blakestacey

joined 2 years ago
MODERATOR OF
[–] blakestacey@awful.systems 13 points 2 months ago

Large Reasoning Models

May the coiners of this jargon step on Lego until the end of days

[–] blakestacey@awful.systems 19 points 2 months ago (12 children)

From an article about a boutique brand that sells books to rich people:

Assouline has made its name publishing tomes that sell for $1,000 or more.

Oh, so they publish textbooks.

"They represent stealth wealth, intended to tell you what your hosts are about and to provide visual evidence: that the owners are people of wealth, education and taste."

🎶 Please allow me to introduce myself 🎶

[–] blakestacey@awful.systems 8 points 3 months ago (1 children)

For some reason, the news of Red Lobster's bankruptcy seems like a long time ago. I would have sworn that I read this story about it before the solar eclipse.

Of course, the actual reasons Red Lobster is circling the drain are more complicated than a runaway shrimp promotion. Business Insider’s Emily Stewart explained the long pattern of bad financial decisions that spelled doom for the restaurant—the worst of all being the divestment of Red Lobster’s property holdings in order to rent them back on punitive leases, adding massive overhead. (As Ray Kroc knows, you’re in the real estate business!) But after talking to many Red Lobster employees over the past month—some of whom were laid off without any notice last week—what I can say with confidence is that the Endless Shrimp deal was hell on earth for the servers, cooks, and bussers who’ve been keeping Red Lobster afloat. They told me the deal was a fitting capstone to an iconic if deeply mediocre chain that’s been drifting out to sea for some time. [...] “You had groups coming in expecting to feed their whole family with one order of endless shrimp,” Josie said. “I would get screamed at.” She already had her share of Cheddar Bay Biscuit battle stories, but the shrimp was something else: “It tops any customer service experience I’ve had. Some people are just a different type of stupid, and they all wander into Red Lobster.”

[–] blakestacey@awful.systems 8 points 3 months ago* (last edited 3 months ago)

Yeah, Krugman appearing on the roster surprised me too. While I haven't pored over everything he's blogged and microblogged, he hasn't sent up red flags that I recall. E.g., here he is in 2009:

Oh, Kay. Greg Mankiw looks at a graph showing that children of high-income families do better on tests, and suggests that it’s largely about inherited talent: smart people make lots of money, and also have smart kids.

But, you know, there’s lots of evidence that there’s more to it than that. For example: students with low test scores from high-income families are slightly more likely to finish college than students with high test scores from low-income families.

It’s comforting to think that we live in a meritocracy. But we don’t.

And in 2014:

There are many negative things you can say about Paul Ryan, chairman of the House Budget Committee and the G.O.P.’s de facto intellectual leader. But you have to admit that he’s a very articulate guy, an expert at sounding as if he knows what he’s talking about.

So it’s comical, in a way, to see [Paul] Ryan trying to explain away some recent remarks in which he attributed persistent poverty to a “culture, in our inner cities in particular, of men not working and just generations of men not even thinking about working.” He was, he says, simply being “inarticulate.” How could anyone suggest that it was a racial dog-whistle? Why, he even cited the work of serious scholars — people like Charles Murray, most famous for arguing that blacks are genetically inferior to whites. Oh, wait.

I suppose it's possible that he was invited to an e-mail list in the late '90s and never bothered to unsubscribe, or something like that.

[–] blakestacey@awful.systems 20 points 3 months ago* (last edited 3 months ago)

From the documentation:

While reasoning tokens are not visible via the API, they still occupy space in the model's context window and are billed as output tokens.

Huh.

[–] blakestacey@awful.systems 9 points 3 months ago (5 children)

The New Yorker gamely tries to find some merit, any at all in the writings of Dimes Square darling Honor Levy. For example:

In the story “Little Lock,” which portrays the emotional toll of having to always make these calculations, the narrator introduces herself as a “brat” and confesses that she can’t resist spilling her secrets, which she defines as “my most shameful thoughts,” and also as “sacred and special.”

I'm really scraping the bottom of the barrel for extremely online ways to express the dull thud of banality here. "So profound, very wow"? "You mean it's all shit? —Always has been."

She mixes provocation with needy propitiation

Right-click thesaurus to the rescue!

But the narrator’s shameful thoughts, which are supposed to set her apart, feel painfully ordinary. The story, like many of Levy’s stories, is too hermetically sealed in its own self-absorption to understand when it is expressing a universal experience. Elsewhere, the book’s solipsism renders it unintelligible, overly delighted by the music of its own style—the drama of its own specialness—and unable to provide needed context.

So, it's bad. Are you incapable of admitting when something is just bad?

[–] blakestacey@awful.systems 14 points 3 months ago (3 children)

I often use prompts

Well, there's your problem

[–] blakestacey@awful.systems 11 points 3 months ago (2 children)

and hot young singles in your area have a bridge in Brooklyn to sell

on the blockchain

[–] blakestacey@awful.systems 8 points 3 months ago (2 children)

So many techbros have decided to scrape the fediverse that they all blur together now... I was able to dig up this:

"I hear I’m supposed to experiment with tech not people, and must not use data for unintended purposes without explicit consent. That all sounds great. But what does it mean?" He whined.

[–] blakestacey@awful.systems 29 points 3 months ago (27 children)

When you don’t have anything new, use brute force. Just as GPT-4 was eight instances of GPT-3 in a trenchcoat, o1 is GPT-4o, but running each query multiple times and evaluating the results. o1 even says “Thought for [number] seconds” so you can be impressed how hard it’s “thinking.”.

This “thinking” costs money. o1 increases accuracy by taking much longer for everything, so it costs developers three to four times as much per token as GPT-4o.

Because the industry wasn't doing enough climate damage already.... Let's quadruple the carbon we shit into the air!

view more: ‹ prev next ›