Reality sets in for generative AI

Welcome to Runtime! Today: the wildest dreams for the future of LLMs appear to be falling apart, Microsoft kicks partners from certain countries out of its early-notification program for vulnerabilities, and the latest enterprise moves.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

Tokens of depreciation

Each technology generation is defined by a hype cycle, a pattern so repeatable that Gartner charges obscene amounts of money to draw the same chart over and over again. But the cycle that followed the rise of generative AI stood out compared to earlier eras thanks to its messianic marketing, and this week doubts that large-language models will ever live up to those promises started to break through.

Investors in AI-related tech stocks decided this week was a good time to take some of their winnings off the table, with Nvidia, Amazon, Microsoft, and Google all feeling the effect. And after OpenAI CEO Sam Altman — perhaps the biggest cheerleader for the sector — acknowledged that generative AI technology isn't evolving as fast as once thought, who can blame them.

“Are we in a phase where investors as a whole are overexcited about AI? My opinion is yes," Altman told a group of reporters in San Francisco last week, according to The Verge.
At the same dinner, he also admitted that "the models have already saturated the chat use case. They’re not going to get much better. ... And maybe they’re going to get worse," according to CNBC.
Altman's newfound skepticism was echoed by former Google CEO Eric Schmidt and Selina Xu, who have pushed the U.S. government to invest heavily in AI in order to beat the Chinese government to the development of artificial general intelligence, whatever the hell that is.
"We worry that Silicon Valley has grown so enamored with accomplishing this goal [of AGI] that it’s alienating the general public and, worse, bypassing crucial opportunities to use the technology that already exists," they wrote Tuesday, and if you've never watched this classic I Think You Should Leave sketch, open it up in a new tab for your enjoyment after you've finished this newsletter.

The real problem for generative AI backers, however, remains the same: most enterprises are finding it too hard to deploy the technology at scale. That was the conclusion of a study released Monday by MIT that made a lot of headlines and likely led to some of this week's sell-off.

The stat that got everyone's attention was that only 5% of generative AI projects "achieve rapid revenue acceleration," according to Fortune.
However, most enterprises see generative AI as a tool for reducing expenses, rather than generating revenue.
But even in that context, something clearly isn't working.
"Generic tools like ChatGPT excel for individuals because of their flexibility, but they stall in enterprise use since they don’t learn from or adapt to workflows," Fortune reported.

Over the last three years, generative AI technology has been a classic example of how everything starts to look like a nail when all you have is a hammer. The good news for enterprise tech vendors and buyers is that MIT's study pointed to some promising trends for generative AI that could help chart a path forward.

"More than half of generative AI budgets are devoted to sales and marketing tools, yet MIT found the biggest ROI in back-office automation — eliminating business process outsourcing, cutting external agency costs, and streamlining operations," Fortune reported.
And according to the report, of the companies that did succeed with generative AI projects, two-thirds of them purchased software or services from a vendor; the third that tried to build it all themselves came away with little to show for it.

But when AI acolytes like Altman and Schmidt start to manage expectations for the future of the technology, it's clear a day of reckoning is coming for generative AI.

LLMs are still a promising tool in the hands of trained professionals, but they are expensive, difficult to manage, and hard to trust.
And last year's promise that autonomous AI agents would fix these issues is nowhere even close to coming true.

After hackers said to be working with the Chinese government got a suspiciously early jump on the disclosure of a Microsoft SharePoint vulnerability last month, security researchers wondered if it was time for Microsoft to rethink its system for sharing vulnerability information with partners. According to Bloomberg, that's exactly what happened.

Microsoft will no longer share specific details about software vulnerabilities with "countries where they’re required to report vulnerabilities to their governments," a company representative told Bloomberg. Instead of sharing "proof of concept" information that shows exactly how an exploit for the vulnerability works, it will now only share "a more general written description" of those vulnerabilities along with the patches, according to the report.

Microsoft needs security partners around the world to help it distribute patches when a vulnerability in a key piece of software is discovered, but after the SharePoint flaws were exploited so quickly it appears to have decided on a more balanced approach. This week Trend Micro released new details about the Warlock ransomware gang, which is targeting unpatched SharePoint servers managed by "organizations across North America, Europe, Asia, and Africa, spanning industries from technology to critical infrastructure."

Enterprise moves

Lee Klarich is the new chief technology officer at Palo Alto Networks, replacing founder and CTO Nir Zuk, who decided to retire after more than 20 years at the company.

Jim Joudrey is the new CTO at UKG, joining the HR software company after technology leadership roles at Amazon and Salesforce.

Eric Freeman is the new chief information security officer at Writer, a newly created role for the former head of security at Patreon.

Sarah Strobhar is the new chief revenue officer at Upbound, joining the infrastructure management company after sales leadership roles at AWS and Applied AI.

Holly Grey is the new chief financial officer at Horizon3.ai, joining the security company after serving as CFO for Exabeam.

The Runtime roundup

Workday reported earnings results that beat Wall Street's expectations, but signaled that funding cuts to local and state governments could impact its results later this year.

Zoom beat expectations for revenue and profit and raised its guidance for the full year, sending its stock up 5% in after-hours trading.

Google's Gemini apps use 0.24 watt-hours of energy and 0.26 milliliters of water to process the median prompt, a disclosure that MIT Tech Review called "the most transparent estimate yet from a Big Tech company with a popular AI product."

Thanks for reading — see you Tuesday!