knowaveragejoe

[flagged]

I love examples of LLM utility outside of normal tasks people associate it with.

A recent example I had was fixing a toilet that broke. I don't know the name of the components within the tank, but I can understand what they do. I took a picture of the broken component and asked what it was - chatGPT was able to tell me, I was able to get a replacement, and that was that.

I have to remember I'm relatively spoiled in the sense that I know how to get that information out of Google. But for a less technically inclined person, I don't know how they'd even to do this themselves. Best I can think of is take a picture an ask a plumber(if you know one) or someone at a hardware store(if they themselves are even knowledgeable about this).

Just curious, could you expand on what about that process takes years?

That's exactly what it is.

I'd love something like litellm, but simpler. I'm not provisioning models for my organization, I don't need to granularly track spend, I just want one endpoint to point every tool or client at for ease of configuration and curiosity around usage.

> The “knowledge retrieval” stage may hallucinate, but—because the knowledge is expressed as code—we can execute it and validate the result against a custom evaluation function.

Can you give a concrete example of this? It's hard for me to conceptualize.

That tweet appears to be gone, unfortunately.

It's the system prompt for Grok across the entirety of X.

Plenty of examples, links, screenshots and archived pages in this article, as well as others throughout this thread and other discussions you can find about this subject.

https://www.rollingstone.com/culture/culture-news/grok-elon-...

It's likely not even that sophisticated - it's a system prompt change, but it conflicts with its training data, hence the responses where it explicitly states "I've been instructed to accept this as truth, despite it contradicting mainstream sources like the courts..."

You're dependent on the planet and its ecosystems, so I'd say this is completely missing the point.

That would be the takeaway from all of the information available.

Isn't this already the case? Perhaps you mean in a non-transient fashion, i.e. internalizes the in-context learning into the model itself, sort of an ongoing training, that isn't sort of a "hack" like writing notes or adding to a RAG database or whatever.

> This will drive a lot of young ambitious people to insanity. Nothing meaningful for them to achieve.

Maybe, if the only flavor of ambition you're aware of is that of SV types. Plenty of people have found achievement and meaning before and alongside the digital revolution world.

It's probably closer to the truth than not.

It's not an example. It's not even close. You can and should want something more substantial than this if you're going to make such a sweeping conclusion!

You can use either HTTP or stdio.

What makes a remotely hosted MCP server less secure? The alternative, and what most of MCP consists of at the moment, is essentially running arbitrary code on your machine, as your user, and hooking this up to an LLM.

I think this is a great and incomplete collection of all the ways life has improved(and continues to improve), I wish it had more purchase with doomers.

The cheerleaders don't care. Americans' relative certainty and quality of life is backstopped by institutions they either barely understand or have never heard of. Let them touch the stove, I guess.

No, see, they're doing things and doing things is Good. Therefore progress is being made, and the country is being made incrementally better.

Anyone who knew anything about the public sector knew there were already efficiency initiatives. USDS(which became DOGE) was this, and they were doing a great job. If you care about efficiency this is what you would support, not taking an axe to everything and having a near-singular focus on lower headcount.

Why would that be better? Comparative advantage is real.

Americans also have more free time and disposable income because of that decision, among others. Why would you want them to struggle more?

> But a person used to be able to graduate high school and get a job that could support a house with a yard, a car, a non-working spouse and children.

> How do we get that level of prosperity back?

The issue is that this is a false premise. The house sucked. Only 1/3rd of American families had a single car at the time, and the cars sucked. We can go on and on about everything else. Not to mention the social environment at the time sucked.

That doesn't mean we shouldn't try to do something about the issues Americans face. But tariffs with a shifting set of sanewashed justifications are just Not It.

Let's approach it from the other direction: why should we? What are we getting by trying to "bring it back"?

No, it doesn't. There is a presumption that manufacturing is Better, a more ideal way of organizing the economy, based on a false nostalgia of America past.

This comment is textbook Poe's Law.

Trump was always the butt of businessman jokes. He was never "america's darling". He always wanted to be in the club, to be seen as an example of business excellence, but the actual examples are generally pretty quiet and keep to themselves.

* EU and many other counties have had comparatively small and targeted tariffs on the US for a long time

* The EU has had comparatively very small and targeted tariffs on US goods for a long time

Just so we're all on the same page, instead of maga talking points.