tomrod

Its worse than that though, as it isn't a loyalty test -- US federal executive agencies have to adopt it whether they like it or not.

Yeah, I dislike it too. As do most of the agencies forced to swap over to it I reckon, since "SI" is a vanity project and "AI" is what the rest of the world understands it to be.

Definitely see where you might feel so. I provided an analogy that may help anchor - why ML fails extrapolation is left as an exercise to the determined reader! :)

The surface for "interpolation" is fairly broad for LLMs but it isn't infinite -- extrapolation still exists!

I don't think it can whistle while chewing gum, no.

I think it can patch together components of tasks represented in its training data to do demonstrably new things, yes, which does count as "perform tasks not in its training data" with the limitation that some form of the task must be present. Example: writing an ode to a sales lead for a job well done in 2032 for the first quarter. Odes exist, the concept of sales performance exist, good job performance, and notions of time. Stringing those together are good examples of interpolation but not extrapolation, as it were.

I'm asking because it sounds both fun and an effective case study.

Loe's Paw?

Thanks, Dang.

Trigger.dev wouldn't rotate customer BYOK. What I was saying is they disable customer use of their service until rotation occurs or is confirmed by the customer.

Like LLMs, insufficient context leads to insufficient answers.

The small snippet, no, you're correct, he doesn't call for expropriation, though on its own it does seem an odd thing to say.

The full podcast, no, you're correct, the interpretation folks make isn't reflected in the podcast itself.

In the full scope of what the public believes it knows about Sam Altman, from allegations of familial abuse, to the board fiasco, to the expropriation of art and other works by humans to train the frontier models with reckless abandon, one sees that the absolute cynicism by the public writ large regarding anything Mr. Altman does or says is probably appropriate.

Perhaps unfortunately for Mr. Altman, as is the case for most public figures, his statements aren't seen as encapsulated logic statements to be taken individually, but in the scope of his broader public persona.

Math envisions and communicates logic-based structures and objects that, as a side effect, may be relevant to certain real world problems decades to centuries after the initial research.

I'd argue your statement is treating this discovery process as an engineering exercise: there exists an unexplored bounded space of logical axioms within which we want to solve the relevant problems and paradoxes. Yes, AI will help under such a framework, by giving mindless and rote repetition to brute force a solution.

But mathematics begins from a different set of primitives - the space of exploration is not guaranteed to be bounded, so the solution context is derived after a lot of trial and error. This is a different philosophical problem being solved.

I'm not sure I'm seeing that the AI solutions offered up by frontier level AI are exploring the same space as mathematics. Overlaps, certainly. But "find new and interesting problems for the sake of fun, beauty, competitive and collaborative prestige" isn't motivational to an AI.

What a strange thing to say when the bulk of consumers are frustrated by a monopsonistic situation.

It does happen to ignore individual property rights and restricting the free market, as well as sapping away consumer welfare from the older market equilibrium to be fed to the frontier lab beasts through increased prices and through utilization by legacy consumers through the LLM mark of the transaction, so there is that.

(I say some of this tongue in cheek).

How do you know parent commentator HASN'T had a negative article written about them?

I'd argue its probably lower than anticipated unless you count Google's AI search mode in the LLM column instead of the search column. As a forced inject, it could go either way.

Huh. Why was this flagged?

My scientific computing and general purpose computing needs beg to differ. Begone, foul stereotype!

I agree, if I am saying or doing cartoonishly evil things without any guardrails and with reckless abandon I too hate it when people call it out directly.

Depends on the TOS, but, blocking service until confirmed or decommed would get the trick done.

To use a fearful, apt analogy: resolving the operational issue isn't just plugging the hole in the dam, its also fixing the damage by pumping the water out and then addressing the damages.

In the pure economics/marketing sense, a Google Maps listing is an advertisement.

I expect for most people that these forms of advertising - directory listings, basic notifications, etc. - aren't a level of cognitive pollution that they notice vs. overly bright billboards, banner ads, search injects, or LLM-injected marginally relevant sources.

Based on your prior comment, I think you mean "effort taken for delivery" in an abstract way. I think that is part of it, but "increased revenue through additional or higher value transactions" is what I think about when I say "increase in business." This is something that can be affected by the true sense of advertising - giving notice about your business - but has diminishing marginal returns.

They are running loads of fiber. If it's worth the trillions then it's worth some easement beside a railroad.

> Why does it need a whole manipulative extraction industry around it?

It doesn't, especially at the aggregate scale it is at!

What do you mean by "real world logic"? Observed behavior, logical abstraction, or handwavy ivorytoweredness?

While we are approaching this with different modes, we broadly agree that advertising is cognitive pollution.

The cafe putting up a sign, hours, and a menu in the window are forms of advertising. Based on your other comment, I suspect you are painting with a brush of "all" when you mean digital advertising or targeted digital advertising.

No, I disagree. I'm a strong cynic of advertising returns generally. However, basic notification that something exists brings business.

This applies to the mom-and-pop selling something around the corner, not Coca-Cola. For Coca-Cola, their advertising money likely has negative returns at the scale they spend it.

I go back and forth. It's useful to maintain certain social connections, which I find harder to do as I'm getting older (weird to say that, I am on the young xennial side).

This is a centrally-run implementation of CloudFlare's VibeSDK. It seems pretty solid for prototypes though there are some missing features - I can't figure out how to specify models in the AI Gateway specifically and the build model is fairly weak, it doesn't seem to set up any infra but workers / doesn't have a JSONC editor directly or let you configure infra. A drag-and-drop or scripting engine would be nice.

It does allow temporary git creds so you can clone to a repo and upload to one of your own, so that's nice and helps with the finish engineering portion.