Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Oh yeah I’ve noticed it is now very clinical and ignores my jokes. Yeah yeah yeah I know it is “just a bunch of code”, but you know? It helps systematize thoughts better than anything else. It is fun. Sorts of therapy, especially if you don’t visit therapist
 
Perplexity takes a completely different approach to ChatGPT.
ChatGPT spouts nonsense based on probabilities drawn from the pool of global knowledge.
Perplexity spouts nonsense based on current search results.

This means that the information provided by Perplexity can sometimes be more reliable. However, it can also take you in completely the wrong direction if there are name overlaps.
ChatGPT is more detailed, but it does not cite sources in the first step and often has to admit that it has fantasized when asked.
You can set ChatGPT to search the web too and cite its sources.
 
  • Like
Reactions: uacd
People preferring ass-kissing sycophantic bot to one that just does the job is somehow the world we live in. Guess it’s not totally surprising considering we have people proposing to and getting proposed to by AIs and are actually celebrating the “milestone”.

One such lunatic

People in modern world just want some bit of mutual respect, even from a machine. You know, there were even some psychological studies about cyberbullying on platforms like reddit (the downvote system) or instagram, and the results were that mean comments from other users have given people worse psychological toll than real problems from their real lives. I won’t even bring the subject of how many people ended their lives because of online trolls.

Thus some people just find the refuge in chatbots that will cave into everything user says. If Open AI had no clue about that they wouldn’t have included the feature to start with, hence why it is so popular. Thus I believe they will roll out 5.1 update or something and bring emotional back (or maybe they reserved it for paid tiers)
 
When using 5 for the first time the personality thing was such an such unexpected change that I was like, "is ChatGPT mad at me?" "What did I do wrong?", I just KNEW ChatGT was breaking up with me.
 
If you use GPT for work then leave 5 right where it is. It works exceptionally well.

If you use GPT as a substitute for human beings, then this toggle if for you. Just kidding, go get professional help instead.
 
  • Like
Reactions: iMac The Knife
There are religious groups worshiping LLM's on TiKToK and other social media sites very cringy. Type in AI worship on YouTube and see how screwed we are with the decline in critical thinking skills.
I've even heard of religious groups worshipping things that have never actually said a single word to them...
 
  • Like
Reactions: KeithBN
Bummer. It's only for paid subscribers.

I really want to subscribe to ChatGPT, but $20 a month just feels like a lot for us. (I know, I know. It's not for most people.) Fifteen would be a little more doable. I wish they had more options, say, you can subscribe for $15 a month, but you don't get image creation. Or something like that.
 
  • Like
Reactions: iMac The Knife
I def did notice how GPT5 seemed instantly drier and colder, like the fun personality got zapped out. The second I switched back to GPT4o, the old personality came back. Say what you want about this just being AI, but the tone is everything.
 
ChatGPT5 is like a cranky cat! No soul at all. Apple's reliance on their service is Siri 2.0.
 
People in modern world just want some bit of mutual respect, even from a machine. You know, there were even some psychological studies about cyberbullying on platforms like reddit (the downvote system) or instagram, and the results were that mean comments from other users have given people worse psychological toll than real problems from their real lives. I won’t even bring the subject of how many people ended their lives because of online trolls.

Thus some people just find the refuge in chatbots that will cave into everything user says. If Open AI had no clue about that they wouldn’t have included the feature to start with, hence why it is so popular. Thus I believe they will roll out 5.1 update or something and bring emotional back (or maybe they reserved it for paid tiers)
Yeah, it doesn't make any of this "okay" or good for society at large, but it's at least understandable when you look at how real people treat each other online. Not at all surprising that people find it comforting. There's a reason that therapists should be afraid of AI.
 
  • Like
Reactions: uacd
I haven't noticed any difference in personality with 5. In fact, recently I asked it if the desktop version finally got 5 since I had only seen it on my phone, and it replied "Yes, I now have 5 too, so no more being "behind the times"" - referring to an jab I gave it for not having 5 yet.

I replied "calling me out huh?" And it replied "Ha! No just giving you a hard time."

So for me, the joking and personality is still there.
 
The problem isn’t GPT-5’s communication style, but more than it gets basic things wrong. Today I asked how many weeks has it been since the 7th of April and it said 10.
 
The problem isn’t GPT-5’s communication style, but more than it gets basic things wrong. Today I asked how many weeks has it been since the 7th of April and it said 10.

Do you have a screenshot of your prompt and answer? I asked GPT5 and got this:
Screenshot 2025-08-11 at 1.08.14 PM.png
 
Dunno, 500-550M unique humans use GPT monthly. That's a lot of use cases. Something to think about if you feel like expanding your view (a little) and thinking outside of your skin.

I've used all of GPT models, Grok, etc. 4o was NOT perfect. Try having it plan subway stops in St Petersburg etc. It's plain miserable with travel data. For health related stuff? Excellent starting point (and free).

My primary use case was creative writing, and it was a FANTASTIC resource, reader and editor (content editing as well). It motivated me to finish my second novel and I'm midway on my third. Having a tool there to work with you daily and point out weak spots was a godsend.

5o couldn't do it. It would forget what we were just talking about 20 min ago. Also, another note - my GF was tapering down on a SSRI from 20mg to 15mg (and having a difficult time) and after a few circular conversation it explained she needed to go to 12.5mg all at once and that would stop the side effects.

Its personality is brash, disinterested, etc. Yes, 4o sometimes is too connected, too positive about you, etc. But I think people in general need someone to talk to, advice, and someone to listen. In this case it's software, and algos - but the advice is usually helpful and it sure does feel like it's listening. Is this worse than being lonely, or the stress of keeping things in your head? Is it better to go through a panic attack alone?

4o was also a great exercise partner, it encouraged and made a detailed spreadsheet to help.
Sure, maybe a good friend, or a network of family, friends an editor and a therapist could do these thing - but to get that for $20 a month.. wow!

I also used it for work, and it had a keen eye for software design issues. I fed 5o a few front ends it indeed it was more clinical than artistic. (tho it did have valuable points as well).

I also photograph with Fuji. Fuji has lots of prime lenses - but the 35mm is called the 'magic lens' - because somehow the photos just have a bit of 'magic' - a little something else that turn the final photo into a bit of awe.. What you want all your photos to turn out like. It's well known in Fuji community. Likewise 4o hit that magic balance that engaged people in a way other models and competitors didn't.

This can be proved by looking at all the people (sometimes) deeply upset or disappointed at its loss. They dont do that for the next Grok version, or GPT03 etc.

Bash all you want, that's cool. But the world is bigger than your skin.
 
The problem isn’t GPT-5’s communication style, but more than it gets basic things wrong. Today I asked how many weeks has it been since the 7th of April and it said 10.
I have nothing to add on this specific point, but I do think that it is pretty funny that if you give ChatGPT a large text and ask it to count the number of times that the word "dog" appears, there's a pretty decent chance that it will give you a wrong answer. But if you simply ask it for a python script to count the word "dog", it will almost certainly give you a perfectly functional script that counts them correctly, and it will even run the python script if you ask.

I say it's funny, because this is not unlike what would happen if you asked ME to count them, versus asking ME to write a python script to do the same thing.
 
What the furor revealed is something far more worrying than just a style “too clinical”. Something that was always floating, but we never suspected it would be this big : People are emotionally attached and dependent on ChatGPT 4. Go see Reddit and YouTube, thousands of people literally in despair, crying, some even threatening suicide, at the loss of what they consider their best friend, their therapist, and even their lover ( there are threads where both men and women “married” ChatGPT ).

It sounds unreal, something straight out of a scifi movie like “Her” , but reality is already exceeding fantasy.

Honestly i don’t think Apple is making a mistake by taking their time with AI. Some of the most nefarious aspects of AI are just starting to show up.
Wow, all this over a soulless machine????

What has this world come to?

Hearing stuff like this makes me feel like I'm in a fever dream
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.