John Mulaney sure knows how to make tech workers laugh.
At Salesforce’s annual conference, Dreamforce, which wrapped up in San Francisco a few weeks ago, the comedian expressed concern over where things might be heading with AI.
After his warm-up act addressed the crowd as Skynet, Mulaney took to the stage to say, “I really wish what you’re doing won’t work, but I’m sure it will.”
“Backstage I heard someone say we need to skill up the human—felt like I was in a movie with some terrifying sh*t about to happen.”
Of course, Mulaney aside, the show is all about accelerating.
Nvidia CEO Jensen Huang explained how the challenges facing humanity are “many and worthy” to be tackled right away. He presented a vision of billions of agents helping humans do all sorts of things, some specialized, some generalized, as AI advances in reasoning, understanding greater nuance and collaborating with others.
“Work is going to be done before we even think of it,” he laughed. “We’re going to come to work and a whole bunch of things we didn’t realize needed to be done will be done.”
“Nobody should miss the next decade,” he said.
In fact, a trillion dollars is expected to be spent on AI by the tech giants over the next five years, according to Goldman Sachs.
And the money keeps pouring in. OpenAI just landed the biggest venture capital round of all time, $6.6 billion at a $157 billion valuation, making the ChatGPT-maker one of the largest startups in the world next to TikTok’s ByteDance and Elon Musk’s SpaceX. Because operating costs are so high, for extra liquidity, the company was also able to secure a $4 billion line of credit from JPMorgan Chase, Citi, Wells Fargo and Goldman Sachs, among others.
Thrive Capital, which led the round, even pledged an additional billion dollars for 2025, reported Reuters. Other investors include Nvidia, Microsoft, SoftBank, Tiger Global, Khosla Ventures, Fidelity, ARK Venture Fund, Altimeter Capital Management and MGX backed by UAE sovereign wealth fund Mubadala, sans Apple which dropped out, according to the Wall Street Journal.
Nearly two years since its viral launch, ChatGPT is now at 250 million users with 11 million ChatGPT+ subscribers and one million ChatGPT Enterprise subscribers. The company has yet to turn profitable and is expecting a $5 billion loss on $3.7 billion revenue for 2024, with $11.6 billion projected revenue for 2025, per CNBC.
Agents rolling out
Meanwhile Disney, Wyndham, OpenTable, Wiley and Saks have been piloting Salesforce’s advanced reasoning engine, Atlas, to create agents that optimize for the consumer experience—whether its helping Disneyland tour guides get real-time ride availability to reduce visitor wait times or the Saks’ AI Shopping Concierge transacting to buy and ship outfits based on customer text and image prompts.
Music producer and Black Eyed Peas frontman Will.i.am demoed for me how he’s been using Salesforce’s low-code Agentforce platform, to enable listeners of his RAiDiO.FYI app to order merchandise from ads, simply by asking for them.
Priceline has been using OpenAI’s Realtime API for Advanced Voice to create an AI agent named Penny Voice that can provide hands-free search and bookings for hotels as well as recommendations for restaurants and activities. CEO Brett Keller told me the goal for Penny is to become a full-service concierge.
“We are working on exciting enhancements to Penny, to give more insight into the on-the-ground neighborhood feel,” he said. “That’s the softer stuff travel websites haven’t been as good at helping customers understand and that’s place we’re leaning into with Penny so people can feel confident in what they’re looking to book.”
United Airlines CIO Jason Birnbaum told me the company is experimenting with agents, but found one of the most compelling use cases for generative AI in migrating legacy mainframe systems written in COBOL, a 60 year old programming language that few humans know, but LLMs can be train easily on.
Worth it?
As businesses figure out their best moves with the technology, Gartner is predicting that 30% of AI projects will be abandoned by the end of 2025.
The biggest challenge is not so much the implementation of the model, but getting the data ready, Deloitte CTO Bill Briggs told me. “Our data modernization business is red hot,” he said. The data needs to be clean and accessible, but no CEO wants to take on a nine-figure project to replace the plumbing. You’re more likely to leapfrog the competition, if you’ve been making the investment all along.
