This is the most California thing I've ever read. Unconstitutional, unenforceable, and a massive increase in costs and bureaucracy; they hit the trifecta! I wonder if printer manufacturers that bake their own bread will be exempt once their checks to the governor's presidential campaign clear.

Incidentally, this is the kind of stupid shit that helps Trump and people like him get elected over and over.

There's a very clear pattern when we look at who benefits in any given gold rush and while there's a few big winners that fuel the mania, the vast, vast majority are losers.

And then there's Nvidia, happily selling shovels all day.

No, of course not, because he's stuck in his dogmatic viewpoint. He doesn't actually know much about LLMs, but he's got a ton of beliefs about them. And you have a hard time changing peoples' beliefs.

That was definitely a problem two years ago. I did just check in ChatGPT, Claude, and Gemini and all reported 3 correctly. The problem with people throwing out these sorts of criticisms isn't that they're all wrong; it's that they're ignorant of the leaps in progress being made. These models are rapidly improving and it's getting harder to find serious gotchas with them. They're still weak in some areas (e.g., spatial reasoning), but for serious power users who know how to prompt them well? They've become insanely powerful tools.

Not gods; tools. But really, really strong tools for huge variety of tasks.

I've used ChatGPT to write code and Gemini to debug it. If you pass the feedback back and forth, it takes a couple iterations but they'll eventually agree that it's all good and I find that's about 90-95% of the way to where I need it to be. Earlier today I took a 6kb script that had been used as something fast and dirty for years - written by someone long gone from the company - and completely revamped it into something much more powerful, robust, and polished in both its code and its output. Script grew to about 20kb, but it's 10x better and I only had to make minor tweaks. Between the two, they found all sorts of hidden bugs and problems with it.

Self-hosted Atlassian products seem to be just fine.

As for what "they said", they also said this cloud shit would be cheaper. It isn't.

Wikipedia is an interesting concept and it works decently well as a place to go read a bunch of general information and find decent sources. But LLMs are feeding that information to people in a customized, granular format that meets their exact individual needs and desires. So yeah, probably not as interested in reading your giant wall of text when they want 6 specific lines out of it.

Remember when Encyclopædia Britannica was crying about you stealing their customers, Wikipedia? Yeah, this is what they experienced.

If you spend time with the higher-tier (paid) reasoning models, you’ll see they already operate in ways that are effectively deductive (i.e., behaviorally indistinguishable) within the bounds of where they operate well. So not novel theorem proving. But give them scheduling constraints, warranty/return policies, travel planning, or system troubleshooting, and they’ll parse the conditions, decompose the problem, and run through intermediate steps until they land on the right conclusion. That’s not "just chained prediction". It’s structured reasoning that, in practice, outperforms what a lot of humans can do effectively.

When the domain is checkable (e.g., dates, constraints, algebraic rewrites, SAT-style logic), the outputs are effectively indistinguishable from human deduction. Outside those domains, yes it drifts into probabilistic inference or “reading between the lines.” But to dismiss it all as “not deduction at all” ignores how far beyond surface-level token prediction the good models already are. If you want to dismiss all that by saying “but it’s just prediction,” you’re basically saying deduction doesn’t count unless it’s done by a human. That’s just redefining words to try and win an Internet argument.

They do quite a bit more than that. There's a good bit of reasoning that comes into play and newer models (really beginning with o3 on the ChatGPT side) can do multi-step reasoning where it'll first determine what the user is actually seeking, then determine what it needs to provide that, then begin the process of response generation based on all of that.

This is not a surprise, just one more data point that LLMs fundamentally suck and cannot be trusted.

Huh? LLMs are not perfect and are not expert-level in every single thing ever. But that doesn't mean they suck. Nothing does everything. A great LLM can fail to produce a perfect original proof but still be excellent at helping people adjust the tone of their writing or understanding interactions with others or developing communication skills, developing coping skills, or learning new subjects quickly. I've used ChatGPT for everything from landscaping to plumbing successfully. Right now it's helping to guide my diet, tracking macros and suggesting strategies and recipes to remain on target.

LLMs are a tool with use cases where they work well and use cases where they don't. They actually have a very wide set of use cases. A hammer doesn't suck just because I can't use it to cut my grass. That's not a use case where it excels. But a hammer is a perfect tool for hammering nails into wood and it's pretty decent at putting holes in drywall. Let's not throw out LLMs just because they don't do everything everywhere perfectly at all times. They're a brand new novel tool that's suddenly been put into millions of peoples' hands. And it's been massively improved over the past few years to expand its usefulness. But it's still just a tool.