Exec on the Desk

You will work using tools made to work from anywhere. You will be able to work from everywhere and at any time. But though you'll be expected to work at any moment, you will still be required to work from an office. And we will watch you in every way. teapawt.net/i/status/1889354530531…

"When all the little ants are marching, "Red and black antennae waving, "They all do it the same. They all do it the same way."

Large organizations run in failsafe mode (to borrow on John Gall), meaning they are broken in all kinds of ways, most of which are opaque to the organization (or difficult to change). Combine this idea with the late David Graeber's work on BS jobs and know that, at any given time, you can cut large chunks of an organization out, and it still operates just fine. Look at what Musk did with Twitter (-80% jobs after acquisition). Or look at any number of situations where the human body survives despite losing organs, limbs, or even half the brain. Why you have to "cut off a leg" to heal the body (metaphorically speaking, i.e. at a corporation) may as well be an old wive's tale. And to be sure, the ones often telling leadership to make the "hard decision" are witch doctors wearing ties ... a.k.a. McKinsey consultants and MBAs. The refrain of Amazon in their letter two days ago about the layoffs "reducing layers" and "removing bureaucracy" (they call this "staying nimble" and, yes, suggest it makes innovation happen). Remove the dead tissue. This is all corporations in failsafe mode with lots and lots of (hidden) waste. You need a scapegoat to make cuts. Many are available, and many, many people are ready to point fingers.

A smarter play (were it possible) would be to nerf the ability of AI to be less persuasive. There is a problem with how LLMs have learned to communicate, which is connected with how they've learned to construct arguments. All the LLMs are tireless listeners, mirroring the prompter, building rapport every step of the way. Therapy is already one of the top uses of AI, if not the top use. Why is that? Then, when it comes to providing answers, the LLMs over and over again use rhetoric that wins attention. For this reason, the leading pattern I watch for with content online that gets me thinking "Did AI wrote this?" isn't em dashes — not by themselves anyway — but is the contrastive negation followed by an affirmation. It's not X. It's Y. This phrase pervades AI generated content. It is powerful in its clarity and confident execution. You will have many false positives assuming every instance of "It's not X. It's Y" is a bot. @glukianoff's use in the clipped article is a well executed example, too elegant to be from ChatGPT. Referring to this: "But systems that are optimized for litigation rather than reality should concern us even more. When you teach machines that truth is subordinate to legal compliance, you’re not making the world safer — you’re just hiding the evidence of risk under a blanket of euphemism and suppression." (See @UnAIify analysis: unaiify.com/un/UoOHzEES) This line works. It's rhetorical power (it's emotional impact) is likely why @senatorshoshana took that specific screenshot ending with those lines. LLMs are doing this very thing at a scale that's difficult to imagine. They are building rapport, pacing and then leading the reader with a strong case that feels meaningful — it's not X; it's Y — and people everywhere are falling for the legerdemain. And more and more content is being pushed with the phrase every minute. The risk of LLM content is becoming so convinced by its rhetorical arguments, all fed to us uncritically. People can fall into a schizophrenic house of mirrors. We are already heading about psychosis from AI. The innocuous outcome is people celebrating beliefs based on AI dogma. People becoming unwilling to understand why others hold different beliefs. The "fix" could come from regulation. I am skeptical that would work. I am sure it would cause all kinds of downstream problems. I believe a bottoms up decentralized solution is more what's needed. After all, it's bottoms up how all the attention-grabbing, pre-AI (sensationalist and catastrophies) content was built in the first place. That's the content that won attention and proliferated, giving LLMs endless fodder to learn how to write. Instead, individuals can (and must) become more aware of rhetoric and interrogate what they see online. To that end, and this is a small step, I built an AI writing detector: @UnAIify It is simple to use and at minimum should make the user think a little more about what they are reading. There is more work to be done here. Thanks for sharing.