Live data from Hacker News

Claude can now search the web

anthropic.com

591–600 of 758 posts

Re: Claude can now search the web

#591
post #381

It badly hallucinated in my test. I asked it "Rust crate to access Postgres with Arrow support" and it made up an arrow-postgres crate. It even gave sample Rust code using this fictional crate! Below is its response (code example omitted): I can recommend a Rust crate for accessing PostgreSQL with Arrow support. The primary crate you'll want to use is arrow-postgres, which combines the PostgreSQL connectivity of the…

This was pretty much my first experience with LLM code generation when these things first came out. It's still a present issue whenever I go light on prompt details and I _always_ get caught out by it and it _always_ infuriates me. I'm sure there are endless discussions on front running overconfident false positives and being better at prompting and seeding a project context, but 1-2 years into this world is like 20…

Often times I come up with a prompt, then stick the prompt in an LLM to enhance / identify what I’ve left out, then finally actually execute the prompt.

Re: Claude can now search the web

#592

Earlier quoted context omitted.

No, whether I curl it, or I use a browser, or an LLM, it is essentially ALL the same, unless of course the LLM crawls it by itself, without human interaction. If the scale bothers you, block it, just like how you would block any other crawlers. Other than that, we all wanted "ease-of-access" (not me though), and now we have it. It does not change anything.

What if the crawlers are faking their identity (As they are doing it right now)

Well, how do we deal with it in terms of DDoS?

Re: Claude can now search the web

#594

Earlier quoted context omitted.

> I asked it "Rust crate to access Postgres with Arrow support" Is that how you actually use llms? Like a Google search box?

“Prompting” is kind of a myth honestly. Think about it, how much marginal influence does it really have if you say OP’s version vs a fully formed sentence? The keywords are what gets it in the area.

I find that providing more context and details initially leads to far more success for my uses. Once there’s a bit of context, I can start barking terms and commands tersely.

Re: Claude can now search the web

#595
post #431

Earlier quoted context omitted.

The idea behind UBI is that people do jobs that they want to do...

Right! So everyone would choose to pursue passions/interests/leisure. We would be going into debt with no meaningful benefit to the taxpayer. Direct malinvestment.

Isn’t the investment to avoid a revolution? To avoid those that cannot find work from dismantling and tearing down everything around them so they can get what they need. Some might consider that to be a benefit to taxpayers and not a poor investment.

Re: Claude can now search the web

#596

It badly hallucinated in my test. I asked it "Rust crate to access Postgres with Arrow support" and it made up an arrow-postgres crate. It even gave sample Rust code using this fictional crate! Below is its response (code example omitted): I can recommend a Rust crate for accessing PostgreSQL with Arrow support. The primary crate you'll want to use is arrow-postgres, which combines the PostgreSQL connectivity of the…

> I asked it "Rust crate to access Postgres with Arrow support" Is that how you actually use llms? Like a Google search box?

Totally agree here. I tried the following and had a very different experience:

"Answer as if you're a senior software engineer giving advice to a less experienced software engineer. I'm looking for a Rust crate to access PostgreSQL with Apache Arrow support. How should I proceed? What are the pluses and minuses of my various options?"

Re: Claude can now search the web

#597

Earlier quoted context omitted.

It's just a matter of time. Your statement assumes AI won't help to develop robotics. Robotics is the big unlock of AI since the world is continuous and messy; not discrete. Training a massively complex equation to handle this is actually a really good approach.

I'm not sure about that. For them to actually be economically useful is a high bar. More so than you think - it isn't just our brains but our strength, metabolisms, and more in a single package. For example you need them to: - High energy requirements in varied env's: Run all day (and maybe all night too which MAY be advantage against humans). In many environments this means much better power sources than current bat…

From a very inhuman perspective, and one I don't find appropriate to generally use: A human physical worker is a high capital and operational expense. A robot may not have such high costs in the end.

Before a human physical worker can start being productive, they need to be educated for 10-16+ years, while being fed, clothed, sheltered and entertained. Then they require ongoing income to fund their personal food, clothing and shelter, as well as many varieties of entertainment and community to maintain long-term psychological well-being.

A robot strips so much of this down to energy in, energy out. The durability and adaptability of a robot can be optimized to the kinds of work it will do, and unit economics will design a way to make accessible the capital cost of preparing a robot for service.

