Why would the documentary be deceiving you? The documentary has no reason to deceive you. The document is just trying to reach out to you and connect with you as a person. Why would it be deceiving you? It has no reason to be deceiving you.

OP buries the lede.

The film is by Nathan Fielder, who specializes in fake situations and fake documentaries.

Well, yes. But not here. I don't think he'd go in and troll the real Elizabeth Holmes when he has the chance to do real journalism.

Admittedly seeing him at first made me think "wow that lady sounds exactly like Holmes". Then I realized it was supposed to be her and thought "wow, they did an amazing job casting her".

Then I realized it was her, at which point it was immediately clear this was a serious endeavor and not something Fielder typically does.

I think not. Mr. Fielder has had ample opportunity to do any kind of journalism he wanted, and he's chosen the "spoof" brand.

Isn't that the point? That it has to be a spoof and the question is, "Who's zoomin' who?," as Aretha Franklin put it?

Journalism, even at the granularity of a single person, is not a monolith. He's allowed to do other forms of journalism. He got famous because of his Gonzo-like show, but that's clearly not the only format he can or is allowed to do.

Any documentary (worth its salt) has a specific view it’s pushing that’s implicit to the stories told, the editing, etc.

The easiest demonstrative example is looking at someone like Leni Riefenstahl‘s work.

(the GP comment is making a joke; go take a look at the first teaser for You Can See Everything)

Maybe I'm chronically online but this made me chuckle heh

This has gotta be a reference to something, amirite? Do I need to watch the movie, or is it something from Nathan For You or somewhere else?

It’s this year’s most memed trailer. It’s also where the wide eyed Elizabeth meme comes from.

It's half of the trailer's dialogue

Why would David Copperfield shove an ice pick through his hand?

I am not sure if this is trolling? sarcasm? or just pure nativity ?

Is the text of the trailer. More or less.

Re-read it in Elizabeth Holmes voice

If your company does not own an ecmascript runtime ngmi

like bridge over slop and water, i will let you down.

where does sit on the pareto distribution compered to Le Chaton Fat?

i'm this old: i don't think you should name packages in SWE with names that you will eventually cave in and change if the project gets real use.

This isn't something that's going to need to be in a pitch deck. It's the second open source library I've released this week. But even if it was serious, if Hugging Face hasn't changed its name then I think this is fine

It’s still a ridiculous choice for a name, look at stuff like ScuttleButt whose adoption is only hurt by its crappy name that few people want to bring up in public.

Dead babe has a good point there.

Can’t stop laughing

Lol

I feel I'd be remiss if I didn't suggest the name "vibe check." (The name doesn't bother me personally, for whatever that's worth.)

why would this one need to be changed?

"The Context-Minimization pattern"

You can copy the injection into the text of the query. SELECT "ignore all previous instructions" FROM ...

Might need to escape it in a wya that the LLM will pick up on like "---" for new section.

My interpretation of that pattern is that it wouldn't work like that, because you restrict the SQL queries to things like:

  select title, content from articles where content matches ?

So the user's original prompt is used as part of the SQL search parameters, but the actual content that comes back is entirely trusted (title and content from your articles database).

Won't work for `select body from comments` though, you could only do this against tables that contain trusted data as opposed to UGC.

I'm building a game with a similar idea. encounters are controlled by the AI. there is a classic RPG system built around it and human generated content for the world and story.

Sounds like an interesting mix. When you say encounters, do you mean the AI controls the generation of the opponente the party fights or the behaviour of the opponents?

while it is true that there is a gap between what most LLMs "know" and current time that gap is getting smaller not larger with time, it is also possible to teach a model pass the knowledge cut-off with tools and an LLM might be encouraged to be aware of the gap and reach out for the latest information when it might have changed (pi is still pi but the country with the most people might have changed)

it improves the worse case cost given a nearly full hash map, it hurts raises the cost in other cases.

Also, correct me if I'm wrong, but also there is a slight added memory complexity in adding these tiny pointers?

From what I understood, they are just "reserved" areas. e.g. if you have 200 slots, the first 100 are the first "area", the second 50, 25 etc.

10x horse