155: Mixed Bag
155: Mixed Bag
The picture of AI "agents" gets slightly clearer; AI art and Copyright
If I had to guess, I acquired Covid at the Soul Coughing show lat week. Ironic, indeed.
I am very grateful to my wife for keeping me fed while I’m in isolation. This bout isn’t as bad as the one in May 2023, but if you can avoid it, you should.
A mixed bag of two subjects today!
“The smarter question is to ask ourselves how ‘AI-as-colleague’ is going to open up new ways to imagine our work patterns.”
On Monday OpenAI hosted the first of three Dev Day events in 2024, announcing improvements to APIs, cost-savings and fine-tuning improvements across its models. But the important news was in Sam Altman’s closing comments focused on AI agents. “These are independent artificial intelligence models capable of performing a range of tasks without human input,” reports Tom’s Guide. Or as Altman put it, “People will ask an agent to do something that would have taken them a month, and it'll take an hour.”
Imagine an agent which: 1) Does research you’ve outlined, 2) Summarizes and tests its summary for characteristics or qualities you’ve selected, 3) Builds a presentation - much like Microsoft and Google suggest, including speaker notes read by one of its voices, 4) schedules time on coworker’s calendars to watch the presentation, then 5) summarizes their feedback and makes adjustments to the presentation. While you do the more important work.
All this is speculation, of course. Agents have yet to fully emerge but the groundwork is there, inside the “reasoning” capability of ChatGPT o1. So I fully agree with Nate Jones, whose quote I headlined above. In this second year since ChatGPT arrived, we’ve moved incredibly fast through single-tactic AI, multimodal AI, and soon a space where AI can do multiple tasks in sequence.
How will we function as individuals, teams, and organizations when “AI-as-colleague” becomes functional reality?
Neil Perkin wrote about what this means last week.
“All the surveys suggest that GenAI tools are currently being used in pretty executional ways to optimise existing tasks and deliver greater efficiences, and there’s a ton of value in that. But as we become more sophisticated in our interactions with AI tools, and as the tools themselves develop, there is a different opportunity for AI to challenge our thinking, to help us to take creative leaps forwards, or to reframe something in a totally different way.”
And if you prefer a more technical deep dive on how AI agents might impact enterprise systems, Silicon Angle has you covered with their Breaking Analysis. They’re bullish.
“The extent of the [AI agent] transformation, we believe, will be more impactful to the application stack than were the changes brought about by innovations seen during the modern graphical user interface, Web, cloud and mobile eras.”
This all looks like technology but it’s really a cultural question.
“The U.S. Copyright Office will register an original work of authorship, provided that the work was created by a human being.”
You may recall Jason Allen, who was denied copyright protection by the U.S. Copyright Office for his Midjourney creation, “Theatre D'opera Spatial” back in September 2023 (view the artwork at that link). Allen has appealed. And it’s worth looking into some of his claims seeking revised copyright standing.
“Creating an image using the Midjourney tool requires more than the bare minimal mental effort. In reality, it is a tedious, complicated, and often frustrating endeavor.” Agreed 100%. No notes.
But the gist of Allen’s complaint is effort. Which the Copyright Office defines as “the human authorship requirement.” Allen asserts:
“The prompt serves as the initial creative input, embodying the human’s original idea and vision for the artwork. The AI system functions merely as a tool to assist in the actualization of the human’s creative ideal, akin to a cameraman, executing the human-conceived “director’s” prompt to generate the final image. Indeed, just as a photographer may take hundreds of photographs, discarding all but the single one that realizes their original idea, Mr. Allen issued more than 600 prompts before generating one that is sufficiently consistent with his conception of what the final product should be. Even then, Mr. Allen had to make individual adjustments outside of Midjourney in order to realize his ideal final product.”
I mean, maybe? I don’t envy the Copyright Office, but I am curious to see how this plays out. If you take Ryan Broderick’s view, a person could, “essentially copyright everything, infinitely, forever. Even random assortments of colors. It would fundamentally break copyright as a concept.”
Stakes is high.
🤖👍🏽 Alberto Romano’s “30 Things I’ve Learned About AI” is worth a look.
🤖🤔 “Yet technology for its own sake counts for nothing. What matters is how it feels to people and what impact it has on societies,” writes Microsoft in announcing it is, “creating an AI companion for everyone.” That, or they’re promoting an update to Copilot across web, mobile and desktop.
🤖🚕 “Look, I know this is objectively ridiculous,” notes Ben Thompson describing his emotional first Waymo ride. I don’t think he’s wrong at all.
“The emotion was really quite powerful! It makes me think that as robots proliferate in daily life that humans are going to develop attachments; when a favorite robot breaks down or needs to be replaced people are going to be sad. I’m not totally sure why this is, but I suspect the verbal aspect of the interaction plays an important part. Communication is meaningful, and the capacity of devices to communicate in a natural way with humans is, well, humanizing.”