Hey John, I write them myself. I also created my own ChatGPT-4o prompt generator which you can get for free here if you want: www.skool.com/learnai/about I also use Anthropic’s prompt generator in their Console
Really useful video, thanks, as I was thinking of testing out these three models myself in similar scenarios. I've always preferred Claude Opus for its writing style, but have shied away from paying $20 a month, as I don't personally use it enough for that cost. Wish the API was available for individuals, not just business.
You should have no problem getting access to the API. If you don’t use Claude much, you’ll pay hardly anything using the API. This might help: docs.anthropic.com/en/docs/getting-access-to-claude
If you have mastery over your subject and have already built up your prompt scripts Claude 3 is still hands down better as you can use the larger context window to leverage the most out of the tool
5 million characters before I have to open new windows but you can down load all the created conversation to your home computer and teach your own model on information over 10000 pages of text from conversations
Finally someone with an honest review. I immediately noticed that the new o model was dumb compared to 4 yeah I hadn't been hearing anybody talk about this until now.
same for me, using 4 as a general assistant for coding and non coding tasks in domains where I have knowledge, really just using it as a time saver over going through my notes or google. I tried 4o when it came out, but was immediately struck by the return to gpt3-like time wasting, i.e. the wasting of my time with it not following my prompts and then me having to try again and again with it, which you quickly learn is just a rabbit hole that you are not coming out of. gpt-4o, to me, is like a glorified version of 3.5 with more capabilities, it is faster than 4, can read images etc. However, gpt-4 actually gives me enough reasoning and accuracy to be helpful and limits time wasting, which is key for me, I would still pay a couple of $ a month for gpt4 over use the free gpt-4o for the time being. I guess it comes down to what you want out of these things.
I just put it to the test as far as image generation, by having it generate an image of a patient who is critically ill in the ICU, and told it to be very very accurate. I gave it great detail. I mean I wrote exactly what the vital signs were, the ventilator settings were, the medications that I wanted in the IV pumps, and went much further than what I'm stating in the comment, and boy did it fail. It created a blurry image of an ICU setting, but nothing I requested was right. It was a down right mess! A down right complete mess of equipment with display screens that were not accurate, and It just made a mess of the job. In a few years, I will see if does any better, at the exact same request, with the same wording as I used tonight, but right now, It is nowhere near ready to do something as elaborate as what I asked it to do. We'll see what it does in a few years with the same request. I smehow don't see it being much better, but I'll try. What I requested it to do, was very very extreme, as I went on and on, giving every single detail.
I've been testing GPT4o since its release. It's a huge step backwards. It's dumber, less personable, makes more mistakes, is lazier, and ignores its instructions even more than before (which was often). I don't see a button for choosing which GPT4 model to use.
Switch back to 3.5 GPT add a memory code in the custom settings now your model will have memory created 1500 lines of code usable code tokenized thousands of words with my phone because it has memory
But gpt4o is free , claude is not free. The free claude version in switzerland is like reading a fantasy book. I asked to give me the 3 best rated dermatologist in my city and the answers where totally wrong .none of this doctors excists. Copilot snd gemini did it correct. So for me as a free user( i dont pay) gemi, Gpt4o, are much more better then the useless claude 3( free version). So its important that you also look at the people who are not paying. For us gpt4o, Gemini...are 100% better then claude ( claude i would only need if i would write a fantasy story. But this I dont)... i testet Claude 3 opus ( on the Ai arena) with the same question and it gave me also totaly wrong answers ...no doctor really exists...so also Claude 3 opus is for me as a normal user totally useless. For my questions i am interested in better is gemini, gpt4o, copilot
it has no internet access. but yes lol it indeed is better for fantasy writing, actually even the free sonnet version is better than gpt 4o here. although gpt 4o is far better at math than any other model what was not tested in the video
Appreciate your feedback. For the dermatologist use case, I wouldn’t use Claude either as it’s not connected to the internet. I was testing on online business use cases, but it’s also a great reminder why we test different models for what we’re trying to accomplish. Thanks for watching 🙌
It’s all about the use case. For me, document reading and writing more natural sounding content, Claude is the way to go. Everything else I use GPT 4, for now.
Great comparison of the various models. Thanks!!
You bet! Thanks for watching 🙌
Brilliant! This was most helpful. Thanks, Rick.
Awesome, appreciate the feedback and glad it was helpful 🙌
Thanks for posting this video, it was extremely helpful. Can I ask where do you get your prompts from?
