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

AI is killing some companies, yet others are thriving – let's look at the data

elenaverna.com

Post reply on HN
Print view View original

101–110 of 318 posts

Re: AI is killing some companies, yet others are thriving – let's look at the data

#101

Post by nitwit005 »

inetknght wrote: post #77
nitwit005 wrote: post #69

Earlier quoted context omitted.

I'm sure I can make a couple of surprising or insightful articles for my current industry, and then I'll run dry. Most topics require data or information in some form, which requires time to accumulate. You end up rate limited. Even at the scale of a decent sized company, you often can only produce interesting content occasionally.

> Most topics require data or information in some form, which requires time to accumulate. You end up rate limited. Even at the scale of a decent sized company, you often can only produce interesting content occasionally. That sounds like a good thing tbqh

Sure, but it implies writing "quality content" won't work as a replacement.

Re: AI is killing some companies, yet others are thriving – let's look at the data

#102

Post by azemetre »

shagie wrote: post #98
azemetre wrote: post #81

Earlier quoted context omitted.

I think the issue is when people start replacing their capacity to think and reason with these machines. There was an intergalactic space jihad about this IIRC.

I also recommend the written story for Zima Blue (the Love Death and Robots version was much more focused on the artistic aspect) and the question of wine.

What are you referring to by "the question of wine?" Searching online didn't help, unless you meant Dandelion Wine by Ray Bradbury?

Re: AI is killing some companies, yet others are thriving – let's look at the data

#104

Post by firstplacelast »

_benj wrote: post #26

I’ve seen this shift in my own usage. I find myself appending “Reddit” to the end of my searches a lot more often, I have pinned Wikipedia to the top of my search results (in kagi) and I haven’t visited stackoverflow in months, although I see that perplexity quotes it quite often when I ask it coding questions… I’m just a sample of one, but it’s certainly interesting to see how apparently I’m just one of many

I've been working on learning PySpark this week, going through exercises to learn syntax and what-not. Avoiding ChatGPT bc I didn't want to be spoon fed anything. I couldn't figure out how to sort something in descending order. I searched SO for 5 minutes, getting out-dated answers or overly complicated solutions not needed for my very simple one-liner. Click over to ChatGPT, give it the line of code I have and say 'make it sort in descending', and spits out exactly what I need.

I haven't used SO in weeks prior and I think ChatGPT has pretty much killed its use case for me. Even two years ago, I would spend 1-6 hours/week searching SO posts depending on what I was working on.

Re: AI is killing some companies, yet others are thriving – let's look at the data

#105

Post by KapKap66 »

cosmic_cheese wrote: post #85
charlieyu1 wrote: post #72

Earlier quoted context omitted.

At least some of them are written by humans and some of them are good. Maybe good 20-50% of the time, still better than 100% SEO crap.

Yep. As poor as the signal to noise ratio is, for many things it’s substantially better there than elsewhere. Though as of late, that’s been eroded too. Increasingly the most useful answers are in older threads more than newer ones, an effect I’d at least partially attribute to the APIpocolpyse a while back that drove away some of the site’s best and most prolific contributors. It’s becoming filled with the same mind…

I do some game development research on reddit sometimes, and with fairly high frequency I find the highest quality results are something like 10+ years old. Occasionally I find something insightful in more recent threads but it's often heavily downvoted or hidden; it almost seems like difficult-or-controversial-but-content-heavy-meaningful answers automatically get thrashed by readers.

Just commiserating, I guess...

Re: AI is killing some companies, yet others are thriving – let's look at the data

#106

Post by shagie »

aleph_minus_one wrote: post #99
shagie wrote: post #88

Earlier quoted context omitted.

Q&A tends to be "chunky" and asynchronous in its communication model. This comes from a reaction to the previous model of forums where it was smaller bits of data spread across multiple comments or posts. I recall going through forums in the days before Stack Overflow, trying to find out how to solve a problem. https://xkcd.com/979/ was very real. Stack Overflow (and its siblings) was an attempt to change this to a "…

> And yet, users often don't know what they want. They want that forum model with interaction and step by step hand holding by someone who knows the answer. This sounds like the seed for a business model to pitch in the next upcoming hype cycle: "short-term hiring of experts with quarter-hourly billing increments enabled by our web- or app-based user interface". :-)

Maybe... the difficulty with how AI is advancing, it might be difficult to distinguish it from an LLM. You might even get to the callcenter problem of "initial support handled by someone following a script" before you can get passed to someone who knows how to solve the problem.

If I want someone to walk me through making muffins from scratch, is a human on the other and of that line (competing with $1/day rates for ChatGPT Pro - its cheaper than that, but that's the comparison) and are they better than what ChatGPT can do?

It would have to be... quite a bit more than what the LLM would be priced at. The minimum it could reasonably be (without any other things) would get close to $4/15m... and that's minimum wage.

