Mind and Iron: What happens when we can speak a video into existence?
But also what Sora can't do. And pain relief without opioids?
Hi and welcome back to Mind and Iron, your one-stop-shop for human-centered tech news. I'm Steven Zeitchik, veteran of The Washington Post and Los Angeles Times and head croupier of this blackjack table.
If you’re new here, welcome. Our mission is to chronicle how tech and science are being used to build the future. Isaac Asimov wanted his novel “I, Robot” to be titled “Mind and Iron” — he thought it captured that epic tension of warm humans and cold computing — and that’s exactly the quest we’re on too. Please consider donating to our cause here. Every dollar truly helps.
The pain of our not meeting last week, it was deep. But it was also for good reason: attendance at the MaRS Impact AI conference, where some of the leading AI thinkers gathered in Toronto to talk about the future. Thanks to everyone who came out to my panel about the ripple effects of AI and the many of you who said hi afterward. Some highlights from the conference below. (E.g., one of the cooler ways to save our planet/preserve our oceans.)
Of course no sooner did we decide to take the week than Sam Altman and OpenAI made major news: they announced Sora, the text-to-video system that turns everyone into James Cameron overnight (allegedly). You by now may have read about Sora — transformational! overhyped! disruptive! destructive! — so we won't belabor the point. But we'll give you a couple nuggets to chew on as this age of AI video dawns.
Also, a little audio bonus — yesterday I appeared on Variety's venerated Strictly Business podcast to talk about Sora. One important question that host Andy Wallenstein posed is whether Sora fears in the creative community are hysterical or prescient. If you want to hear some answers — or if you’re just over this whole reading thing — you can give a listen here.
And finally, on non-AI fronts, can we construct a world where pain isn’t a fact of life but opioids aren't either? A look at a new biotech groundswell.
First, the future-world quote of the week.
“I just hope that as people are embracing this technology, and as companies are moving to reduce costs, there’ll be some sort of thought and some sort of compassion for humanity and the people that have worked in this industry and built careers and lives.”
—Tyler Perry on AI video, squinting to see the light through his Eeyore glasses
Let's get to the messy business of building the future.
IronSupplement
Everything you do — and don’t — need to know in future-world this week
Sora, superhero or super-monster; listening for whales; a future without opioids?
1. THE PRESENTATION WAS DAZZLING. NOW WHAT?
OpenAI recently released demos for Sora, its so-called “text-to-video” program.
If you missed them or want to revel anew in their otherworldly beauty, here they are. All of these were created, the company says, with nothing more than someone typing in the accompanying prompt. No animators, coders, directors or humans in sight.
So, does this instantly turn everyone into a high-level creative, and what does creativity even mean if it’s so easily achieved? Does it transmute social media from the homespun to the polished — will TikTok videos soon be rife with people moving in this gauzy, semi-real fashion? Is this stylization to the next five years of social media what doing the Renegade and lip-synching to Drake have been to the last five?
Will Sora enable new art forms? Does Sora transform some Hollywood jobs and wash away the rest? (Tyler Perry certainly thinks so; he just hit pause on his studio expansion because of it.) Will Sora make disinformation even more pervasive and convincing? Does Sora conjure an epic battle between people who believe in handmade creative work and people who believe machines do it better?
The answer to many of these is — amazingly —a host of qualified yesses. We are in the midst of turning one of humanity's most impressive endeavors — capturing or fabricating reality — an endeavor that for most of existence humans couldn’t even do, into something a machine can now do for us. We’ll be dealing with the legal, financial and moral consequences for years.
More to come on all this in later issues, especially after Sora is released commercially. (Right now it’s not even in the hands of developers.)
For now, three takeaways.
This is a next step in our modern media evolution.
Here’s the 30,000-foot view of what’s happening. For basically a century visual images could only be distributed in the most painstaking, centralized way possible, first with giant projectors in movie theaters and then with antennae and cable signals on clunky television sets. About 15-20 years ago Web video (and smartphones) came along and changed the distribution game. Suddenly a teenager could put her skateboard tricks in the public square and the world could see them. Suddenly a studio could put a new drama online and anyone with a phone could watch it. The democratization of video distribution was transformational in, well, all the ways we know today's streaming and mobile world has been transformed.
