Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Anthropic: "Applicants should not use AI assistants"

simonwillison.net

Post reply on HN
Print view View original

381–390 of 419 posts

Re: Anthropic: "Applicants should not use AI assistants"

#381

Post by lupire »

plagiarist wrote: post #342
imchillyb wrote: post #295

Earlier quoted context omitted.

There are LEOs that were prosecuted by states and the federal government for not taking action while children were being shot by another child. LEOs are expected to take fire to protect civilians. Protect & Serve is their credo. I wouldn’t trust LEOs to protect me, so I sure as hell fire am not trusting a low paid rent-a-cop to perform a similar duty. Nope. I believe that my mindset is prevalent and not an outlier.

I would trust a security guard more. They have consequences for misconduct or failure to do their job. (Assuming they aren't an on-duty LEO who is "overemployed.") Not enough that I think they'd protect me in a situation that requires a gun. Just more than a cop.

A security guard's job is to act as a witness and deterrent, not to intervene to protect you.

Only a private bodyguard can be expected to fight for you.

Observe and Report.

Re: Anthropic: "Applicants should not use AI assistants"

#382

Post by sshine »

raincole wrote: post #87
hollowturtle wrote: post #46

Earlier quoted context omitted.

Can't disagree more. Talent is built and perfected upon thousands hours practice, LLMs just make you lazy. One thing people with seniority in the field don't realize, as I guess you are, is that LLMs don't help develop "muscle memory" in young practioners, it just make them miserable, often caged in an infinite feedback loop of bug fixing or trying to untangle a code mess. They may extract some value by using it for…

Use LLM. But do not let it be the sole source of your information for any particular field. I think it's one of the most important disciplines the younger generation - to be honest, all generations - will have to learn. I have a rule for myself as a non-native English speaker: Any day I ask LLMs to fix my English, I must read 10 pages from traditionally published books (preferably pre-2023). Just to prevent LLM from…

I use LLMs as a translation tool, and make sure to generate JSON flashcards.

Sometimes it is more important to get a point across in another language than it is to learn that language. Computers being automatable, you can use it to create a backlog for when you skipped learning so that you can maintain some control of your habit of not learning what you're saying.

Re: Anthropic: "Applicants should not use AI assistants"

#383

Post by sshine »

eknkc wrote: post #162
hollowturtle wrote: post #46

Earlier quoted context omitted.

Can't disagree more. Talent is built and perfected upon thousands hours practice, LLMs just make you lazy. One thing people with seniority in the field don't realize, as I guess you are, is that LLMs don't help develop "muscle memory" in young practioners, it just make them miserable, often caged in an infinite feedback loop of bug fixing or trying to untangle a code mess. They may extract some value by using it for…

I drove cars before the sat nav systems and when I visited somewhere, I'd learn how to drive to there. The second drive would be from memory. However, as soon as I started relying on sat navs, I became dependent on them. I can not drive to a lot of places that I visited more than once without a sat nav these days (and I'm getting older, that's a part of it too). I wonder if the same thing will happen with coding and…

On a roadtrip ten years back we chose to navigate by map and compass, and avoid highways.

With sat nav I don't even try to read the exit signs; I just follow the blue line. It takes me 10-20 drives somewhere before I have the muscle memory, and I never made an active mental effort.

Going somewhere by public transportation or foot, e.g. a large homogenic parking lot complex, I consciously make an effort to take mental pictures so I can backtrack or traverse perfectly the second time; in spite of that being mentally challenging, it's still the easiest way I have.

I cannot assemble the hardware that I write code for. This is in spite of having access to both the soldering equipment, the parts and the colleagues who are willing to help me.

At some point all skills become abstract; efficiency is traded for flexibility when you keep doing the same thing for a very long time.

I can still drive a stick shift, but maybe not in 20 years.

Re: Anthropic: "Applicants should not use AI assistants"

#384

Post by lupire »

Muromec wrote: post #159
radu_floricica wrote: post #115

I'll be the contrarian and say that I don't find anything wrong with this, and if I were a candidate I'd simply take this as useful information for the application process. They do encourage use of AI, but they're asking nicely to write my own texts for the application - that's a reasonable request, and I'd have nothing against complying. sshine reply above is coming from a very conflictual mindset. "Can I still use…

It is very sensible position and I think the quote is a bit out of context, but the important part here is who it is coming from -- the company that makes money on both cheating in the job application process (which harms employers) and replacing said jobs with AI, or at least creating another excuse for layoffs (which harms the employees). In a sense, they poisoned the well and don't want to drink from it now. Looki…

It's worse. Anthropic is the one claiming to be a water-enhancing company that you should drink from instead of the posioned well. But they are begging you not to feed them their water.

Re: Anthropic: "Applicants should not use AI assistants"

#385

Post by lupire »

YurgenJurgensen wrote: post #224
jack_pp wrote: post #186

Earlier quoted context omitted.

Making your comms friendlier or whatever is one of the myriad ways to use LLMs. Maybe you personally have "LLMs shoved down your throat" by your corporate overlords. No one in their right mind can say that LLMs were created for such a purpose, it just so happens you can use it in this way.

