[joenot443]

Sometimes I wonder if we as internet people are special in that it seems we can pretty immediately spot vanilla-LLM generated text. I'm sure others have noticed the same, but maybe 20% of my inbound recruiter pings on LinkedIn are straight from GPT and subsequently are immediately dismissed.

I posted a job on UpWork last month and probably 60% of the replies were LLM, some of them seemingly automated as they came in moments after I posted. Obviously these were immediately rejected.

Is this something we'll have to get used to?

I got my first internship in 2015 by cold-emailing the CEO of a Waterloo startup that I thought was really cool. At the time, reaching out directly over email with a thoughtful and earnest introduction was a good way to set yourself apart. I'm not sure that's the case anymore.

[kjs3]

Is this something we'll have to get used to?

On one hand, I have potential candidates that I mentor who gleefully tell me how smart they are that they have LLM-backed bots replying to every remotely related job post on every platform. The idea that this approach might not be appreciated is generally met with either profound disbelief or dismissal as "luddite attitudes" or some such. "AI is the new hotness, how can it be bad that we're using it!"

On the other hand, I talk to HR folks, who are just as pleased with themselves that they've deployed ML-driven candidate filtering apps so only candidates that are 'perfect' matches for my job description are even seen by a person. And they do not appreciate it when I point out that must be why the only resumes I've seen from them are buzzword-bingo lists of 'qualifications' obviously included to game the filters, as most subsequent interviews make clear.

So...it seems to be an arms race that we all loose. Get used to it.

[tonyedgecombe]

>Is this something we'll have to get used to?

The problem for UpWork and their customers is what happens when 99% of the replies are LLM generated. The same when everything Google links to is LLM generated or even worse when their LLM summary is generated from LLM content they have been slurping up. The whole sad edifice may well collapse in on itself.

[prescriptivist]

I was listening to a podcast the other day and one of the commercials I think was from Facebook (IIRC, it could have been Google...). It was directed toward students, so I'm pretty sure it was an organic ad, but there were several aspects to it that felt fairly personalized to me (bicycling and outdoor related) that I actually wondered if it was generated on the fly using an LLM.

FB knows me and what I like and they have enough of data on my searches that they could customize a pretty relevant audio ad that, now with LLMs, can feel really relevant and natural, especially with audio gen being so good.

My point though is I wasn't sure if it was LLM generated or not and that's stuck with me. Random ChatGPT copy-pasta is easy for me to pick out – most people do not write that well. But a sophisticated application of this tech probably approaches the it-could-go-either-way territory.

[ToucanLoucan]

I genuinely loathe the over-familiar vibe you get from targeted ads in such a way where no matter how relevant to me the given product is, if you advertise it like that, I'm immediately suspicious of your intentions. If I'm talking about longboards a lot around my house and get ads for longboards, I would go so far as to say I would go out of my way to not buy any of the brands shown to me.

I think this is an underrated metric in terms of how advertisers are organizing their spend; the creep factor. If your first impression with a potential customer is creeping them out, how are the odds of them giving you a sale?

Just my 0.02.

[LaundroMat]

I have a similar annoyance: the obvious ways marketeers try to get me in their funnel (I call it conversion aversion). It's hard to measure, so it's not taken into account by marketeers, but I'm quite certain the amount of people annoyed by conversion optimization tactics represent a lot of potentially lost sales.