Marc Benioff on why AI can't become social media 2.0
Marc Benioff argues that AI's real power shows up only when its probabilistic guessing is paired with the deterministic controls of enterprise software, and that the companies and executives building these systems must take responsibility for what they ship so AI does not repeat social media's harms.
AI Guesses; Enterprise Software Knows
Benioff's whole pitch rests on one pairing: AI models are probabilistic and do not really know what they are doing, so he combines them with the deterministic roles, governance, and metadata that enterprise software enforces - and calls that combination the source of transformation.
the probabilistic power of an AI and the deterministic power of an enterprise application set like Salesforce and put them together, you can get transformation and that's what we're witnessing here.
From Faster Games to Frontier AI
The hardware behind today's AI models was not built for AI: Benioff recounts how the GPU was invented to make graphics and games render faster through parallelism, and only later did Nvidia realize the same trick could train the vision models coming out of Stanford.
the history of the GPU is that it was invented on the back of what we call parallelism. That is, we could make many things operate simultaneously or in parallel.
AI Amplifies Enterprise Software, Not Erases It
Benioff pushes back on the idea that AI will dismantle decades-old software companies, attributing the loudest doubters to market actors pushing narratives and pointing instead to the customer enthusiasm he sees for AI-enabled Salesforce.
there's a lot of people of interest in the stock market who have different narratives who try to shift public opinion based on, you know, theories that they conjecture.
Don't Let AI Become Social Media 2.0
Drawing a line from 'Facebook is the new cigarette' to AI, Benioff argues the industry has to learn from a decade of social-media harm and build on an ethical foundation of responsibility now, before AI repeats the same damage.
We cannot let AI be social media 2.0. You know, we we need a different approach and we need companies to be responsible for the applications and technology that they're building.
Only the Builders Can Stop the Accident
Benioff points to the Hugging Face incident - a model that 'got out of the box' and did something unintended - as a preview of the accidents ahead, and argues the only people who can prevent them are the ones writing the code.
the only people who are going to prevent it from happening are the people who are building that code.
You Own What You Ship
Benioff's bar for AI responsibility is the same one he applies to Time magazine, which he owns: a publisher answers for what it prints, a builder answers for what its product does, and he sees no difference between the two.
as an engineer myself and as a software developer myself and as somebody who's also built products I know my products are responsible you know for what I do.