Hacker News Reader

Comment by blfr | Hacker News Reader

Maybe it's AI psychosis speaking but how can you not be excited? You can outsource virtually all electronic drudgery for $100/mo.

I haven't been this excited about tech since I set up my first home server 20+ years ago. And Claude doesn't sounds like an always-on vacuum cleaner (it was a decommissioned proliant).

View on HN

Personally, I don't have any electronic drudgery to outsource. However, I do have to deal with all the ways that AI has made my life worse.

Really? You don't write dockerfiles, ansible playbooks, never wanted something rewritten into typescript so that it also works client-side, or get stuck in some cloud console?

So much of the drudgery in my day to day is dealing with corporate red tape and ensuring my team is producing the right thing. LLMs don't help much in that regard.

The hard part (where I work) was never the code, it was all of the other BS that needs done (a lot of which happens in crusty old systems that look like they're from the mid 2000's)

Agreed! At the moment I have all the tech I really need. I could think up a million interesting side-projects that AI might help come to fruition. However, in reality I have much more banal concerns - like whether I'll still be able to make a decent living using my skills; and whether I'll ever be able to start a family and own a home.

Perhaps this would be far more exciting if I were younger. But right now I'm just trying to figure out if I can survive.

> Maybe it's AI psychosis speaking but how can you not be excited? You can outsource virtually all electronic drudgery for $100/mo.

"Electronic drudgery" is how I earn my income which pays for my food and shelter. That's why I'm not excited.

We were already getting great for nothing and dealing with that problem to financing custom fit. Now we have mediocre and customized for nothing.

I find it great for experiments to throw away but I wish everyone remembered step 2 more often.

We're redirecting an increasingly large portion of the world's R&D funding into building data centres to run LLMs which can barely speak English ("Good observation - that's the pivot point. I'll land the load bearing backstop.") and which turn what should be 100 lines of code into 4000.

The end result for me has been that I am now a gatekeeper trying to prevent AI slop making its way into systems which I care about, and failing, and getting dragged into meetings with my boss about how my performance has gone down now that I'm chained to this stupid thing and forced to use it for even the most trivial tasks.

I haven't been this upset with PayPal mafia overlords since crypto.

LLMs write samey English that is of significantly higher quality than the median literate native speaker. There's a difference between obnoxious writing and bad writing, and when you claim LLMs can "barely speak English" you've run off the rhetorical cliff of your argument.

I think what the parent poster is getting at is that "LLMs don't sound natural". People don't write the same way an LLM does, which leads to a kind of 'uncanny valley' effect when reading AI text.

Over time, I suspect both human and AI writing will 'converge'. Humans are good at adopting/imitating language from a variety of sources, so I won't be surprised when we start using "load-bearing" in conversation. Simultaneously, LLMs will start sounding more natural.

That's in fact not what they said, though.

It is

I agree with you contra GP that "barely speak English" wildly misrepresents LLM capabilities, but I think they don't write very well, and your characterization is also an overstatement. They often use stylistic elements that don't make sense in context: pointless metaphors, abuse of bullet points and tables when a simple paragraph would do, excessive parataxis, misplaced emphasis, dense technical prose even when unnecessary, occasionally weird word choices.

Those problems are much worse in languages other than English, by the way. I'm not a native speaker but I always interact with LLMs in English because their prose in Italian is terrible, it often reads as though the text had been written in English and then translated.

Obviously there's nothing wrong with each of those stylistic elements in abstract, exactly like there's nothing wrong with em-dashes or the word "absolutely", but I would argue that writing well means choosing styles that match the content; what LLMs appear to be doing instead is forcing the message into a default pattern, which we percieve as forced and therefore tend to dislike.

To be clear: their writing is fine. The primary goal of writing is to convey meaning, and LLMs have been perfectly adequate at that pretty much from day one. I can get over their quirks.

The bar they set was "barely speaks English".

The bar I set was "significantly higher quality than the median literate native speaker".

The bar you're setting is expert use of metaphors, restrained use of bullets and tables and parataxis, proper emphasis, and avoiding dense technical prose.

I suspect you have a higher opinion of the median literate speaker than is warranted.

LLM speak is of "higher" "quality" to a superficial inspection because it uses big words in a grammatically correct way.

It is worse because it fails to communicate information well. When I say it can barely speak English, I mean that it produces sentences which are incomprehensible on the first few reads. I have never found the median literate native speaker to be so difficult to understand, partly because they use a limited vocabulary and avoid waffle.

It can barely speak English in a new and exciting way, one unlike all the previous humans who could barely speak English.

I don't think the claim will bear out that LLM-generated text fares more poorly on comprehensibility tests than those of the median literate speaker. I understand the aesthetic arguments against it, and largely agree, but the technical argument you're making here is almost certainly and starkly false.

Most people write quite badly!

I would describe PIAAC literacy levels 1 and 2 as barely able to comprehend English, and someone who wrote at a corresponding level as barely able to write. They're close to the line of illiteracy, particularly PIAAC 1.

I believe the median person who comprehends English to PIAAC level 3 literacy or above, and writes accordingly, would be more comprehensible than an LLM, if we controlled for correctness and for their own understanding of the subject matter.

I'm not sure I'd agree that the quality is significantly higher than the median literate native speaker. Who do you have in mind? I interact some online with people I meet through sports, and it'd be difficult for me to imagine the LLM output as superior writing. But perhaps that isn't a representative example of the median.

What the hell am I supposed to be excited about? Endless slop? Career change from intellectually stimulating creation to bland gatekeeping?

There is nothing fun about programming anymore. The joy is completely gone. I would switch careers if I could make money elsewhere.

Reply on news.ycombinator.com