LLMs aren’t making your comms friendlier; they’re just making them more vapid. When I see the language that ChatGPT spits out when you tell it to be friendly, I immediately think ‘okay, what is this person trying to sell me?’

You are you, here. Think of the statistically average person and what they consider friendly.

Re: Anthropic: "Applicants should not use AI assistants"

#386

Post by lupire »

myfonj wrote: post #317
einsteinx2 wrote: post #264

Earlier quoted context omitted.

Wow I completely didn’t notice that until I read your comment. My brain must have automatically filled in the correct word. I had to go back and re-read it to confirm.

Well English is not my first language, so I probably tend to go through text more slowly and/or scan the text differently, so I have higher chance stumbling upon these oddities. (I can confirm that sometimes I see unexpected amount of misspelt homophones or usage of strangely related words.) Seeing two distinct LLM chats gloss over this particular nuance in almost identical way was really interesting. Grok, on the ot…

Typical English language learner, apologizing for having a better grasp of English than a native speaker.

Re: Anthropic: "Applicants should not use AI assistants"

#388

Post by lupire »

ecocentrik wrote: post #355
nertirs1 wrote: post #326

Earlier quoted context omitted.

The same could be said about a lot of things, like being able to write a functional solution to a leet code puzzle on a black board in front of an audience. IMHO, an effective interview process should attempt to mimic the position for which a person is applying. Making a candidate jump through hoops is a bit disrespectful.

IMO the interview process should help the employer correctly identify qualified candidates. Respect for the candidates is important but some candidates should absolutely be prepared to jump through hoops.

Jumping through hoops isn't the only way to demonstrate professional skill. I am the lion tamer, not the lion.

Re: Anthropic: "Applicants should not use AI assistants"

#389

Post by blagie »

knuppar wrote: post #368
blagie wrote: post #344

Earlier quoted context omitted.

The extreme scenario: I live in a home surrounded by miles of fields. There is no one within miles to hear me scream. Without a gun, anyone could come by my home, kill me, rob my home, and be gone before the police would even show up, if I even had a chance to call them. If I didn't call the police, they could literally move in and stay for months before anyone would notice. The reason this does not happen is because…

> The reason this does not happen is because everyone has a gun. Everyone knows I have a gun. If I see you coming on my property, I WILL shoot you. You don't know if the first shot will be a warning shot, birdshot, buckshot, or a 5.56×45mm NATO. You might get lucky and I might not spot you. Or you might be crippled for life. Without guns, crime is free. With guns, crime doesn't pay. Your perceived safety might be hig…

So the problem with a survey like this is that it does not break out among the scenarios I listed:

1) Rural, minimal police, minimal government, large plots, no collective security.

2) Dense, urban, heavy policy, significant government, right housing, extensive collective security.

Indeed, it focuses on the latter. Virtually all of the addresses, photos, and stories talk about cities, or at least towns.

I don't want to over-post so I'll answer the other comments too:

1) Violence does not require more than "very few humans" to "like committing violence." The point of security isn't to protect against the typical individual but the violent outlier.

2) Most violent individuals aren't sophisticated. What's more, one instance of violence has little impact. Serial violence does. If an individual robs one house, that's not enough to live off of. If an individuals robs houses regularly, in an area with guns, they will be shot. That's a pretty good deterrent.

For gun safety to move forward, both sides need to understand each other, and everyone needs to address the major issues of gun advocates, such as:

1) Day-to-day safety (on the scale / in the settings I described)

2) National safety (if Jan 6th had worked, and we had a coup; if China invaded; etc.)

3) Rule-of-law (we do have a 2nd amendment, and changing that would require an amendment)

Otherwise, it's simply a push of more guns versus less guns, with idiotic laws being shoved through opportunistically on both sides.

Re: Anthropic: "Applicants should not use AI assistants"

#390

Post by blagie »

robertlagrant wrote: post #319
blagie wrote: post #260

Earlier quoted context omitted.

> Everyone has a gun on their person with no licensing requirements. Yet at the NRA conference they ban guns. That's not what the NRA is pushing for, any more than there are Democrats pushing for mandatory sex changes for all kids (yes, this is cited on similar right-wing comedy shows, and individuals on the right believe it). Pushing for a right doesn't mean 100% of the population will exercise that right. And yes,…

> yes, this is cited on similar right-wing comedy shows, and individuals on the right believe it Can you give an example? Of course you can find 2 people in the US who believe it, and they held 2 comedy shows where it was said, and it's technically true, but I don't think I've ever seen anything like this said.

I don't log all comedy shows I see, so I can't provide a citation off-hand, but I've heard it plenty of times. However, to see consequences, I might start by reading executive orders:

https://www.whitehouse.gov/presidential-actions/2025/01/prot...

And follow the trail back to how they got there.

https://www.breitbart.com/education/2022/07/07/aclu-national...

You can look around. You'll see many other articles like this one. As with most things, this is distilled into more inflammatory posting once it hits social media or comedy.

Post reply on HN