I gave it assignments but I'm not a physics guys, not a maths guy, not an engineer so if you'd like to know how it did watch another video on how it compares.....
Models excel at "replaying" information they've seen before. When you use tests essentially pulled from Google, the models have likely encountered those exact scenarios, so it’s no surprise that all versions perform similarly well. This is especially evident in your Harry Potter story test-since the models know this story very well, they simply replicate it. You could ask a child the same prompt and likely get a similar answer. In almost every test, you say, "I don’t have the knowledge to read this output," and then compare how it looks a little different. This doesn’t provide any meaningful insight. For example, in the Python code tests, the differences between the models were minimal and could just as easily be attributed to randomness within a single model's output. Asking the same model the same question (with minor tweaks) often produces slightly different answers. A single comparison of the same prompt across three models doesn’t provide much value. Three models giving slightly different answers is no more informative than one model giving slightly different answers on repeat trials. All in all, I didn’t learn anything about Pro, except its pricing and that it’s slow. I honestly don’t understand how you concluded at the end, that $200 might be worth it for writing code, when it's not something you know anything about, and the tests didn’t demonstrate anything useful. After that downpour, I also want to say that your channel looks great. Very pro setup.
Yea after I paid for the subscription I wasn't too sure how I could test it emphatically. I'm definitely not their target market haha I was hoping there would be a much more obvious difference. If I ever do another AI comparison I'll keep your notes in mind!
Quality is amazing for a channel that just started man.Keep on trucking the views will come! Congrats in advance haha
Cheers for the props!
I'd get chatgpt pro since I'm using it for solo DND but even the monthly subscription is just a lot
I know it's crazy right? hopefully my section for DND showed enough of a difference for you!
can the o1 pro-mode read long document like 4o (uploaded)?
Unfortunately not. Neither o1 or o1 pro mode accept pdf. Only gif, jpeg, png & webp for the time being
I gave it assignments but I'm not a physics guys, not a maths guy, not an engineer so if you'd like to know how it did watch another video on how it compares.....
Hey gotta help out my fellow content creators am i right?
Models excel at "replaying" information they've seen before. When you use tests essentially pulled from Google, the models have likely encountered those exact scenarios, so it’s no surprise that all versions perform similarly well. This is especially evident in your Harry Potter story test-since the models know this story very well, they simply replicate it. You could ask a child the same prompt and likely get a similar answer.
In almost every test, you say, "I don’t have the knowledge to read this output," and then compare how it looks a little different. This doesn’t provide any meaningful insight. For example, in the Python code tests, the differences between the models were minimal and could just as easily be attributed to randomness within a single model's output. Asking the same model the same question (with minor tweaks) often produces slightly different answers.
A single comparison of the same prompt across three models doesn’t provide much value. Three models giving slightly different answers is no more informative than one model giving slightly different answers on repeat trials.
All in all, I didn’t learn anything about Pro, except its pricing and that it’s slow. I honestly don’t understand how you concluded at the end, that $200 might be worth it for writing code, when it's not something you know anything about, and the tests didn’t demonstrate anything useful.
After that downpour, I also want to say that your channel looks great. Very pro setup.
Yea after I paid for the subscription I wasn't too sure how I could test it emphatically. I'm definitely not their target market haha
I was hoping there would be a much more obvious difference. If I ever do another AI comparison I'll keep your notes in mind!