My company has blanket ban on using AI, too, and all domains are blocked.
It is fairly comical to see people rant and froth at the mouth about Google and privacy, then they willing and happily give openAI all their data.
Last edited:
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.
OpenAI Debuts New GPT-4.5 Model for ChatGPT
- Thread starter MacRumors
- Start date
- Sort by reaction score
My company has blanket ban on using AI, too, and all domains are blocked.
It is fairly comical to see people rant and froth at the mouth about Google and privacy, then they willing and happily give openAI all their data.
Except when you search through ChatGPT, you never know if what it finds is real or if it hallucinates. You would have to re-check everything anyway to make sure that it didn’t hallucinate.
I can't see Apple not gatekeeping this somehow. They have to control everything and in the case of AI, dumb it down so it's less useful than the vanilla service.
I have no idea what sort of agreement they might have, but you can choose to have your ChatGPT account linked with Siri in Settings. I can’t imagine what point there is to that if you don’t get the benefits of your ChatGPT account. 🤷♂️
I just started my tax refunds and the tax software has ChatGPT built in. It literally saved my at least an hour of research!
AI has already helped me save tons of time and even money. It's like having a buddy you can ask anything. Like your best friend is a cook and you can ask him about some cooking ideas or secrets. When you need your buddy to be a tax expert he can be just be that. Or a history teacher. Or whatever you need to be done.
Just this morning I asked ChatGPT how I could recalibrate the hands of my chronograph because they weren’t exactly at 12 o’clock when reset. Within a minute, I found out that I don’t need to go to a jeweler and can do it myself, which already saved me 200 euros or more. I was also shown how to do it. The hands had been like that for years, and I didn’t feel like searching for the manual somewhere in the house or going to a jeweler because I was too stingy.
Surely I could have googled it or read some watch forums and try to find a solution but this was so convenient and fast. And its just one example of many things ChatGPT has helped me over the last months.
The rule is not in effect yet since the countries have to implement it. My company is having a deal with ChatGPT to use a sandboxed version so all our company data stays within our company. The public ChatGPT link is blocked since the data is not protected.
I just think it’s crazy to pay $200/month on practically anything, I mean I don’t even pay that much a month on car insurance lol
They know companies are paying for it. Just like some health related apps are ridiculously overpriced (there is a tinitus app for like 300 euro) because they know some health insurances subside it or how some tiny furnished apartments are over 1k in Berlin because US based companies are willing to pay for it for their expats abroad
I don't have Pro but I have Plus and API access so I was able to try it with one of the Apps I wrote using a standard test I use.
Yes it's better but it's
OpenAI's reasoning models seem pretty good though. I really like the new "Deep Research" button in the App (Mac & iOS) for researching online for best buys etc. e.g. I asked it for suggestions for a new car and it asked me questions to get the exact criteria then went off for 10mins and came back with a pretty sound report including financing etc.
Our son had a rash from surfing on the beach, we told it we wanted suggestions for products from local pharmacy warehouse and it went off for 10mins, using that pharmacy's website as the source, and came back with some good suggestions including one that we hadn't thought off which was specific to the situation and symptoms.
This is a real tool. Google is no where. Why isn't google in this race? I no longer "google" for answers...
Yes it's better but it's
- 3 times slower than any model I've used via the API, unusable I think in app for now
- expensive. One Q/A costs about 30c so I can see why this is for Pro users only at the moment.
OpenAI's reasoning models seem pretty good though. I really like the new "Deep Research" button in the App (Mac & iOS) for researching online for best buys etc. e.g. I asked it for suggestions for a new car and it asked me questions to get the exact criteria then went off for 10mins and came back with a pretty sound report including financing etc.
Our son had a rash from surfing on the beach, we told it we wanted suggestions for products from local pharmacy warehouse and it went off for 10mins, using that pharmacy's website as the source, and came back with some good suggestions including one that we hadn't thought off which was specific to the situation and symptoms.
This is a real tool. Google is no where. Why isn't google in this race? I no longer "google" for answers...
Indeed! It would cost $2k monthly if they paid royalties to the sources from which they stole the contents …
But I’m not talking about their team or enterprise plans. I’m talking about their pro plan, which is an $200 individual subscription for one user.
I completely understand companies buying their enterprise plans though. But I’m not exactly sure who a $200 plan that can’t even be shared with anyone is intended for.
You can tell it to provide sources and check them yourself. It’s still faster than trying to search for it all manually, especially if it’s information you were ignorant to previously, like OP’s grant potential. I generally use Perplexity more myself.
It’s way better now that you can link to the internet and see all the links. It will make stuff up still from time to time but I use it dozens of times per day and almost never see made up things anymore.
/Note about the term “hallucinate”. I know this is what’s used, but a more correct term is “confabulate”. Hallucination is sensing things that aren’t there. Confabulate is filling in gaps in knowledge or memory with things that aren’t true.
Well, then don't help them. What is the problem? With your logic why are you even on the internet?
IMO, OpenAI is losing - they charge a lot more for slightly inferior products.
Recently been using Claude 3.5 (and 3.7) side-by-side with 4o for a big analysis project (in Cursor) - ChatGPT consistently shows worse results and more hallucinations.
Not sure if much, much more expensive 4.5 would be THAT good, considering...
Recently been using Claude 3.5 (and 3.7) side-by-side with 4o for a big analysis project (in Cursor) - ChatGPT consistently shows worse results and more hallucinations.
Not sure if much, much more expensive 4.5 would be THAT good, considering...
There's Perplexity for that. Their subscription pays for itself if you use if for work. Especially in reasoning mode, you just can't beat it. It does literally hours of research in minutes.
Last edited:
That is just stupid. LLMs, if used properly, provide insane advantages if you deal with any kind of analysis, coding or content-creation tasks.
What would take me hours digging through code in a huge analytics problem, it literally finds my issue in seconds. It’s a neutral tool, it’s all about the user and their intentions. To ban it in a work setting is nothing short of sitting down while the competition marathons by.
If you use a free version, you will learn not to waste your queries. I think the free version used to have a limit of ten queries within three hours and I never need that many. However processing large files is not possible in the free plan.
Paying for AI can be limiting though. Right now I use several free models, but if you pay, you will likely want to pay for only one model.
People say that a subscription is worth it, as it generates more money than it costs. That is the problem though: You need to generate money. Adobe subscriptions have the same problem. Most people do not generate money with their queries or with their photos. It always annoys me if commercial use drives up the prices for a product.
If an employee saves time with AI, he might no longer be needed soon.
Paying for AI can be limiting though. Right now I use several free models, but if you pay, you will likely want to pay for only one model.
People say that a subscription is worth it, as it generates more money than it costs. That is the problem though: You need to generate money. Adobe subscriptions have the same problem. Most people do not generate money with their queries or with their photos. It always annoys me if commercial use drives up the prices for a product.
If an employee saves time with AI, he might no longer be needed soon.
We don't code. It's more important to make sure someone isn't uploading the internal sales or revenue reports to save some time going through it themselves.
Right now they are auditing (is that a word?) Copilot to make sure it's also GDPR compliant (probably isn't).
I am just stating how it is btw. I still use it working from home to rewrite my emails to c-level every day haha