Now OpenAI and its ilk (Google won’t be far behind) are coming along to change the production landscape. Suddenly a teenager (or grandfather) with zero technical skill could type or speak and high-quality video appears. Suddenly an entertainment company doesn't need very many employees to create video. This democratization will be even bigger than the last. (The savvy Substacker The Mediator has a good deep dive on this.)
Netflix, YouTube and TikTok changed how we consumed moving images by removing many of the laborious efforts required to distribute those videos. Now imagine what happens when computer programs remove the laborious human efforts required to create that video in the first place.
But here’s what Sora isn't.
Sora isn’t storytelling. Because video is great but storytelling is something else — it’s cause and effect, exposition and revelation, tension and release.
And nearly everything we’ve seen so far from AI video has been non-narrative. Much of it, in fact, is glorified screensavers. And while there’s reason to be impressed by how damn good that screensaver looks, there’s no reason yet to believe it transcends the form— no reason to believe it can do anything more than merely supplement a good story. (And sorry, ChatGPT ain’t writing that story.) You can make animals pedal bikes on water or a lonely man move through the Tokyo snow. Great. Now give me a reason to watch it.
The French pioneer Georges Melies made “A Trip to the Moon,” one of the first narrative films, 120 years ago. He’d blush at the technological tools we now own. And yet a machine is still decades behind him on the storytelling front.
So highly original storytellers are safe. In fact, they’ll become even more central. Anyone can create video. Find me the person who can tell me a compelling story. (Tyler Perry has no reason to personally worry.)
Sora also isn’t long. A minute is a ton of time in moving images — hundreds of frames stacked one after the other. That means massive computing power. But it also means that anything resembling more than the shortest short won’t be here for a while; you can forget about a feature film or sitcom episode.
Honestly the biggest target for AI video right now is short content from social creators — exactly the opposite of the longer stuff Hollywood pays a lot to produce and we pay a lot to watch.
Oh, and Sora doesn’t have sound — that’s a whole other kettle of fish (bicycling on an ocean).
Sora also isn't perfect. In fact it's stubbornly imperfect. Five-pawed cats, stairs to nowhere, cars that disappear, waves that move backward. You might think these are easy to fix. They're not. The model is autonomous, and the faux pas are a function of the fact that computers haven’t experienced physics; they’re limited by Steven Soderbergh’s famous principle that an AI never felt insecure or made a meal for someone it loved. Maybe some new programs will be tailored to look for this stuff and correct it (though they might run into the same problems).
Some creators will also no doubt incorporate these mistakes into their storytelling — sci-fi and horror just got a boost. But the imperfections put a limit on Sora. For the foreseeable future, to make something look undetectably real you’ll still need to physically stage it. (This presents, at the very least, a small bit of reassurance on the disinformation front.)
Sora makes everyone a creator. And I mean everyone.
Remember when MySpace came along and the hype was that everyone was now a filmmaker? (Sadly, I am old enough to.) Anyone who’s watched the Web video we’ve had in the two decades since would laugh at that comment. No more laughing. Now anyone really can make a video that looks that good.
What will this do to the technical end of creativity is a wide-open question. Imagine a world in which anyone can make something look as good as Pixar; how does Pixar define itself then?
And what this will do to these social platforms — who might excel when the video language has been so transformed and the technical skill no longer required? The creator economy just got smashed to pieces. There are going to be new creators who master this form. And there are going to be popular creators who find themselves left behind.
But I’m more interested in what this does to us as a society. The aesthete in me loves that polished beauty will be everywhere; that badly lit homemade shaky-handed film is a relic of the past (unless you order Sora to make something look like a badly lit homemade shaky-handed relic of the past). But a lot is lost with that, a lot of texture and authenticity. In an age when phones let us airbrush out any photo expression we don’t like and social-media performativeness means no one ever had a bad day, our whitewashing just got taken to a new level. In the Sora age, no one ever has to produce a video that doesn’t look good. Which is kind of bad.
AI arrives with the promise of creating a world of beauty. In the process, though, it will paint over a lot of reality.
[The Hollywood Reporter and The Mediator]
2. COULD WE EASE PAIN WITHOUT RELYING ON OPIOIDS?
Even with the 16 percent decline in opioid prescriptions over the past few years, Americans are still taking the stuff at alarming rates — some 40 prescriptions per every 100 adults. And opioids, as we tragically know too well, are a factor in far, far too many deaths — some 70,000 every year.
But opioid prescriptions don’t happen randomly — pain from surgery and various conditions can be intense. More than 130 million people were prescribed opioids in 2022 for just this reason.
So biotech companies have been scrambling to find a better alternative. That race is now heating up in a major way.
Last month, Boston biotech firm Vertex found success with a late-stage clinical trial of a drug named VX-548, which relieved pain (in groups of patients who had minor procedures) about as well as Vicodin did. You may recall hearing about Vertex in this space a few months ago, when the firm discovered an effective CRISPR treatment for sickle cell disease. This continues the trend — a cutting-edge approach to an age-old medical challenge.
The company’s chief science officer says they’re going to apply for FDA approval of VX-548 as soon as this summer, which puts the drug in the realm of the plausible future for the 15 million Americans who have surgery every year.
Vertex isn’t the only one closing in on the goal of replacing opioids with a potentially safer alternative. (Or replacing NSAIDs, less effective/tolerated.) A Southern California company named Latigo Biotherapeutics said this month that it has secured $135 million in fresh funding. That money will allow it to move forward with research and testing on a similar drug to Vertex’s. Theirs is called LTG-001; it’s in the first phase of a clinical trial.
Both drugs work by targeting the peripheral nervous system instead of the brain, as opioids do. Specifically, they go after something called sodium channels, which serve as a kind of backroad system within cells. Sodium channels are genetically associated with levels of pain sensitivity, so these companies hypothesized that by targeting them they can prevent pain signals from being transmitted — far preferable to targeting the brain.
Plenty of concerns abide. Sodium channels have long been thought promising, but false starts have been common. A study five years ago found that pain drugs targeting the channels didn’t show themselves to be effective. And side effects are a long way from being documented. Just because they're not addictive doesn't mean they can't do harm.
But there’s no denying how promising all of this scientific and financial energy is. As Americans live longer and undergo more procedures, pain will be an increasingly relevant fact of life. These companies are trying to construct a future in which there’s less of it — and in which hundreds of people don’t have to die every day to make that happen.
3. ONE OF THE NICE THINGS ABOUT TRAVELING TO TECH CONFERENCES IS HOW MUCH OF A SENSE you get of the prevailing winds. What are future-minded investors and entrepreneurs focusing on? What should we be excited for? What should we be worried about?
Here are a few comments I heard throughout the MaRS Impact AI event last week.
-Kris Bennatti, an AI thought leader at a company called Hudson Labs, talked about getting a call from a financial-services firm that wanted to replace several dozen equity analysts with AI; she described it as “an extreme example but not uncommon.” (We’ll leave the feasibility — and desirability — of all these for another time.)
-A vp at Unity Health Toronto, Muhammad Mamdani, talked about using AI to glean information about patients instantaneously without so much as a lab or a test. The ancients tested for diabetes by tasting urine, he said; an AI could literally sniff it out.
-One use case that stood out to me was Whale Seeker, an AI company that can, well, detect whales.
Right now sensing what’s in the depths of the ocean takes expensive cameras, sensitive instruments — and a lot is still missed. But by using modeling and other AI techniques, the startup says it can pick up the presence of whales and other marine mammals in a given area and then send that information to ships, shipping companies and others to aid with their decision-making — literally preserving species through data. (And serving a far more constructive purpose for planet-heating energy-guzzling AI.) More than 100,000 whales have been detected so far.
“Our goal,” co-founder Emily Charry Tissier told me, “is to be the AI company that protects marine life.” Sometimes all this tech will disrupt the planet. And sometimes it will give it a boost.
The Mind and Iron Totally Scientific Apocalypse Score
Every week we bring you the TSAS — the TOTALLY SCIENTIFIC APOCALYPSE SCORE (tm). It’s a barometer of the biggest future-world news of the week, from a sink-to-our-doom -5 or -6 to a life-is-great +5 or +6 the other way. Last year ended with a score of -21.5 — gulp. But it’s a new year, so we’re starting fresh — a big, welcoming zero to kick off 2024. Let’s hope it gets into (and stays) in plus territory for a long while to come.
SORA WILL MAKE ANY OF US INSTANT VIDEO CREATORS: It’s hard to get one’s head around the exponential number of consequences, but for now mild skepticism feels right. -2
PROGRESS FOR NON-OPIOID PAIN-RELIEVERS? Tentatively hopeful. +2.5
AI CAN ALSO HELP SAVE THE PLANET EVEN AS IT HEATS IT UP: +1.5