Which makes it very humanlike, indeed.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
How to Access ChatGPT From Your Mac's Menu Bar
- Thread starter MacRumors
- Start date
- Sort by reaction score
I would have thought that was the reason you keep clamoring for a 15-16" iPad!
How did you get the iPhone 15 Ultra already?
That proves ChatGPT is completely useless. Everyone knows ketchup is a berry jam.
Like a president. Good point.
I showed this technology to my mom. The first thing she did was ask it to make a chicken cacciatore recipe. She actually thought it looked good and printed it out lol
I was thinking more like a car. Or a snake. Or a computer program that coldly exterminates humans without ever actually "thinking" about what it's doing. Or a president who's actually a robot with other tiny robots inside of it, working it with levers.
Why do they keep calling my Samantha ChatGPT?
"It's a hell of a lot better at providing an answer to complicated questions (but often an inaccurate one)"
There you go. Fixed!
See attached screenshot. I just asked it to do some basic math. Its answer was off by 100. It's fine if you ask it "what's 5 + 10", because the odds of it coming across that particular combination of numbers in its training data is pretty good. But it can't actually do the math itself. It's just looking things up in a "what words and numbers are likely to follow this combination of words and numbers" statistics table.
Then, when it comes to information it actually has been trained on, there are two issues:
1. ChatGPT has no idea whether the information it was trained on is true or not, and
2. There is a degree of randomness that ChatGPT injects into its answers to be "creative". This means it knowingly skews the results it provides.
Here's another good example: https://mastodon.social/@L_howes/109879173479929748
ChatGPT made up an article in response to a search query and attributed it to an author who never wrote it.
Or this example, wherein a company wondered why they had a sudden influx of people sign up to test their service, then stop using it almost immediately. Turned out that ChatGPT invented a new service that the company did not offer, and even provided developers with sample code for the imaginary service.
Don't believe ChatGPT - we do NOT offer a "phone lookup" service
TL;DR ChatGPT claims we offer an API to turn a mobile phone number into the location of the phone. We do not.
Don't get me wrong, ChatGPT is pretty cool. But it is not a search engine and shouldn't be treated as such. At least not for anything requiring accuracy. Definitely fun to play around with and certainly an interesting tool for generating realistic-sounding-but-not-necessarily-true content.
Attachments
PPL: Hi Big Tech, any solutions for public transport?
BT : Yeah robotaxis! Each one requires no driver, only holds two passengers, and needs a a team of remote support staff.
PPL: Why can’t you just help develop good software and sensors for existing buses that can transport a lot more people around?
BT : Nah, we can’t IPO your public buses. Anything else?
PPL: Any solutions for affordable healthcare?
BT : Nah, but instead of seeing a doctor you can speak to this chat bot. It consumes masses of electricity and GPUs and can give you the wrong answers but at least it is always there.
PPL: That’s not what people are asking for. Next you’ll tell us that you will fix the web by making all the web into 3D.
BT : Yes, we are working on making a 3D internet that consumes 10X more energy. Then all you poors can die in the cold and only us rich people and our nepo babies can have the planet to ourselves.
Last edited:
Great, so now we can have a hallucinatory, lying, racist-trained little icon in our menu bar that will tell us it wants to be alive, and it wants us to break up with our partners because they don't love us but it does.
No thanks.
(And before you @ me, yes I know that ChatGPT and Bing/Sydney are two different things. If you just want to reply as a pedant, you're missing my point).
No thanks.
(And before you @ me, yes I know that ChatGPT and Bing/Sydney are two different things. If you just want to reply as a pedant, you're missing my point).
Don't install the app that the story mentions. It's not part of the OS so there's nothing to block. Eventually AI/machine learning will become more tightly integrated with search engines, so I don't know what your options will be once that happens - depends on how Google/MS/Duck/Brave/Yahoo/etc. deploy it. Maybe some search engines will use this tech and some won't, and you can determine for yourself what you want to use.
None that aren't about Big Tech getting filthy rich and usually not really ever solving much of anything for actual people
Is that an Iphone20 (without the camera hump)?
I was on the fence, but I agree with several people here that it's far more dependable to use search engines and get my info from my choice of reputable sources. I did take a look at the site for my own yucks and found that MacGPT is not necessarily free unless you put a zero in the amount portion of the payment window. Still nope.
You have a masters degree in MIS. You should be learning to embrace it rather than shun it.
AI is the future. Get on board now. It already is helping me to do my job for example.
I installed it but my icon is (I think?) the OpenAI logo and not a little brain icon… did I download something different?
The "app" itself seems like a wrapper for a webpage so it's kinda dumb.
As for the ChatGPT... well... it's way better and useful than Siri
As for the ChatGPT... well... it's way better and useful than Siri
To be honest this is pretty common.
If ChatGPT was created and released by Apple, and was going to be incorporated into many Apple product most here would be wetting their pants.
Because it's something better than Apple has, if Apple has anything at all, then there will be a lot of negativity
Sadly after reading these forums for many years this is a pretty typical scenario.
Not really, it makes mistakes that people wouldn't make. For instance, I asked it to name the chapters of a Classical Japanese book. A person who was asked that question would list the chapter names they know, then the ones they are unsure of with a comment that they are unsure, then admit they don't know the rest. ChatGPT very confidently listed the chapter names, some of which were correct, and others totally made up. People just don't do that kind of thing. They know what they know and what they don't know, and they know it's good practice to admit what they don't know.
ChatGPT doesn't "know" anything. It's basically a more complicated version of autocomplete, just stringing together words that are statistically likely to follow each other. It's impressive that it's able to form coherent and grammatical sentences in this way. But there is no guarantee what it says is accurate. The problem is because its output is coherent and grammatical, people assume the content is factual. Which it sometimes is and sometimes isn't. And there's no easy way to tell when something it tells you is inaccurate.
I’m not scared, it’s just boring after a while.
What I definitely notice compared to chatbots of the past is that you need to spend more time with it until you realize “yeah, I’m definitely talking to a machine”. In the past you could tell pretty quickly, but it takes longer and longer as these models improve.