Musings on the promises of AI and what it can do for us humans

We end the calendar year with a bunch of coal in our stockings: bankruptcies, rising loan rates, diminishing demand for discretionary goods, and the specter of AI. 

But, before we spike the eggnog, we remember that demand will bounce back — it always does. We remind ourselves that interest rates will stabilize and that America’s economic bloodstream will flow freely again. We seek to learn from the cautionary tales of Mitchell Gold and Klaussner, to name just two industry-bruising belly-ups. 

And we seek solace in the promises of AI rather than mope in the gloom of yet another threat to our creativity, individuality and humanity. After all, a lot of the rhetoric associated with the arrival of AI is reminiscent of circa 1995 and the angst regarding the World Wide Web. 

Man in the Mirror

I teach a course in media law and as part of that course this semester we organized a visit to our campus by AI and copyright expert Matthew Sag, on faculty at Emory University’s School of Law. The native Australian brought a clear-eyed sense of the promise and perils of AI that helped me better understand what it is and and why we should probably begin figuring out how to regulate it.

Sag admonished us not to be angry with AI because its quirks, prejudices and biases all came from us. AI trains on data we made ourselves. And AI’s appetite for data is breath-taking, such that the demand for ever more data will continue to mushroom. 

The reason that we’re all talking about AI these days is rather simple: computing power. Huge advances in the speed of computing and in the chips that supercomputers run on are fueling the rapid rollout and adoption of machine-learning. This vastly improved computing power is being applied to all sorts of problems, often accompanied by stunningly sophisticated graphics. 

AI is being trained on more than 15 million websites and hundreds of millions of copyrighted works, Sag told a nearly packed house at Berry College. Billions of images are being used to train MidJourney, Stable Diffusion, DALL-E and other image generators. 

But, the machines learn at an abstract level, seeing abstractions rather than copyrightable expressions. Outputs can, therefore, be novel and surprising. 

Wanna Be Startin’ Somethin’

We’ll look back on 2023 and, with respect to AI, to May 25. On that one trading day on the Nasdaq, the AI pioneer Nvidia saw its stock value rise by $200 billion, one of the largest gains in a single day stock trading had ever seen. The big day coincided with news that ChatGPT trains on a supercomputer running on chips from Nvidia, according to Stephen Witt’s account in the New Yorker magazine, “The Chosen Chip.” By the closing bell, Nvidia had become the sixth most valuable corporation on earth, putting it ahead of Walmart and Exxon-Mobile . . . combined!

In a rare interview, which Witt writes about, Nvidia CEO Jensen Huang is quoted as saying, “There’s the doomsday AIs — the AI that somehow jumped out of the computer and consumes tons and tons of information and learns all by itself, reshaping its attitude and sensibility, and starts making decisions on its own, including pressing buttons of all kinds.” 

He nails our general sense of angst about this stuff. But just when we might be tempted to blow it all off as just another over-hyped phenomenon, Huang adds, “Reasoning capability is two to three years out.”

Sag told us that he sees “a great deal of evidence that the differences between AI and humans will persist” because, among other reasons, the AI model has no internal governor. Humans do. 

“I can lie, for example,” Sag said. “I can impart my intentions to my dog or my Roomba.”

Then he asked a provocative question the answer to which I haven’t fully resolved: “How many of you have queued up to see the Mona Lisa? Would you queue up to see a perfect copy?”

The answer is, of course, no, we wouldn’t pay for tickets for the Louvre, then stand in line for a few seconds before Da Vinci’s masterpiece, precious time most tourists use merely to look through their iPhone camera lens to snap a pic. But generative AI doesn’t traffic in perfect copies; it creates new things by learning from the images already created. I agree with Sag’s main point, that there (still) is a human factor.  

Human Nature

For good reason, there is a tremendous amount of concern about AI’s potential impact on creative fields, including all of the knowledge industries, communication, writing, art and computer coding. These fears might be warranted, but for the near term, AI promises to be more an unprecedented helper than a replacement. The Skynet of the Terminator movies still is a long way off. 

