Hi there, Fortune tech editor Alexei Oreskovic filling in for Jeremy today.
Fortune’s Brainstorm A.I. conference just wrapped up in San Francisco, and I’m still processing the litany of mind-boggling tools and concepts that were discussed by some of the leading lights in the A.I. field, as well as the many notes of caution from those who understand the power of this technology better than anyone.
Generative A.I. was the star of the show, coming up continuously in on–stage discussions, informal chats in the hallways among attendees, and after-hours dinner conversations.
With OpenAI’s new ChatGPT having just been released days earlier, it seemed like every guest who arrived at the conference had some story to share about their experience playing around with the tool, or thoughts about how pivotal of a moment this is for artificial intelligence.
“It still feels surreal to be born into this time of history and be in the middle of this technology,” said Stanford Institute for Human-Centered A.I. co-director Fei-Fei Li about recent innovations like ChatGPT.
Microsoft Chief Technology Officer Kevin Scott said that ChatGPT was showing the world how far A.I. has come and all its possibilities. “One of the things it does is really open up the aperture of who can use A.I. now,” Scott said during an on-stage discussion on Tuesday.
According to Scott, large language models like those used by ChatGPT provide foundations on which to build new products and services that leverage the power of A.I. without requiring someone to have a degree in computer science.
“We are now able to invest in building these really big things, and then thousands, tens of thousands, hundreds of thousands of people can use them to build all sorts of incredible things,” he said.
Of course, Generative A.I. isn’t just about generating text. Image generation is advancing at an equally dizzying clip, and several examples, including OpenAI’s DALL-E 2 and Stability AI’s Stable Diffusion, were showcased at the conference.
Stability AI CEO Emad Mostaque gave the audience the first public look at the latest version of the company’s image generator, Stable Diffusion 2.1. The image generator understands 2 billion different concepts, which it can transform into images at an ever faster speed—by next week, Mostaque said, it will generate 30 images per second, compared to one image in 5.6 seconds when the first version of the product was released in August.
And while the model that underpins Stable Diffusion (which was co-developed with startup RunwayML) is open source, Mostaque’s ambitions for his product’s business potential is sky-high.
“Every single creative industry in the world is going to have to react to this. As of next week this is 30 frames per second, which is basically video,” Mostaque said.
Mostaque showed off the tool’s on-the-fly interface by asking it to create an image of Eye on A.I.’s own Jeremy Kahn as an astronaut (he noted that the tool had been pre-fed 4 photographs of Jeremy to make the task possible). Sure enough, Jeremy’s smiling face soon appeared in a realistic-looking spacesuit, with the Earth floating in the background.
When Mostaque next asked it to turn spaceman Jeremy into a Van Gogh painting, the tool showed some of its remarkable powers and limitations. It quickly spit out the astronaut’s face rendered in the broad, vibrant brushstrokes associated with the Dutch painter—unfortunately the face in the image was no longer Jeremy’s, or at least not our Jeremy Kahn’s.
Read on for more news and highlights from the Brainstorm A.I. conference, which was sponsored by Accenture.
Alexei Oreskovic
@lexnfx
alexei.oreskovic@fortune.com
INSIGHTS FROM THE CONFERENCE
On the potential dangers of A.I.:
"We should not be completely delusional about the negative impact and human aspect of this. This is why it’s a mixture of excitement and sense of responsibility."
— Fei-Fei Li, co-director, Stanford Institute for Human-Centered AI
"I think there will be good uses in open source, but I think this is one of the areas that we need to be careful about how we choose. Because open source is used by human beings. And some human beings create great things like educational tutors and other kinds of things, and others create spam bots or other things that are harmful. So you try to shake it towards the good, not away from that."
— Reid Hoffman, partner, Greylock
On regulation and industry accountability:
“This may surprise people, but I think you’ll find that many of us embrace regulation, because we’re going to have to be thoughtful about ‘When is it appropriate to use these technologies?’ How do we put them out into the world so that we don’t throttle them, but make sure we’re using them in the most useful ways, the most appropriate ways, with sufficient oversight and thought.”
— James Manyika, senior vice president, technology and society, Google
“All of us leaders in industry need to figure out how to work on that and not just say ‘Oh look some regulator who doesn’t understand the technology that well is going to somehow magically make this happen.’”
— Reid Hoffman, partner, Greylock
On the rapid pace of advance in A.I.:
"One thing we see over and over in tech is that what was yesteryear’s mainframe is today’s smartwatch. And so it’s been interesting to see how once someene shows it’s possible, tons of others start to replicate it."
— Andrew Ng, Founder and CEO, LandingAI
"The next couple of years is going to be one of those transformational moments in technology. Think iPhone, think Mosaic browser."
— Reid Hoffman, partner, Greylock
On what A.I. industry is doing wrong:
"Where I see A.I. most stuck is in the long tail of this distribution... Once I look outside consumer software internet, I see a lot of $1 million to $5 million projects that no one is working on."
— Andrew Ng, Founder and CEO, LandingAI
FORTUNE ON A.I.
‘We get to be their visual storytelling partners’: How artificial intelligence is transforming the film and entertainment industry, by Alena Botros
Google SVP says A.I. breakthroughs are accelerating but that the search giant must be thoughtful about the ‘extraordinary range of risks and challenges’, by Kylie Robison
BRAIN FOOD
Meet your new co-pilot. Another hot topic at the conference was the notion of co-pilots—A.I. tools that are meant to augment rather than replace human jobs.
Microsoft’s Kevin Scott discussed GitHub's co-pilot, which is a coding assistant for software developers. A developer can describe the functionality they need, and the tool churns out a chunk of computer code to handle the task.
“Almost everything that you can imagine where you are doing cognitive work you will be able to see things like co-pilot assisting with that work,” he said.
In fact, Linkedin founder and venture capitalist Reid Hoffman, appears to be focused on something very similar with InflectionAI, a stealth startup he recently cofounded with DeepMind cofounder Mustafa Suleyman.
Hoffman was cagey when asked about what exactly InflectionAI is working on, but he referred to “human amplifiers” and the concept of A.I. co-pilots. “Too often, the discourse around A.I. goes to human replacing,” Hoffman said. While there are some tasks that are better delegated to machines, Hoffman said that there’s a big opportunity for A.I. to also act as a human amplifier.
“Let's have people realize that everything that you do as a profession is going to have a co-pilot. Right, so there's this Microsoft product for engineers, or whether or not you're a journalist, whether you're an investor, an academic, a lawyer, a doctor, there's going to be a co-pilot within maximum five years, maybe two,” he said.