I really don't think that humans are competitive on that timescale or rate.

It would probably be better to hire people at some higher rate to write content for your private model. Brandon Sanderson is considered one of the faster writers (in the fantasy genre) and averages at about 2500 words / day ( https://famouswritingroutines.com/collections/daily-word-cou... ) - and while he makes a lot more than most authors, lets go to a more typical $75,000 USD / year. 250 working days per year and we're at $300 / day. And we're to $0.12 per word. ... Which puts a person in the intermediate to experienced price per word range https://uxwritinghub.com/writers-salary/

Not that I'm suggesting that's the way to do it, but something for LLMs to consider - hire experts to write content for their LLM. $125 per 1000 word blog post.

298 words. I'd like my $37.25 please. Not that I'm asking you for that, but rather that's what my words as training material would be worth.

Re: AI is killing some companies, yet others are thriving – let's look at the data

#107

Post by nitwit005 »

mitthrowaway2 wrote: post #45
vecter wrote: post #36

I've noticed that since I've started using ChatGPT, I've almost entirely stopped using Google (except for the rare case where I need a specific website but don't remember the URL). In addition to a bunch of technical questions related to my work, my ChatGPT chat log has the most mundane things like: - What is platos frios - Can you download Netflix videos to your local device - Who composed the Top Gun theme - Who ha…

After some experience and testing, I've become well aware not to use LLMs to ask questions like "who did X" and "what is company Y's policy about Z", because they tend to hallucinate responses (even for well-known people). What I've not yet figured out how to deal with is how to handle being surrounded by a society of people who go ahead and trust LLMs for their factual answers anyway. I think even if I'm careful abo…

I suspect this is more of a case of garbage-in garbage-out. The existing web results have invented answers.

People created websites to "answer" people's search queries about celebrity net worth, if some celebrity is gay, if they are in a relationship, etc. They obviously frequently did not know, and made a guess, or relied on tabloids as a source, who also frequently make things up.

Re: AI is killing some companies, yet others are thriving – let's look at the data

#108

Post by shagie »

azemetre wrote: post #102
shagie wrote: post #98

Earlier quoted context omitted.

I also recommend the written story for Zima Blue (the Love Death and Robots version was much more focused on the artistic aspect) and the question of wine.

What are you referring to by "the question of wine?" Searching online didn't help, unless you meant Dandelion Wine by Ray Bradbury?

In the story, wine is one of the topics of conversation between Carrie and Zima. The first mention of it - and the question - "Red or white?"

    Zima indicated that I should take one of the seats. His hand dithered over two bottles of wine.

    ‘Red or white, Carrie?’

    I opened my mouth as if to answer him, but nothing came. Normally, in that instant between the question and the response, the AM would have silently directed my choice to one of the two options. Not having the AM’s prompt felt like a mental stall in my thoughts.

    ‘Red, I think,’ Zima said. ‘Unless you have strong objections.’

    ‘It’s not that I can’t decide these things for myself,’ I said.

    Zima poured me a glass of red, then held it up to the sky to inspect its clarity. ‘Of course not,’ he said.
The use of wine and the decision is one that goes through the entire story and plays a role at the conclusion.

The "AM" is the Aide Memoire. It is a small robot resembling a hummingbird in size that accompanies people and is an assistant for remembering things for humans who have a lifespan of hundreds of years.

I strongly recommend the story - it is so different from the animated version.

Re: AI is killing some companies, yet others are thriving – let's look at the data

#109

Post by lz400 »

mvdtnz wrote: post #48
_benj wrote: post #26

I’ve seen this shift in my own usage. I find myself appending “Reddit” to the end of my searches a lot more often, I have pinned Wikipedia to the top of my search results (in kagi) and I haven’t visited stackoverflow in months, although I see that perplexity quotes it quite often when I ask it coding questions… I’m just a sample of one, but it’s certainly interesting to see how apparently I’m just one of many

It's shocking to me that people would intentionally seek Reddit threads. The quality of the discussion on that site is absolutely appalling beyond belief.

That's not true for the kind of searches we're talking about here. If you are looking for "best mechanical keyboard" or "reviews of shimano bicycle gearsets", reddit will regularly be of higher quality than the median google first page.

Re: AI is killing some companies, yet others are thriving – let's look at the data

#110

Post by diamond559 »

Legend2440 wrote: post #70
RandallBrown wrote: post #40

Earlier quoted context omitted.

Isn't that because ChatGPT is trained on those QA platforms? If real humans aren't answering questions on the Internet, how will LLMs learn the answers to those questions?

Not entirely. LLMs are smart enough to answer questions from documentation or from other sources, even if the exact question wasn't asked anywhere. But Q&A websites do contain information that might not be in other sources, so there would be some loss.

That's just copying from better sources, doesn't mean the ai is "smarter".
Post reply on HN