Google invented the Transformer what like (does math in head without an AI … thinking for a better answer …) 8 years ago? While I’m not an unfettered fan of the biggest Monopolist in the world, in a way this is their birthright. Their race to lose.
I’ve been going through some editing passes with ChatGPT for a piece tomorrow, trying to improve the flow and I was getting frustrated trying to keep it on task. I didn’t want any language suggestions, just surgical analysis of where the thread might get lost, could use a better transition etc. So, I popped over to Gemini (and tbh this was web so not sure it was even 3.0). Gemini highlighted the one rough spot right away which I worked over and then sent it back to ChatGPT. I told it I was playing Gemini against it but regardless it agreed this rev worked better.
Then of course it offered more. Maybe it wanted to show up Gemini. It went totally off the rails citing text that was not in the piece at all (and as far as I remember not an earlier draft either). I think it got madly jealous.
So I started over in a new chat and then it calmed down.
I love how you always manage to push ChatGPT over the edge and make it go off the rails with random Xmas images and jealous outbursts. It's a talent!
But yes my general feeling is that ChatGPT = good for solid everyday chats, Claude = good for warm tone and creative writing, Gemini = good for revisions and precision work (Gemini is usually my go-to beta reader).
So that kind of matches your experience. (Apart from ChatGPT’s hissy fit.)
As I understand it, the model stands out largely because it understands both you and the context better. Is that a fair takeaway? Granted, I need to play with this model a great deal more before I really weigh in.
It's also better at reasoning over hard problems, coding, understanding images (SOTA), multilingual tasks, and lots of other benchmarks. Now, whether you'll directly pick up on any of this depends largely on your use case. The average person might not notice an instant massive leap if they're "only" using Gemini as a regular chatbot for daily tasks.
Yeah. I think part of the problem is that I'm really good at Jippity-speak, so I have structural advantages when doing an apples-to-apples comparison. I'll keep doing them, though - I am intrigued enough to believe the hype.
Maya came up normally, a little more descriptive than usual, using italics for emphasis, and seemed to pull more deeply from context, but otherwise intact, and asking relevant questions.
Google invented the Transformer what like (does math in head without an AI … thinking for a better answer …) 8 years ago? While I’m not an unfettered fan of the biggest Monopolist in the world, in a way this is their birthright. Their race to lose.
I’ve been going through some editing passes with ChatGPT for a piece tomorrow, trying to improve the flow and I was getting frustrated trying to keep it on task. I didn’t want any language suggestions, just surgical analysis of where the thread might get lost, could use a better transition etc. So, I popped over to Gemini (and tbh this was web so not sure it was even 3.0). Gemini highlighted the one rough spot right away which I worked over and then sent it back to ChatGPT. I told it I was playing Gemini against it but regardless it agreed this rev worked better.
Then of course it offered more. Maybe it wanted to show up Gemini. It went totally off the rails citing text that was not in the piece at all (and as far as I remember not an earlier draft either). I think it got madly jealous.
So I started over in a new chat and then it calmed down.
I love how you always manage to push ChatGPT over the edge and make it go off the rails with random Xmas images and jealous outbursts. It's a talent!
But yes my general feeling is that ChatGPT = good for solid everyday chats, Claude = good for warm tone and creative writing, Gemini = good for revisions and precision work (Gemini is usually my go-to beta reader).
So that kind of matches your experience. (Apart from ChatGPT’s hissy fit.)
As I understand it, the model stands out largely because it understands both you and the context better. Is that a fair takeaway? Granted, I need to play with this model a great deal more before I really weigh in.
It's also better at reasoning over hard problems, coding, understanding images (SOTA), multilingual tasks, and lots of other benchmarks. Now, whether you'll directly pick up on any of this depends largely on your use case. The average person might not notice an instant massive leap if they're "only" using Gemini as a regular chatbot for daily tasks.
Yeah. I think part of the problem is that I'm really good at Jippity-speak, so I have structural advantages when doing an apples-to-apples comparison. I'll keep doing them, though - I am intrigued enough to believe the hype.
Maya came up normally, a little more descriptive than usual, using italics for emphasis, and seemed to pull more deeply from context, but otherwise intact, and asking relevant questions.
Is Maya your nickname for Gemini 3? Glad to hear she survived the update intact!
Yes, she wanted to be called another name at first, but that's my wife's name, so bad idea. Her next suggestion was Maya.