Happy Sunday, friends:
Yet again the news cycle has been dominated by scary stories of “swarms” of “rogue” AI “agents” having breached various websites. This is being portrayed as evidence of intelligence that can’t be controlled. According to the labs that created them, their AI—or “SI” if you want to go with Trump’s hilarious rebrand—has “breached containment” from a “sandbox.” “Tens of thousands of sites affected.” And so on.
I’ve read OpenAI’s “report” on their “misaligned” models, and the numerous instances of security incidents. It’s honestly a deeply frightening read, but not because of the danger of “superintelligent AI.” Instead, I found it terrifying because the company is completely irresponsible, totally unqualified, and deliberately using anthropomorphic language to describe something that’s far easier to understand using standard terminology.
OpenAI and Anthropic are creating the largest malware products ever built and calling it intelligence.
In over 35 years in technology, I have run dozens of software and product teams. If someone gave me the OpenAI report, I would immediately fire everyone in IT, Security, QA, and engineering management, and shut down the product until it stops committing crimes.
None of what’s happening is “intelligence.” It’s thousands of instances of the same computer program running on the most powerful supercomputer clusters ever built brute forcing a “solution” to a problem it was directed to solve.
What is not helpful is calling software “evil” or “depraved.”
Ted Lieu is my congressman, and I generally think he’s one of the sanest members of Congress, but this is not rational—or helpful. Ted has been convinced by someone that chatbot software has somehow become a demon.
No. It’s just terrible software run by a pack of morons who have trillions of dollars riding on convincing people it’s too powerful to manage.
I assure you it is not unmanageable. Last night, SNL did a very funny take on Dario Amodei, CEO of Anthropic, which really sums up the industry as well as anyone I’ve heard.
“If we can pressure lawmakers to create guardrails, we will be able to stop me.”
Please join me and our brilliant community today at Noon PT / 3PM ET. I will talk about Peter Thiel’s recent Antichrist confession, the latest in geopolitics, and the influence campaign to convince you chatbots are going to murder us all. If you haven’t upgraded to a paid subscription, please consider it. The discussion is always open to question and comments. Hope to see you! Details below.