AI as a helper still relies on human creativity, human expression and human skill in directing it as to what is wanted and how what is wanted should look, which is another way of saying form and function. 

Reflecting on Huang’s sobering prediction and Sag’s subtle optimism, I conjured a help wanted ad we’re likely to start seeing versions of in 2024, perhaps even in home furnishings, an industry not typically associated with cutting-edge technologies and supercomputing. 

Help Wanted: Prompt engineer. Job duties include engineering complex text prompts to guide generative AI platforms in the creation of a range of specific visual and textual outputs. Desired outcomes include tailored, high-quality images that match a designer’s creative vision and, in automating creative tasks, creating new efficiencies throughout the organization. Requirements include fluency in the terms, styles, specs and “language” of home furnishings, interiors and architecture. 

I’ve begun using chatGPT for fairly simple tasks, such as appealing my county’s property tax assessment (it worked!), summarizing Zoom calls and generating meeting minutes, and writing drafts of letters of recommendations for people I don’t know all that well but well enough not to turn down their requests for some help. The benefits are immediate: saving time, preserving precious brain cells for more rewarding activities, and overcoming the inertia of lift when starting an iterative process.

Perhaps the biggest benefit of all, however, is not the machine-learning, but the human-learning. As I use the technology, my prompts improve. I’m (slowly) learning which kinds of prompts produce the best results, with the main takeaway being specificity. Unlike Google’s search box, which rewards concision, chatGPT doesn’t mind a lot of parameters. 

Generative AI is “going to make us all more productive, and it’s going to make new forms of research possible,” Sag said. “GAI will also be disruptive. I lived through this already with the World Wide Web. And there have been other precursors: the camera, the phonograph and the spreadsheet. We thought the advent of the spreadsheet would eliminate the need for accountants. We believed that the camera would put a lot of artists out of business.”

Speed Demon

I walked away from Sag’s presentation (and a lovely dinner with him and our three panelists) thinking that perhaps a helpful analogy is driverless cars. I recently bought a car that I joke is (much) smarter than me, anticipating events before I can, providing a steady stream of real-time data, and most wondrously of all making the morning commute a mostly pleasurable experience (thanks, Harman Kardon!).

Yes, there have been some problems with driverless technology resulting in crashes; there are bugs to iron out. And, yes, no technology is jerkwad-proof. It took a deep fake video generated by AI to inspire President Biden’s executive order seeking to install guardrails on the technology. (Biden’s staff showed him the deep fake to make the point that an AI system could create a presidential statement that never occurred, perhaps even touching off a national security crisis. “When the hell did I say that?” Biden asked them.)

We will need rules, and we will need regulation, rules and regulation that change as we learn the dangers of the technology in the many wrong hands that inevitably take something meant for good and immediately begin weaponizing it to terrible harm. Why should AI be any different?

OK, now pass the eggnog! It’s time to get this party started.

Editor’s note: On this, the 40th anniversary of Michael Jackson’s groundbreaking Thriller album, we thank the King of Pop for all of the subheads in this column. 

Brian Carroll

Brian Carroll covered the international home furnishings industry for 15 years as a reporter, editor and photographer. He chairs the Department of Communication at Berry College in Northwest Georgia, where he has been a professor since 2003.

View all posts by Brian Carroll →

2 thoughts on “Musings on the promises of AI and what it can do for us humans

  1. Brian – only you could have written such an article that combines true insight and wit and make it relevant to ” . . . home furnishings, an industry not typically associated with cutting-edge technologies and supercomputing. “

    1. So good to hear from you, Jeff. Thanks for the kind words, and happiest of holidays to you and your family. (I just asked chatGPT how to best string lights on a tree!)

Leave a Reply

Your email address will not be published. Required fields are marked *

Subscribe to our Newsletter for breaking news, special features and early access to all the industry stories that matter!

Sponsored By: