ChatGPT's explosive growth shows first decline in traffic since launch | Hacker News Reader

ChatGPT's explosive growth shows first decline in traffic since launch

reuters.com

27 pointsmfiguiere13 comments

I was very nervous about ChatGPT and disruption of both software development and general content creation a few months ago. I paid for GPT4 and used it. I'm no longer that nervous. Having something be wrong 10% of the time is extremely tedious. It's got its uses but I think people are going to choose quality over quantity in a lot more cases than we first thought.

Just 10%? Most of the time I’ve tried to use for anything hard it’s much worse than that

I think it is wrong 75% of the time but it’s easy to get it to generate another answer over and over again until it gets it right.

It's pretty clear to me the reason why the traffic went down as ChatGPT is used by many students.

In the Gartner hype cycle, you are now at the ‘Peak of Inflated Expectations’ of this LLM hype.

It is quite clear that the AI bros and grifters have squeezed the LLM hype to the point where any serious use-cases outside of summarization have just been untrustworthy to the users and relies on triple checking the output.

Summarization has proven to be the most reliable use case and everything else is useful as a bullshit machine packaged as a snake-oil product.

I very rarely use google anymore, as it's a worse bullshit machine. I just ask chatgpt and tell it to give me sources. I also never use cookbooks anymore, as it's much more interesting to have a chat about what I can make with the ingredients in my fridge. It's also absolutely fantastic at making meal plans (tell it restrictions like lactose, protein you want per day, calories per day, and any other preferences you have) and workout plans (tell it the equipment you have in your house and your goals). If you're wanting it to code the next big thing by itself it probably won't do it properly, but for everyday things it's a very useful assistant. More useful than anything else out there by far.

Try using Perplexity.AI [no affiliation], which automatically provides a citation for every sentence it produces. Typical paragraph will have five citations.

NO LOGIN. NO SIGNUP. NO EMAIL.

Is it good though?

As with many of these organization-controlled LLMs: it used to be a whole-lot-more capable, pre- "as a large language model, I cannot..."

I am sorry you get downvoted for saying things as they are. We'll see ourselves in 2028 and taking bets now - the hype of LLMs would be all but gone.

LLMs or generative AI in general are NOT a product a business can rely on. Even it is not good enough to come up with a good logo for your brand. Really.

Unless you lower your standards to the rock bottom and accept whatever permutation of pixels as your brand identity.

An AI Winter is not a remote possibility. A breakthrough however seems very remote.

It's 2023 and by each passing day, models are getting outdated.

How do they know how much gets outdated every day? And what kind of and what exact training data as a "patch" would be required to keep the "accuracy" on the higher side.

Seems like no one knows. These for now are expensive computational toys.

I'm seeing decline in ChatGPT virtue signaling in my Linkedin feed too.

ChatGPT usage is not a virtue...?

Reply on news.ycombinator.com