The tech industry may seem like it’s in a lull, plagued by widespread layoffs at major tech companies and a down economy, but that air of doom wasn’t apparent at a gathering of techies and investors in San Francisco on Tuesday.
Instead, there was an overarching feeling of optimism.
They were there to discuss the latest craze capturing the attention of the tech world: generative artificial intelligence. The technology is known to the larger world through ChatGPT, which has captivated imaginations with its ability to generate creative text via written prompts.
Generative AI is a catch-all term describing programs that use artificial intelligence to create new material from complex queries, such as “write a poem about monkeys in the style of Robert Frost” or “make an image of pandas draped over living room furniture.”
While AI more generally refers to software programs that can make themselves better by “learning” from new data, and which have been used behind the scenes in all kinds of software for years, generative AI is a fresh consumer-facing spin on the concept.
About 1,000 people from all over the world, including AI researchers and content marketers, attended Tuesday’s Gen AI Conference, which was organized by startup Jasper. It was a lavish affair, held at Pier 27 on the Embarcadero, overlooking San Francisco Bay.
Attendees noshed on free farm-to-table lunches and bonne bouche delicacies and sipped their coffee from mugs, not the disposable cups typical at most tech events. In the “Art Experience” room, guests could zone out staring at the computer-generated visuals that covered the walls, featuring scenes of multicolored cityscapes and abstract, morphing shapes.
“To me it feels like it’s cracking in a way that Web3 felt like in 2021,” said Ken Walton, vice president of growth for Azra Games, which incorporates blockchain technologies and is backed by Andreessen Horowitz.
“There’s a sense of wide-open possibility,” he told CNBC.
Rising interest rates and the resulting cryptocurrency meltdown of 2022 battered the tech industry, as venture-backed titans such as FTX and BlockFi imploded and many digital coins lost significant value.
The mood in Silicon Valley and the surrounding San Francisco Bay Area was dour.
Then came ChatGPT, from Microsoft-backed startup OpenAI. The underlying AI software powering ChatGPT, a kind of machine-learning technology known as a “large language model,” isn’t new. But the chatting program’s simple-to-use interface meant that the masses could now play with cutting-edge software that was previously limited to AI researchers and techies.
Suddenly the tech sector seemed exciting again. The venture capital community poured $1.4 billion last year into startups specializing in the technology and has amped up the rhetoric.
As Bessemer Venture Partners’ Sameer Dholakia told audience members, generative AI could change “the lives of billions of people.”
Conference organizer Jasper received $125 million in funding in October from investors such as Bessemer, Coatue and IVP. Jasper incorporates technologies from OpenAI and others in its software that generates promotional copy for marketers, among other uses.
But the field of generative AI is so new, startups are still trying to discover appropriate business use cases and figure out how to make money. Because language models like OpenAI’s GPT family of software have gotten much better at producing readable text, investors believe that content marketing represents an easy sale.
Conference attendee Arshavir Blackwell, a machine learning expert and principal at Arvoinen Consulting, told CNBC that he’s interested in using generative AI technologies such as ChatGPT to produce more compelling Facebook ads for clients as part of his consulting business. Blackwell said he believes that the text-producing software has improved so much that it could be possible for advertisers to come up with promotional copy that resonates with users in ways they didn’t imagine.
Blackwell credits OpenAI and ChatGPT with showing people what’s possible with generative AI, shining a spotlight on the industry at large.
“They were not afraid to take risks,” Blackwell said, noting that the AI startup kept releasing new iterations of the software despite its propensity to generate inaccurate information and produce the occasionally offensive comment.
At the same time, advances in computing, particularly the evolution of a kind of computer chip known as a GPU, have made it easier to develop the machine-learning software that lets these programs create more realistic text and images.
“The bottleneck has been the computing,” Blackwell said.
Still, he notes training these massive AI technologies “costs like $5 million.” For now, startups such as OpenAI and Stability AI, which developed a popular open-source image-generating tool, depend on big investors to provide them the money to create their tools.
During one conference session, Dario Amodei, the CEO of AI startup Anthropic, told audience members that firms are getting more comfortable spending a ton of cash on AI because they see the software getting more capable each day.
Until a year ago, Amodei said, “you could only justify a few million dollars” on AI projects. “Folks are spending $100 million now,” he said.
Still, software that incorporates generative AI can prove unwieldy, especially when users make it perform tasks that it wasn’t intended to do. For instance, Microsoft’s recently released experimental chat tool for its Bing search engine has been generating bizarre and sometimes disturbing conversations based on certain prompts.
“The model at times tries to respond or reflect in the tone in which it is being asked to provide responses that can lead to a style we didn’t intend,” Microsoft said in a blog post that addressed the debacle.
It’s part of the mystery of developing and using generative AI.
“I’ve never seen a technology that is produced as such where on one hand where you put all this data in, you’ve got thousands of GPUs, it costs $10 million, maybe $100 million, to train, and once you’ve trained it, you don’t necessarily know everything it can do,” Amodei told attendees.
As Jai Das, a president and partner of Sapphire Ventures, noted on Twitter, “Generative AI is an interesting alternative for content creation, but let’s be honest, the content is very boring and generic.”
“Basically, it’s just a tool,” he added in another tweet.
But for one day in San Francisco, generative AI was more than just a tool. It was the next leap forward in computing.
“We know we’re in one of those moments,” Dholakia told attendees. “This is, as we have seen before in previous generations, a technical leap forward — like cloud computing, like mobile, like the internet — that fundamentally changes the way we operate.”
Meanwhile, as tech enthusiasts and investors debate its importance, we can enjoy the results, such as ChatGPT’s effort to write a poem about monkeys in the style of Robert Frost:
In the woods, where trees stand tall and grand,
A band of monkeys roam the land,
With nimble hands and playful eyes,
They chatter, leap, and swing on vines.
Amidst the green, they find their fun,
In a world where joy is never done,
Where every tree is a playground new,
And every fruit a prize to pursue.
Like mischievous sprites, they dart and dance,
With a joy that seems to take a chance,
And in their playful, wild abandon,
They seem to laugh at life’s demands.
Yet, in their eyes, a wisdom lies,
A knowing glint that underlies,
Their antics, pranks, and playful ways,
A sense of life that goes for days.
So, let us learn from these creatures fair,
And let our hearts take to the air,
With playful spirits, like monkeys wild,
In a world that’s happy, free, and mild.
Watch: Charlie Munger weighs in on ChatGPT3.