Hacker Newsnew | past | comments | ask | show | jobs | submit | sdwr's commentslogin

Maybe human part inspectors should get treats for inspecting parts correctly

Awesome idea! The entity tracking is very exciting, most interesting part imo

I think the budget is noticeable in the sentiment analysis unfortunately, the tags and entity recognition are good but the sentiment ratings themselves seem pretty sloppy.


I think it's mostly prompting, but I will be experimenting with this more. The prompt currently is garbage IMO

    You are an expert analyst of the Hacker News community. Analyze submissions for
    the underlying ideas, concepts, technologies, and entities being discussed.

    Write all summaries in third-person analytical prose. Do NOT start sentences
    with "The user", "The commenter", "The author", or "This post". Instead, lead
    with the substance: describe the idea, argument, or phenomenon directly.

    Good: "Decentralized identity systems could reduce reliance on corporate
    gatekeepers." Bad: "The user discusses how decentralized identity systems work."

(Source: https://github.com/devrupt-io/ethos/blob/67670eb2855b84d389d...)

Garbage, why? That is the insightful bit you chose to omit. How would you do it instead?

It leaves a lot of interpretation to the model. For example it doesn't give any guidance on concept naming or disambiguation, which leaves all of that work to the JSON schema.

In my experience it's much more effective to reference key terms or ideas in the JSON schema and then explain those and their constraints in the system prompt.

This is one reason why people often think one model performs better than another for tasks they are both capable of. The real question IMO becomes, does porking all of that extra input prompt (a) eat too much context or (b) increase cost too much.

We will put an update on this in the future and post it in our blog, https://blog.devrupt.io/


Yeah, or just the steam text guides would be a huge advantage.

I really doubt it's playing completely blind


Thanks to another comment here I went looking for the strategy guides that are injected. To save everyone else the trouble, here [0]. Look at (e.g.) default/STRATEGY.md.jinja. Also adding a permalink [1] for future readers' sake.

[0]: https://github.com/coder/balatrollm/tree/main/src/balatrollm...

[1]: https://github.com/coder/balatrollm/blob/a245a0c2b960b91262c...


Gwern is the nerd's nerd

I'm sure the goal is that reading files is something you debug, not monitor, like individual network requests in a browser.

An error message helps people skirt the restriction, by providing immediate feedback on what does/doesn't get flagged.

Same idea as shadow banning, ban waves, and generic errors for sensitive actions


That is only acceptable for non paying customers.

> arXiv "emergent" (the count of AI papers about emergence) has a clear, unambiguous R² maximum. The other four are monotonically better fit by a line

The only metric going infinite is the one that measures hype


What does clean room mean in this context? They built it from the assets with the game as a reference , but didn't look at the engine source code?

It means they didn't reference any existing or decompiled code from the original client. None of it is directly infringing on any copyright, though it may be doing so indirectly since there have been plenty of lawsuits for tools that contain no copyrighted information can but can used to facilitate infringement (e.g. a tool that decompiles a game ROM)

How about the assets?

I assume you, the player, have to provide the assets yourself, and the game won't run without them. Since the code does not contain the assets, there is no copyright infringement.

As long as the assets dont contain code, they're kind of fair game. The rule of thumb is you cannot redistribute them, but if the person owns a legal copy you can point to them on their local system. It is not up to you to figure out if they're pointing to a pirated copy or a legitimate copy mind you.

It's very indirect. The message is "the government is soft on fraud, partially because of liberal values", but the author does everything possible to not actually say it.

I don't think that's indirect at all. It's pretty clearly what did in fact happen in Minnesota. I don't read the author as claiming it's endemic to liberal values, any more than the isomorphic pathologies are endemic to the finance industry (which Patrick also writes about), or the defense industry. Again: it's easy to find Democratic sources saying the same thing.

Why is it so difficult for people to acknowledge that Minnesota fucked this up badly? What is that going to cost us? The attempts to downplay it seem pretty delusive.


Well, we live in a world where someone ran for president on the basis of "stopping welfare fraud" that turned out to be mostly a myth, so, you know, context is a thing.

As the article literally says, a whole bunch of people got sent to prison, that seems like pretty solid evidence.

The question is: now what?


Run the rest of them down? Figure out the total scope of the fraud, so we can enact countermeasures to prevent anything like it from happening again?

Fraud targeting social welfare programs is a grave crime; it strikes at public support for those programs. It enriches criminals very specifically at the expense of those people who would most benefit from the program.


All I remember hearing about this is how creepy and racist it was to look for kids at a black-owned daycare. It was a scam the whole time?

According to OP, there is substantial evidence indicating about 50% of the daycares are scams. I've seen Nick Shirley's video, I don't think he demonstrated any concrete about any of the sites he visited (he's not a very good investigator), but if the 50% number is correct... well, the broken clock was probably right at least a couple of times that day.

The 2019 OLA report "Child Care Assistance Program: Assessment of Fraud Allegations" is what makes the claim that greater than 50% of reimbursements to child care providers under these specific programs were fraudulent. That estimate is broadly and bipartisanly considered to be directionally true.

Have you read the document? I often find these things that we believe to be true wind up being a game of telephone. We further have no prior (is the metric of this sort of like how everyone has committed a crime given sufficient prosecutorial attention)?

I've also never heard of this report before this year as someone quite attuned to what happens in Minnesota.

One document contributor contends, notably, that if "adult “employees” spend hours in hallways chatting with other adults [...] the entire amount paid to that provider in a given year is the fraud amount."

I'll leave the assessment of that definition to readers. https://www.auditor.leg.state.mn.us/sreview/ccap.pdf


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: