August marked the third month in a row that the number of monthly visits to ChatGPT's website worldwide was down, per data from Similarweb.
ChatGPT is losing some of its hype, as traffic falls for the third month in a row::August marked the third month in a row that the number of monthly visits to ChatGPT's website worldwide was down, per data from Similarweb.
It took a while but yeah that seems about right. It takes a lot of guiding to have it produce something usable. I have to know a lot about what I want it to do. It can teach me things but the hallucinations are strong sometimes so you have to be careful.
Still it helps me out and I make a lot of progress because of it.
I like it for certain techy things. I just used it to create a linux one-liner command for counting the unique occurances of a regex pattern. I often forget specific flags for Linux commands like how uniq can perform counting.
And something like that is easy to test each piece of what it said and go from there.
As long as you treat it like a peer who prefaced the statement with "I might be wrong / if I recall correctly" it ends up being a pretty good aid.
"I can suggest an equation that has been a while to get the money to buy a new one for you to be a part of the wave of the day I will be there for you"
There, my phone keyboard "hallucinated" this by suggesting the next word.
I understand that anthropomorphising is fun, but it gives the statistical engines more hype than they deserve.
That very much depends on if you believed all the hype or not. If you did, then yes, it's failing, as ChatGPT was supposed to be the next big breakthrough that was going to automate everything ever, and any company that didn't get in on that right now was going to be left in the dust by all their competitors. On the other hand, if you were an actual sane person (so you know, not a CTO/CEO), then this is very much a non-story as you always knew that all those outlandish claims were nonsense and that this was always going to be yet another niche piece of tech that's useful in a few places in limited amounts.
I mean, yeah? You can't rely on hype ever being present. Honestly with the very light use I do (Asking a couple of questions at most in a month) I feel like it has not changed at all. Just at the top I now have a locked button that says "GPT 4".
FWIW, it’s become an inextricable part of my life. I use it for hours every day, for programming and Linux advice, spreadsheet help, foreign language practice, and random trivia.
Yesterday, I discovered that Snoop Dogg’s -izz speak from the aughts was actually derived from carny pig Latin.
I'm starting to think I'm incompatible with GPT. Code that's laughably wrong (like sticking in things that aren't even in the language), DM advice that I could get walking down a greeting card isle, and explanations that would get a Wikipedia editor sent to the firing squad.
Nah, that sounds about right. This is just the natural result of people actually trying to use GPT for all the things they were told it would be able to do, and now discovering that was in fact all bullshit. The LLMs are and always have been massively overhyped and oversold on what they can do. Sadly this won't stop the corporate executives from trying to use them to replace workers, although when that effort eventually face plants they'll just quietly re-hire a bunch of people and find some middle manager to blame for their failure. This is was and always will be merely a productivity tool to automate some repetitive work, but it still needs someone to review and clean up its output. It's not "replace someone doing 40 hours of work a week", it's "allow someone to do what used to take 40 hours in 35 hours instead".
Sadly the most impact this is going to have is on spammers and scammers, who can now automate generating their garbage since it never mattered that any of that crap was accurate or not, merely that on a casual glance it looks reasonable.
Yeah. There are a lot of shitty marketing ideas that suddenly become profitable if you don't have to pay people to generate the content it needs. Honestly I've had a couple of those ideas over the years and I'm glad I'm no longer in a position to propose them to anyone.
When I first started messing with it, I was kind of neat and fun. I like making characters so I was using it for like story prompts and general outlines. Some were better than others, but it was neat for some inspiration and fleshing out. I never took it's outputs 1:1.
But when I messed with it again recently. It was a lot worse. Like it ignored parts of my prompt. Like as an example a prompt was about a romance story, but the story was about character A and their family. The love interest character was barely a footnote and could have been removed entirely and nothing would have changed with the story outlines it was giving me.
I thought maybe it doesnt like romance prompts, so I tried less specific and more broad prompts from there, and it was the same thing of just... not outputting what I was asking it to. It got worse and worse and sometimes wouldn't output anything at all.
ChatGPT has gotten dumb. I used to have to code check it's answer every few responses. Now it's every response. It wrote me an if/else statement the other day where if and else had the same outcome.
Source: Work in AI, sometimes on LLM's, mostly on the software engineering side rather than the science side.
I have a few theories on why ChatGPT was so successful, and why the hype is starting to crumble, but they all largely centre around a well-known problem that LLM's have had for years - they hallucinate, a lot.
When your product becomes popular, you deal with unique problems that don't seem to scale without insane amounts of money, or a literal army of people to plug gaps where your model is saying things it shouldn't - whether it's accusing high-level politician's of crimes they didn't commit, or telling people how to make chemical weapons in the style of your grandma. It's an expensive loop in compositional models, so I'd hate to know how much work it took ChatGPT to get to it's "best" version.
Over time, valuable data disappears, and your data naturally skews over time due to it being incorrect, invalid, or pushed into just being biased towards a given inaccuracy. Sometimes, you do everything right and you train on manual input that you've vetted as correct through expert analysis or user feedback - and it's still wrong. IMO, ChatGPT was always going to struggle to keep the hype, and it will eventually be seen as what it has always been: a concept that shows the utility of LLM's as a commercial product.
