ChatGPT Traffic Slips Again for Third Month in a Row (reuters.com) 49
OpenAI's ChatGPT, the wildly popular artificial intelligence chatbot launched in November, saw monthly website visits decline for the third month in a row in August, though there are signs the decline is coming to an end, according to analytics firm Similarweb. Reuters: Worldwide desktop and mobile website visits to the ChatGPT website decreased by 3.2% to 1.43 billion in August, following approximately 10% drops from each of the previous two months. The amount of time visitors spent on the website has also been declining monthly since March, from an average of 8.7 minutes on site to 7 minutes on site in August. But August worldwide unique visitors ticked up to 180.5 million users from 180 million.
School coming back into session in September may help ChatGPT's traffic and usage, and some schools have begun to embrace it. U.S. ChatGPT traffic in August rose slightly, in concert with American schools being back in session. "Students seeking homework help appears to be part of the story: the percentage of younger users of the website dropped over the summer and is now starting to bounce back," said David F. Carr of Similarweb, who regularly tracks ChatGPT and its competitors.
School coming back into session in September may help ChatGPT's traffic and usage, and some schools have begun to embrace it. U.S. ChatGPT traffic in August rose slightly, in concert with American schools being back in session. "Students seeking homework help appears to be part of the story: the percentage of younger users of the website dropped over the summer and is now starting to bounce back," said David F. Carr of Similarweb, who regularly tracks ChatGPT and its competitors.
Wildly popular for what exactly? (Score:2, Interesting)
Wildly popular for making ticktock and youtube videos? Wildly popular for its mistakes making it into the news? Wildly popular as an excuse for eliminating jobs? Wildly popular for CEOs lie about?
The traffic is slipping because being popular isn't the same as being useful. Traffic is equalizing with its actual usefulness. It will continue to be a slow decline while stakeholders fight to keep their investments from failing.
One day it could be more. Today, it's just search results summarized by an authoritati
Re: Wildly popular for what exactly? (Score:2)
I donâ(TM)t understand why we arenâ(TM)t all on the same page, the publicly accessible ChatGPT is just publicity promoting what they are selling eventually. Who cares if traffic is down, or if you get weird answers - this was released just to get us to talk about it.
Reality is that an actual usable LLM is trained on specific data, and is inaccurate far less often than the widely trained ChatGPT on their website.
When a business uses LLM, there are guardrails all over the place, and verification to
Re:Wildly popular for what exactly? (Score:4, Interesting)
Wildly popular for making ticktock and youtube videos? Wildly popular for its mistakes making it into the news? Wildly popular as an excuse for eliminating jobs? Wildly popular for CEOs lie about?
The traffic is slipping because being popular isn't the same as being useful. Traffic is equalizing with its actual usefulness. It will continue to be a slow decline while stakeholders fight to keep their investments from failing.
One day it could be more. Today, it's just search results summarized by an authoritative sounding LLM, often incorrectly.
It's useful when used appropriately.
With coding it can be a big time saver in fleshing out ideas or figuring out APIs, you just need to double check the work carefully.
For computer issues it can be fairly useful in diagnosing and fixing issues. It's far from 100% but neither are the various forum fixes either.
For search, it's not great as a general search engine. But when I'm dealing with a particularly unfamiliar topic where I don't even know what search terms to use it can give me a good answer or at least give me enough grounding to search the rest.
Having it write you an article or a paper of some kind? It's a neat parlour trick but not really helpful.
Re: (Score:2)
It's useful when used appropriately.
Agree completely.
It’s super-useful to me, like being able to interview the google or maybe an [somewhat leaky] expert in the field.
Examples?
Re: (Score:3)
Except it's exhausting to read the generated output and think every second, could this be a total hallucination? How about this? Or this? Especially knowing that hallucinations appear when you press it for details.
Re: (Score:2)
Re: (Score:3)
(...) when I'm dealing with a particularly unfamiliar topic where I don't even know what search terms to use it can give me a good answer or at least give me enough grounding to search the rest.
That's my main use case. I can describe something I'm vaguely aware of but not enough to know the technical jargon, and after back and forth it provides me enough technical terms I can then easily find more (and more accurate) information by searching for those. Here's an example of a recent prompt of mine:
Me: What the technical, most common name for the managerial style in which the manager gives a list of tasks with deadlines to the employee, keeps constantly asking the employee when he'll deliver the res
Anybody surprised? (Score:2, Insightful)
Anybody surprised? IMHO, people are simply realizing what and over-hyped piece of crap chatGPT is. Well, there are situation where chatGPT can be used maybe but it's definitely over-hyped and over-sold...
Re: (Score:3)
Re: (Score:2)
Well, it is an LLM. It can only do things reliably it has seen lots of correct examples for. Outside of that, it will hallucinate and there is no way to fix that. In fact, it has become obvious (and it was not unexpected in the first place) that if you try to make an LLM better in one aspect, it will get worse in others.
The only use this tech really has is for very limited specialized problem domains. It is not a general-purpose tool at all, unless you are just asking about the very basics of a mainstream t
Have they fixed any issues? (Score:1, Interesting)
Re: (Score:3)
novelty prompts have started to wear thin
They made some massive PR claims out of the gate, and many people were interested and wanted play with it for a while to find out for themselves. Well, now they've mucked about with it and had some soak time -- looks like it isn't as ground breaking as the PR department was making it out to be after all.
It can write short stories based on given prompts, and it will lie whenever the black box statistical algorithm doesn't happen across the real answer. It's a neat toy, and a useful tool to get past some mino
Re: (Score:2)
That nicely sums it up. But actually, an LLM does not "lie" when it hallucinates. It simply does not know any better and it has no way of knowing how well it knows a specific thing. So hallucinations cannot actually be eliminated from an LLM.
Re: (Score:2)
an LLM does not "lie" when it hallucinates. It simply does not know any better and it has no way of knowing how well it knows a specific thing. So hallucinations cannot actually be eliminated from an LLM.
Technically, yes. However, the consumer experience is what I was targeting here, and I feel that from the consumer perspective any answer given that is untrue is effectively a lie. Ie, it doesn't really matter how the algorithm came up with the wrong answer, it only matters that the answer itself is wrong.
Re: (Score:2)
Ah, yes. From the consumer perspective, calling it "lying" does make sense, I agree to that. It does, after all, tell made-up crap using language that assures a high level of confidence.
From the tech perspective, it is different. If it were "lying", it would know it is lying and could stop lying, i.e. the problem would be fixable. Unfortunately (or fortunately?) this problem cannot be fixed for LLMs and we do not have generalized fact-checkers that could be used to filter LLM output. My take is that using r
Re: (Score:2)
An LLM "hallucinating" in any question a tiny bit more advanced is something that cannot be fixed. This behavior is inherent to the model type.
Fix the website (Score:3)
The hosting configuration is fuxxored.
Re: (Score:3)
pretty sure just someone reversed the modem polarization.
People Discover (Score:2)
the man behind the curtain (Score:2)
Re: (Score:2)
Indeed. This thing can do a lot of very, very basic things with impressive results. But knowledge workers can only be replaced by something that actually has an understanding of things and can do fact-checking. LLMs cannot do either.
Recent example: I saw somebody demonstrating an LLM to make a simplistic teaching course. Yep, nice, but I need a reasonably sophisticated teaching course or the students will be pissed and rightfully so. But that LLM cannot actually go above the simplistic version, it is litera
It's Just A Set Of Paints. (Score:4, Insightful)
It needs an artist to bloom. Give it time. Also, people don't really want to 'rent' it. They want to run their own instance and that requires, from what I can tell, about $60,000 to just get a basic setup. Again, give it time.
Teachers and students (Score:2)
School coming back into session ...
Teachers have been found to use ChatGPT for their lesson-plans. Since teachers have domain knowledge, hopefully this just filling-in a template and not writing the whole lesson because the AI-generated books found on Amazon demonstrate that current AI is frequently and dangerously, inaccurate.
Not interested (Score:2, Insightful)
I tried it once. I had no interest in returning.
Re: Not interested (Score:4, Informative)
Internet 'may be just a passing fad as millions give up on it', Daily Mail, Tuesday December 5. 2000.
Can't think of anything happening in the last three months... the months of summer... that could cause usage to drop. Nope, nothing.
Re: (Score:2)
The "AI" fad is nearly over. It will be as relevant as 3D TV by Christmas.
Competition is heating up (Score:3)
For a while, ChatGPT was the only place to go. Now, Bing Chat offers a very similar experience, built on the same technology, and people are using it. Google Bard isn't as capable, but "good enough" for a lot of purposes, and easier to access. Some use cases, like programming help, are now being packaged as integrated IDE tools, such as GitHub Copilot.
Whatever you think of each of these options, the fact is, each will be used by some portion of people who would have gone to ChatGPT before. As new competitors enter the market, they'll all have to work harder to keep people on *their* platform.
Re: (Score:3)
ChatGPT has fallen badly behind.
The model is trained on data from 2-3 years ago and Bard, Claude etc doesn't have those problems.
Re: (Score:1)
That's ... certainly an idea. Far more likely, however, is that we've simply moved past the peak of inflated expectations and are sliding into the trough of disillusion as reality displaces fantasy. Fewer people are using these things because they've realized that they're mostly useless.
This was entirely predictable. As I've explained many times before, these models are simply not capable of doing the things people believed that they could do. If the comments here are any indication, a good number of pe
Re: (Score:2)
You and I are apparently having a very different experience with ChatGPT.
Personally, it's helped me with a wide variety of things, including:
- An idea for safely removing a very large mirror from a wall
- Instructions for performing a complex operation on a HOG stage lighting console (not listed in the manual)
- Diagnosis of an obscure error message related to C# code
- Construction of complex XPATH queries in XSLT
- Code sample to implement authentication with Amazon Cognito in C# (not easy to find on Amazon's
Re: (Score:2)
You and I are apparently having a very different experience with ChatGPT.
Personal experience has nothing to do with it. When I say that the technology can't do something, that's not an anecdotal generalization. It's a simple statement of fact. There's a very big difference between what laypeople think is happen and what is actually happening.
I have noticed that people tend to overstate the utility and often blaming themselves for low-quality output. "It's not the fault of the model that you suck at writing prompts" is something you often hear from the apologists. I've noticed
Re: (Score:2)
Personal experience has nothing to do with it.
I have yet to see ChatGPT save anyone time.
*You* have yet to see? I suppose personal experience should only be discounted if it's someone else's personal experience?
My personal experience is relevant in the sense that I frequently ask ChatGPT for answers, on many different topics, and get good answers. Often, I'll then repeat the question as a Google search, just to compare how long it would actually take to get the same results the "old" way. Many times, it takes a lot more clicks and reading various links, to find what I want, as compared to ChatG
Re: (Score:2)
Sigh... That's what I get for replying to a known troll.
I don't need personal experience to know what models like this can and can not do. I personally haven't seen what they can do save anyone any time. This is because the these things can not do the things they're being asked to do.
This isn't complicated.
Re: (Score:2)
At least I'm known for something!
Re: (Score:2)
I use ChatGPT on a daily basis - about half for work, half for personal queries where I'm trying to quickly scratch a curiosity itch. It's significantly better than "objectively awful" at a whole range of prompts I throw at it. I only came to ChatGPT a couple of months ago once most of the hype had died down, and came in deeply skeptical - it's a productivity tool, nothing more.
Could I wish it were better? Of course, but then it really would start taking over human jobs, which isn't on the horizon for a
Re: (Score:1)
it's a productivity tool
It's not a productivity tool. It's a silly toy. Pretending that it's a tool is stupid and dangerous, as many unfortunate people have discovered.
I'm perfectly happy to have the debate with you as well
What's to debate? There's fantasy and reality. Like the willing victim of a strip mall psychic, you want to believe so you highlight the 'hits' and ignore the 'misses'. It's really that simple.
that's one example amongst hundreds.
I have a massive collection of ChatGPT absurdities as well. If you think trading stupid anecdotes is a "debate" or meaningful in any way, well, there's really nothing I
Re: (Score:2)
Yep, pretty much the response I expected.
Dismissive, arrogant, yet at the same time completely uninformed and unwilling to learn any better. Dunning Kruger in full swing.
Re: (Score:1)
unwilling to learn any better. Dunning Kruger in full swing
I was thinking the exact same thing.
One more actual fact for you: Your ignorance is not as good as my actual knowledge.
Pathetic.
Re: (Score:1)
Despite your big talk, you've yet to demonstrate any actual knowledge or experience with anything in this space, as opposed to the people who are actually using it productively every day and have a fundamental understanding of how the technology works, not just a theoretical one they've read on Facebook or Twitter.
The rest of us will continue to enjoy the productivity gains while you wallow in your bitterness, screaming out "but it's a toy! listen to me! I know what I'm talking about guys, honest!".
What a s
Re: (Score:2)
LOL! Looks like you haven't read any of my posts! I'm not even a little bit surprised. You're in the habit of talking about things way outside your depth.
Not only do I have the benefit of an extensive formal education in this particular area, I'm also published. I'm more than qualified to speak authoritatively on the subject.
You, in contrast, have played with a silly chatbot for a few weeks. What were you saying about Dunning-Kruger again?
It's beyond pathetic. I have no patience for the willfully igno
Re: (Score:1)
Keep screaming into the void dude. Your opinion doesn't affect my productivity. Simple fact.
Re: (Score:2)
"Productivity"? LOL! That's really funny coming from the guy who spends his time "debating" things he knows nothing about online.
Why not put some of that energy towards actually learning something? If nothing else, you'll have less time to smear your uninformed nonsense all over the internet. Think of it as a public service.
Re: (Score:1)
Haha yes, the productivity improvements I realize from ChatGPT afford me the spare time to do this sort of thing. Game, set, and match :)
Re: (Score:2)
Wow, could you get any more pathetic? Here in reality, you just have nothing important to do.
The arrogance of uneducated trash never ceases to astonish.
Re: (Score:1)
You really do seem to be getting upset at being contradicted. I guess there's no point in continuing the conversation, you are completely unwilling to accept anything that violates your preconceived notions, despite your apparent lack of (practical, rather than theoretical) experience on the subject matter, or to draw any value from the direct experience of others who are currently having real-world success with making these things work for them.
I guess we'll see in a few years time whose perspective has a
Re: (Score:2)
What's the next hype? (Score:3)
Just asking for a friend in HR who tried hiring ChatGPT experts with +15 years experience willing to work for a box of bananas.
Re: (Score:2)
Rookie error. Tell your friend to try offering peanuts instead of bananas.