With new and more advanced artificial intelligence, what’s old is definitely new again, and everyone loves a comeback story. Then again, based on the vibe at the SXSW festival currently taking place in Austin, Texas, the resurgent interest in artificial intelligence might just test that notion.
Since the festival’s opening on Friday, AI has been the top tech on everyone’s lips, thanks to a year of surprising advancements that pulled off what seemed impossible — it made people forget about the overly hyped, but ultimately disappointing versions that came before via overly stiff, easily vexed bots. The latest wave of AI in its sophistication is a different breed capable of understanding more, communicating naturally, learning as it goes and generating brand new works, even photos and art.
It may be no surprise to data scientists that machine intelligence can resemble human intelligence this much, not in some far-off future, but now. For everyone else, especially anyone who ever struggled to make Alexa understand a command, skepticism is the only logical route, at least until there’s a chance to check it out firsthand.
This is how OpenAI’s ChatGPT, a bot designed to hold human-like interactions, went from quiet launch for a free app to internet famous. Social media drove an astounding number of users, turning the debut into a watershed moment for AI that proved the public’s interest.
The developer’s SXSW session on Friday ran through the numbers: Within two months of the debut in November 2022, ChatGPT saw an unprecedented rush that broke 100 million users. Meta, then Facebook, took more than four years to hit that figure in 2009, although smartphone usage was far less prevalent at the time. A more recent example, TikTok, crossed this threshold much faster, but it still took nine months.
Greg Brockman, president of ChatGPT’s parent OpenAI, seemed as surprised as anyone at the explosive interest. That’s not a dig at his own product, and the company packed plenty into it, starting with deep learning algorithms called large language models. LLMs do what the name suggests: The bot ingests massive volumes of data to speed up natural language processing, so it can learn the way people speak more quickly. It’s fascinating, but the models aren’t, Brockman explained, as OpenAI had already built the tech nearly a year prior.
The app also wasn’t the first AI release either, certainly not in the company’s eight-year history, and not even during 2022. AI image-creation tool Dall-E 2, an upgraded version of the original Dall-E from 2021, beat it to the punch by launching earlier last year.
The app’s popularity probably comes from other details, he said, given the attention paid to the user experience. The team did extra work to make it accessible, going with a simple interface and ensuring it was easy to talk to or command. It was also free, which essentially threw open the doors, so people could try it out and judge for themselves.
“I think that the thing that was very interesting was that, as this app really took off and people started using it, we could see the gap between what people thought was possible and what actually had been possible for quite some time,” he explained.
That matters to Brockman, because the entire AI field, which includes OpenAI, should keep people well-informed about the possibilities. That’s the foundation for crucial and deeper conversations “to figure out how to absorb [AI] in society,” he added. “How do we get all the positives? How do we mitigate the negatives?” This spirit of collaboration also continues through the technology, since OpenAI introduced a new ChatGPT API in early March so other developers and businesses can bring AI-powered natural language chat to their own sites and apps.
Pundits, analysts, politicians, top technology executives and more have been weighing the same thing across ethics, equity and a multitude of other factors for years. Google, one of the biggest developers of artificial intelligence, even created its own ethics council in 2019 to guide its decisions and development for AI.
But there’s little to show for it, especially from the tech giant’s “blink or you’ll miss it” advisory panel. In an infamous pivot, it disbanded the group after just one week due to a political controversy over one member. Pesky humanity got in the way of the machinery again: An internal AI ethics researcher tried to fill the gap, only to learn that Google doesn’t actually like to be criticized for its choices, or so reporting at the time described the 2021 dismissal of staff scientist Margaret Mitchell.
Of course, that didn’t stop Google (about which more below) or others from applying AI in myriad ways.
Just days before SXSW, a CNN segment showcased how the world of “deepfakes” is expanding from videos that realistically replicate a person’s likeness to audio that can convincingly impersonate someone based on a voice sample. When reporter Donie O’Sullivan’s deepfake voice bot called his parents, they thought their son sounded a little odd, but they didn’t doubt it was him.
It’s both impressive and unsettling, and that’s just one application of the tech. But it’s worth remembering that AI is not a singular tool. It’s a set of systems and complex algorithms fed by data — lots and lots of data, the more the better — and they can apply to virtually anything, or multiple things, from augmented reality to quantum computing, health care to fashion.
Whether that inspires optimism, fear or both can vary by person. The SXSW festival’s agenda broadcasts something for everyone then, across sessions such as “Combatting Climate Change with AI” and “Unlocking the Power of AI: Insights & Innovations” to “Will AI Revolutionize or Wreck Criminal Justice?” and “Can There Be AI Art Without Artists?” Others explore topics like what leadership means in an AI-driven world, how to balance humanity and machine, and more.
It may not be in the title, but AI was a notable part of another SXSW panel, landing on MIT Technology Review’s list of “10 Breakthrough Technologies of 2023.”
Generative AI stems from generative adversarial networks, a deep learning scenario that pits two neural networks against each other, so they can learn from one another. The effort has led to meaningful breakthroughs that do more than merely increase an AI’s knowledge, but can actually generate convincing new works.
“Things have really picked up more broadly in the category of generative AI, and essentially, all of them are really about AI creating something in a certain format — in this case, images,” said Elizabeth Bramson-Boudreau, the MIT publication’s chief executive officer and publisher, referring to Google’s Imagen and OpenAI’s apps.
The latter pair, Dall-E and the upgraded Dall-E 2, allow users to create realistic photos and artworks using natural language — which makes sense, coming from the ChatGPT developer. But consumers may be more familiar with Google’s Imagen because it created corgis out of thin air and placed them in houses made of sushi in photos that went viral last year. It’s hard to look at those adorably happy pups and remember that they aren’t real.
The implications of AI in visually driven sectors like fashion are not only logical, but already being explored in projects. For instance, a new collaboration between backpack brand Sprayground and architectural designer Tim Fu of Zaha Hadid Architects will use AI to design a “hyper-futuristic vision of streetwear,” according to the announcement.
Naturally, generative AI covers more than images. (Hollywood writers are apparently keeping an eagle eye on where bot-driven text goes, as are marketers.)
“There’s tremendous disruption coming for all industries — my industry, I’m sure most of your industries too, and there’s going to be a lot of disruption in image-making for creative industries,” Bramson-Boudreau added. “You [may wonder], do you need to have as many designers if you can use these tools? I know they’re being integrated into Photoshop, and soon they’re going to come to Microsoft Office.”
The rest of the list covered an array of ground-breaking efforts, ranging from somewhat predictable to genuinely alarming. These included gene-editing technology; a new standard-bearing mobile processor likely to reshape how tech giants operate and compete; military drones; the impact of fusing telemedicine and abortion pills; genetic engineering for animal-to-human organ transplants, and more. AI nestled comfortably among that lineup, which, on its face, seems rather telling.
Not that it’s perfect. Even data scientists and engineers admit that artificial intelligence is still young and nowhere near fully baked. But even in its infancy, the tech is already impressive and obviously learning at an accelerating clip. Hopefully, by the time the bots go from tots to grown-ups, society, world governments, commerce and culture will have figured out “how to absorb it,” as OpenAI’s Brockman put it, to get all the positives while mitigating the negatives.
This is the other side of the work, and that has barely begun.