPT-3, released in 2020, began with as much as 40 times the amount of web scraped data in C4.” Where did GPT get its source material? You and me. And everyone else.
Adobe has also revealed text-based video editing within Premiere. They’re clearly playing catch up with Descript. But the process of making Art continues to become easier and easier. Video “editor” is now a skill accessible to a broader swath of humans.
And, D-ID’s AI-generative presenters can now integrate with Canva, i.e. Build your presentation and embed an avatar who conveys your talking points and subtext. Tutorials are becoming much easier for anyone to build.
Encoding and enabling real life
Now imagine augmented tutorials and applications applied to real life. That’s what Vova Kurbatov demonstrates in the video linked above, i.e. AR recognizes a book and calls out the highlights you made, it reminds you of your watering schedule for a plant, helps you adjust room temp, etc.
I think this will be the short term benefit of AR: Annotate real life, and leverage gestures to enable interactions and data collection. (My prediction: This is what Apple will attempt to reveal with its initial AR-wearable.) Now apply this capability to someone interacting with a machine for the first time, or getting off the plane in a foreign country. Then add audio. And mix all of the above with a GPT focused on training or travel.
The future of knowledge management
One of my favorite writers is the culture pundit Bob Lefsetz. He offered sage perspective on AI last week. And then (as he frequently does) published responses from readers. Including this from author David Meerman Scott (bolding mine):
“Nearly everybody is missing the bigger opportunity for creators with AI - applying the tools to our own content instead of relying on public web content. Imagine the power for a musician to upload all their work, released or not, to a proprietary database that they control. Then AI tools can help them to generate more music from their own existing work!
I did that with my writing. I created a database with nineteen years of blog content, over 1,500 individual posts plus six of my books and I use a closed ChatGPT interface. I can create new posts from my own work, summarize my own ideas, craft email responses based on my words, and much more. The future of AI is using it with our own stuff.”
This is a huge idea.
And not just for individual creators. It speaks to the future of Knowledge Management. Consider the entire corpus of PDFs, emails, videos, chats, PPTs, Excel docs, etc. lingering inside your company, going back however long your retention policy allows. What if every bit of dialogue in a video, or presenter notes or comments inside a PDF were immediately accessible?
Firms like Minneapolis-based Lucy have seen this future and are positioning themselves to educate and enable organizations to unlock their own wisdom.
Microsoft is promising as much with Copilot.
Knowledge Management, Part 2
One of the keys to unlocking specialized knowledge management will be an ability for individuals and firms to spin up generative tools quickly and securely. Oh, is that you, Amazon? In announcing Bedrock and related ML toolkits, we’re seeing AWS being positioned as a “Switzerland” of Generative AI. As the blog post puts it, “customers get really excited because these generally capable models can also be customized to perform domain-specific functions that are differentiating to their businesses.”
Again, the key is training the AI model on your data only, and securely. See how lawyers keep showing up in these tech-future narratives?
Anyway, as Meerman Scott suggests, now an organization can summarize itself—unlocking domain expertise across an enterprise. Think of it this way: Expertise retires or quits every day. What if their comments, responses and notes could be leveraged long after they’ve left the building? Or the insights and ideas across thousands of employees could benefit the work of them all?
“Ideas that are distributed across a large network of minds can’t be lost. Instead, they accumulate and become more complex than any individual could grasp.”
Knowledge Management, thy name is Cultural Ratcheting
This 6 minute film by Joss Fong and Áron Filkey for Space10 is definitely worth your time. Creativity in the Age of AI is essentially about knowledge management, just through the broader lens of civilization and creativity.
“The output from a generative model isn’t the end of the process. In skilled hands, it’s just the beginning.”
And if the generative model is trained on your own data, or your organization’s data, you can have greater confidence in your abilities to take action.
Last but not least
Kudos to the One Club and TikTok for spinning up a 20-week, part-time, upskill training program for content creators in cities across North America. ONE Creator Lab launches this July. They’re looking for in-house and external agencies to engage, too. Love the free/pay-it-forward model.