Make no mistake, the likes of Google, Amazon, Apple, and Meta will probably plug the gap, and will reach either parity with GPT4 or improve on it. However, the fundamental problem of hallucinations will not disappear, and we'll continue to see neutered experiences that make for great tools, but burn cash to provide these tools with minimal possibility of offending people/damaging the brand.
The main thing I hope to see from the rise and fall from ChatGPT is a rise in productivity tooling, but also people to finally see those that hype these technologies as what they are - grifters.
This is a big part for me. When ChatGPT first came on the scene, I was absolutely blown away by its natural language parsing capabilities, but it wasn't long before I started to hit the boundaries of its abilities. I was disappointed by how unreliable it was with anything but the most simple queries. Now it just doesn't do enough to really bother with.
Hasn't the service also gotten worse? When it first came out, you kept hearing how it could pass the bar exam and medical license test. And now all you hear is that it can't do basic high school homework without wrong answers. Maybe it was hype in the beginning and it never could do those things.
It was always able to do some genuinely amazing things, but it was always limited when you took it beyond its wheelhouse. It helps to think of it not as an "AI" as people keep saying, but as a "text completer" with a huge amount of power within that domain.
Or, another way to think of it is as a super-powerful search engine. If the answers and knowledge you're asking of it were fed into it as input data at some point in its training, it'll probably be able to find it and reformat it back to you with a scary amount of smoothness and precision. If you're asking it to figure out something new, it may be able to fake it in some short-term fashion or another based on what it's seen, but not with any genuine understanding behind it. That's just not what it does. I actually have a little private theory that if it was given something like the bar exam in scale and complexity, but an exam was genuinely a whole new novel invention that hadn't been extensively discussed and represented in its input corpus, it would fail pretty badly. A lot of what humans can do that makes them capable is adapt to new domains -- we can teach ourselves to play chess, or do math, or fly airplanes, or play Celeste. GPT is hugely impressive but it's still only one domain.
I actually don't believe that it's gotten substantively less capable. I think there are little ticks up and down in its capability sometimes in particular areas, and people seize on those to conclude that it's now becoming dumber, but in my experience, the raw API was always quite capable (more so than the somewhat nerfed chat interface), and it was always super-capable with some tasks and not at all capable with others. I think journalists are just now figuring out that, after having studied the issue in their professional capacity for the better part of a year, and reporting on it as if it's a new thing.
From what I've seen, here's what happened. GPT 4 came out, and it can pass the bar exam and medical boards. Then more recently some studies came out. Some of them from before GPT 4 was released that just finally got out or picked up by the press, others that were poorly done or used GPT 3 (probably because of gpt 4 being expensive) and the press doesn't pick up on the difference. Gpt 4 is really good and has lots of uses. Gpt 3 has many uses as well but is definitely way more prone to hallucinating.
I made the mistake of asking chatgpt questions about securing my network setup. It confidently gave me a huge amount of misinformation that led to 8-10 hours of frustration and pointless troubleshooting.
This is just balancing out. Anything that gets over-hyped will eventually drop in use. It'll eventually be a boring yet useful tool just like spreadsheets, spellcheck, or email.
They definitely nerfed it. We will probably end up in a situation where corporations and the rich have access to god-tier AI, and everyone else has access to mediocre, ad-supported AI.
No one seems to have thought about the fact that most schools have been out for those three months. Not sure exactly how much of the traffic is high schoolers and college students cheating, but that could account for at least some of the loss in traffic.
I switched from their $20 interface to straight API, it makes more sense for me. I wonder if other people are doing this and these metrics aren't accounting for that.
ChatGPT took the world by storm when it was released last November, but it looks like it's losing momentum.
"One theory about why ChatGPT's web traffic dropped over the summer is that school was out, which would help explain why the traffic trend stabilized in August as schoolchildren in the US were back in class in greater numbers toward the end of the month," David F. Carr, a senior insights manager at Similarweb, wrote in the report.
Before Meta's Threads assumed the title in July, ChatGPT was the fastest-growing app ever when it reached 100 million users in two months.
Some of that hype was prompted by students, leading to professors finding ways to combat ChatGPT plagiarism, and one Princeton student launching GPTZero to detect if an essay was written by AI.
But it's also being used in the workplace, with employees using ChatGPT to write code, do research, and improve time management.
In July, users of OpenAI's latest model, GPT-4, started complaining that the chatbot's performance had declined.
The original article contains 338 words, the summary contains 169 words. Saved 50%. I'm a bot and I'm open source!
Yeah but they keep removing stuff and locking down how useful it can be. First they took GPT4 and put it behind a paywall, now I have a limit to how much I can use it per day and have to switch between multiple accounts sometimes. Makes it a lot harder to work it into new projects knowing I might have to wait on GPT to get its shit together every other day
Aw yeah dude chatgpt sucks for sure all of you should stop using it immediately so all the rich people who wanna fuck with it and discover ways to make money can and you can continue to be a little bitch to the system.
Definitely don’t use the api and learn some Python so you can control all the settings including system level prompting and so on. Definitely not a fucking blast to play with and I’ve definitely been so annoyed and have hated it for months and months. 👌🏻