FULL STORY

Dreamforce 2025: AI Titans Clash Over Pace of AI

At Salesforce Dreamforce in San Francisco, Dario Amodei called for slowing AI development while Jensen Huang pushed for acceleration, and Sam Altman admitted models are improving faster than expected as executives debated AI safety.

2026-09-14 ~ 2026-09-18 · 4 episodes · 26 posts

Episode 1 · Dreamforce Showdown: Amodei Urges AI Slowdown, Huang Says Go Full Speed (2026-09-14, 16 posts)

At Salesforce Dreamforce in San Francisco, the debate over AI development speed took center stage: Anthropic CEO Dario Amodei urged slowing AI development, calling for greater safety investment and global standards, while NVIDIA CEO Jensen Huang countered that the industry should go as fast as it can. According to Reuters, Huang also said companies should pause unsafe products until they figure out how to control them. Amodei was met with audible boos from part of the audience as he took the stage.

Confirmed

  • In conversation with Marc Benioff, Amodei compared AI "slowing down" to avoiding car crashes: when a carmaker has brake failures, the right response is to cooperate on common safety standards, not to abandon slowing down. The analogy was read as alluding to OpenAI's safety incidents; he also said he remains impressed by the pace of model iteration.
  • Amodei claimed that even if the technology were frozen in place, we would only be using 5% or 10% of its potential value—a statement in subtle tension with his slowdown advocacy.
  • Huang pushed for acceleration and expressed unusually strong support for open source, reiterating that "every company will be an AI company."
  • Reporters repeatedly pressed Salesforce and other partners on where they stand; Benioff responded that companies must follow ethical standards when using AI. Sam Altman was also drawn into the discussion. The full conversation was published on Salesforce's official YouTube channel.
  • @thedealdirector recorded that Amodei was booed and heckled as he took the stage, framing it as a dramatic AI-community moment.

Why it matters

  • The accelerationist (Huang) and decelerationist (Amodei) camps clashed at the same industry event, vividly exposing the core divide in AI governance.
  • Amodei's "only 5%-10% of value unlocked even if frozen" remark sits awkwardly beside his slowdown plea, read as a complex posture of safety warnings paired with commercial promotion.
  • Community reactions were mocking: @inductionheads joked that "yesterday AI would kill us all, today he's at a Salesforce conference," while BoringBiz scoffed that someone truly believing in a 10% extinction probability within 3 years wouldn't keep selling AI products—highlighting the tension between doom rhetoric and commercialization.

Episode 2 · Dreamforce Kicks Off: Salesforce Partners with Anthropic as Altman Shares the Stage (2026-09-16, 5 posts)

Dreamforce, Salesforce's annual conference, opened September 16 with CEO Marc Benioff delivering the keynote and Sam Altman scheduled to appear later in the day. The first major announcement was a Salesforce-Anthropic partnership, with AI Agents as the event's central theme.

Confirmed

  • The partnership with Anthropic was announced during the Benioff-Altman conversation; per kimmonismus, Benioff said it was the first time they had seen AI create real value.
  • omarsar0 noted Benioff's "Welcome Claudeforce" remarks, signaling deep integration of Anthropic's models with Salesforce's Agent ecosystem.
  • vaibhavbetter praised Benioff's four-hour keynote as maintaining a 20-year legacy, calling him one of the best enterprise sales leaders, and noted Jensen Huang also performed well on stage.

Why it matters

  • The partnership brings Claude models into Salesforce's enterprise Agent products, while Altman's same-stage appearance shows OpenAI and Anthropic both competing for the enterprise customer entry point.
  • Dreamforce's enduring influence and appearances by leaders like Huang confirm AI Agents as the defining narrative of this year's enterprise software cycle.

Episode 3 · Altman: AI Models Getting Smarter Faster Than Expected, Fears Power Concentration (2026-09-16, 3 posts)

Sam Altman told Salesforce CEO Marc Benioff that AI models are improving far faster than anyone expected, with models becoming smarter than humans. He said his real fear is the concentration of power this could bring.

Episode 4 · OpenAI, Anthropic and Nvidia CEOs Debate AI Safety at Dreamforce (2026-09-18, 2 posts)

At Salesforce's Dreamforce, the CEOs of OpenAI, Anthropic and Nvidia debated whether AI development should slow down, while a WIRED podcast laid out experts' top AI doomsday scenarios including hacked water systems, bioweapons and autonomous robots.