Hey John, I write them myself. I also created my own ChatGPT-4o prompt generator which you can get for free here if you want: www.skool.com/learnai/about
I also use Anthropic’s prompt generator in their Console
Thanks. May I ask what ERP and CRM you use?
Really useful video, thanks, as I was thinking of testing out these three models myself in similar scenarios. I've always preferred Claude Opus for its writing style, but have shied away from paying $20 a month, as I don't personally use it enough for that cost. Wish the API was available for individuals, not just business.
You should have no problem getting access to the API. If you don’t use Claude much, you’ll pay hardly anything using the API. This might help: docs.anthropic.com/en/docs/getting-access-to-claude
If you have mastery over your subject and have already built up your prompt scripts Claude 3 is still hands down better as you can use the larger context window to leverage the most out of the tool
5 million characters before I have to open new windows but you can down load all the created conversation to your home computer and teach your own model on information over 10000 pages of text from conversations
Finally someone with an honest review. I immediately noticed that the new o model was dumb compared to 4 yeah I hadn't been hearing anybody talk about this until now.
Funny I’ve had the complete opposite experience. It’s chatting ability is way better
same for me, using 4 as a general assistant for coding and non coding tasks in domains where I have knowledge, really just using it as a time saver over going through my notes or google. I tried 4o when it came out, but was immediately struck by the return to gpt3-like time wasting, i.e. the wasting of my time with it not following my prompts and then me having to try again and again with it, which you quickly learn is just a rabbit hole that you are not coming out of. gpt-4o, to me, is like a glorified version of 3.5 with more capabilities, it is faster than 4, can read images etc. However, gpt-4 actually gives me enough reasoning and accuracy to be helpful and limits time wasting, which is key for me, I would still pay a couple of $ a month for gpt4 over use the free gpt-4o for the time being. I guess it comes down to what you want out of these things.
I just put it to the test as far as image generation, by having it generate an image of a patient who is critically ill in the ICU, and told it to be very very accurate. I gave it great detail. I mean I wrote exactly what the vital signs were, the ventilator settings were, the medications that I wanted in the IV pumps, and went much further than what I'm stating in the comment, and boy did it fail. It created a blurry image of an ICU setting, but nothing I requested was right. It was a down right mess! A down right complete mess of equipment with display screens that were not accurate, and It just made a mess of the job. In a few years, I will see if does any better, at the exact same request, with the same wording as I used tonight, but right now, It is nowhere near ready to do something as elaborate as what I asked it to do. We'll see what it does in a few years with the same request. I smehow don't see it being much better, but I'll try. What I requested it to do, was very very extreme, as I went on and on, giving every single detail.
Interesting, I tested it the other day with an image for my AI-ify newsletter and did a good job. Although it wasn’t as extensive a description.
It’s not complete yet. Gotta wait for the image and audio features.
Yeah, I’m excited to see what comes with the audio modalities.
@@RickMulready Me too mate. Probably have to wait a bit longer now that the Sky voice that was used in the demo has been put on hold.
brilliant!!
🙌
I've been testing GPT4o since its release. It's a huge step backwards. It's dumber, less personable, makes more mistakes, is lazier, and ignores its instructions even more than before (which was often). I don't see a button for choosing which GPT4 model to use.
Switch back to 3.5 GPT add a memory code in the custom settings now your model will have memory created 1500 lines of code usable code tokenized thousands of words with my phone because it has memory
But gpt4o is free , claude is not free. The free claude version in switzerland is like reading a fantasy book. I asked to give me the 3 best rated dermatologist in my city and the answers where totally wrong .none of this doctors excists. Copilot snd gemini did it correct. So for me as a free user( i dont pay) gemi, Gpt4o, are much more better then the useless claude 3( free version). So its important that you also look at the people who are not paying. For us gpt4o, Gemini...are 100% better then claude ( claude i would only need if i would write a fantasy story. But this I dont)... i testet Claude 3 opus ( on the Ai arena) with the same question and it gave me also totaly wrong answers ...no doctor really exists...so also Claude 3 opus is for me as a normal user totally useless. For my questions i am interested in better is gemini, gpt4o, copilot
it has no internet access. but yes lol it indeed is better for fantasy writing, actually even the free sonnet version is better than gpt 4o here. although gpt 4o is far better at math than any other model what was not tested in the video
Appreciate your feedback. For the dermatologist use case, I wouldn’t use Claude either as it’s not connected to the internet. I was testing on online business use cases, but it’s also a great reminder why we test different models for what we’re trying to accomplish. Thanks for watching 🙌
It’s all about the use case. For me, document reading and writing more natural sounding content, Claude is the way to go.
Everything else I use GPT 4, for now.
Gp to