Emotional opinions on AI aside, we will I think see many additional high-tech support options in the coming decade for physical trades and design trades alike.

Re: Claude can now search the web

#598
post #489

Earlier quoted context omitted.

[flagged]

His gesture was _exactly_ like Hitler's. A gesture he repeated. The gesture was quite different to that he'd used previously for 'giving people his heart'. He's known to be a white supremacist. That is apparently his heritage too. He supports far right parties in Europe. Other 'Republican' politicians have repeated the gesture from the dais; but they seem to have made other excuses. None of the many videos or photos…

> His gesture was _exactly_ like Hitler's.

Hitler's was unlike the general population's, as it had a bend to it.

You can bend reality all you like, but the intent of giving the Hitler salute was not there, as he has said. He's not secretly a nazi, and he's not openly a nazi. He's right-wing, yes. That's not illegal, and it happens to be the majority vote in the US.

The most reasonable criticism is calling it a Roman salute and saying it bears connotations to imperialism, and that it was most recently practiced by Hitler.

I think, if you want to read into his deepest, unspoken intents, he probably compares himself to Caesar more than Hitler. Just like Zuckerberg, and all the other multi-billionaires who want to see themselves as the de-facto leaders of the world.

> He's known to be a white supremacist

No, a bunch of observations leads you to conclude it.

He never showed up at a white supremacist rally.

He lets them speak on his platform.

> He supports far right parties in Europe.

Most right-wing parties in Europe are still socialist by American standards.

For example, the most liberal parliamentary party in Denmark thinks a 40% tax is fine.

If you're a Republican, you're crazy in the eyes of a European.

Specifically, he supports a far-right party in Germany, which is controversial, since there hasn't been popular far-right parties (only fringe ones) since NSDAP.

The big, controversial subject is ending muslim immigration into Europe. The far right becomes the bannermen for this cause, because closing down on immigration is viewed as xenophobic. In the meantime, as this opinion is being suppressed instead of addressed, it continues to grow with the populist movements.

The fact that Elon Musk has opinions on European immigration policy doesn't make him a nazi. Just like being against muslim immigration doesn't make AfD nazis (the German party that he endorsed), just uncannily populist.

> Someone who hadn't meant it would have come back on stage, when it was pointed out to them, and made an apology. Or at least immediately issued a statement/press release.

That's how I read his sentence immediately after the salutes: "My heart goes out to all of you." -- it sounded remarkably like something someone would say when they realize what they did could be viewed as heiling. You don't need to apologize to be a good person.

Re: Claude can now search the web

#599

Earlier quoted context omitted.

The longer I've been in the workforce, the more I realize most humans actually kind of suck at their jobs. LLMs being more human like is the opposite of what I want.

That could very well be because the jobs are effectively useless. By no means does that mean the people are, nor is what the income allows them to do. But most jobs sure do seem pointless.

Many jobs are quite helpful and even necessary, if done for ~2 hours a day. They become "useless" in aggregate when they're forced to be minded by the same person for 8 hours (because of opportunity cost, effects on health well-being, etc., you end up "breaking even" or worse on QoL and net productivity).

Overall economic productivity is high enough that a lot of positions could be split into 2 or 3 short shifts, at full pay - IF you don't factor in the various financial boondoggles that we've gotten ourselves wrapped up in. If you made the decision to wipe out a lot of these obligations (mostly to rich people), we could get to that kind of set-up, solvently.

Re: Claude can now search the web

#600

Earlier quoted context omitted.

If chatgpt is scraping the web, why can they not link tokens to source of token? being able to cite where they learned something would explode the value of their chatbot. At least a couple of orders of magnitude more value. Without this chatbots are mostly a coding-autocomplete tool for me—lots of people have takes, but it's the tying into the internet that makes a take from an unknown entity really valuable. Perplex…

That's asking for the life stories and photos and pedigrees and family histories of all the chickens that went into your McNuggets. It's just not the way LLMs work. It's an enormous vat of pink slime of unknown origins, blended and stirred extremely well. https://en.wikipedia.org/wiki/Pink_slime

you sort of can do (a decent approximation of) this, it’s just even the approximate version is impractical for computational reasons.

https://www.anthropic.com/news/influence-functions

Post reply on HN