GPT has had free image generation and running of python code for a few months before Deepseek R1 came out, so the dethroning happened on the currently available GPT model. Though I don’t find deepseek categorically better for everything, the online version they have tends to overcomplicate things where GPT makes a much more to-the-point answer (both with the exact same copy-pasted prompt). Like for code Deepseek started going into an overcomplicated JS function one time, adding more and more into it to make it work despite that it clearly didn’t. GPT got it right on the first try. Though I appreciate that deepseek considered accessibility in its code from the start.
There’s places where deepseek can be better but for me it’s now more of a side-by-side thing. And if I want to run code online for data analysis and stuff then there’s only chatGPT. If deepseek restores online search though that will be OP. It used to easily pull from 50 different sources, that’s 50 search results you can look at.
I think my bigger problem is that I feel like with every new generations the models are becoming more and more neutralized lol. I feel like you need to be asking EXACTLY what you want it to do and nothing short of that, but if I’m asking AI it’s because I can’t formulate it exactly for myself! I shouldn’t have to be writing a full one-page prompt to ask just one question. Google used to answer my questions just fine with just a sentence prompt before they completely broke their search engine. Now GPT talks to you like you’re its friend with emojis and stuff and I had to specifically go into the custom prompt and tell it not to do that lol. You’re a machine, talk to me like one lmao. It started using slang and basically talking like a teenager.
GPT has had free image generation and running of python code for a few months before Deepseek R1 came out, so the dethroning happened on the currently available GPT model. Though I don’t find deepseek categorically better for everything, the online version they have tends to overcomplicate things where GPT makes a much more to-the-point answer (both with the exact same copy-pasted prompt). Like for code Deepseek started going into an overcomplicated JS function one time, adding more and more into it to make it work despite that it clearly didn’t. GPT got it right on the first try. Though I appreciate that deepseek considered accessibility in its code from the start.
There’s places where deepseek can be better but for me it’s now more of a side-by-side thing. And if I want to run code online for data analysis and stuff then there’s only chatGPT. If deepseek restores online search though that will be OP. It used to easily pull from 50 different sources, that’s 50 search results you can look at.
I think my bigger problem is that I feel like with every new generations the models are becoming more and more neutralized lol. I feel like you need to be asking EXACTLY what you want it to do and nothing short of that, but if I’m asking AI it’s because I can’t formulate it exactly for myself! I shouldn’t have to be writing a full one-page prompt to ask just one question. Google used to answer my questions just fine with just a sentence prompt before they completely broke their search engine. Now GPT talks to you like you’re its friend with emojis and stuff and I had to specifically go into the custom prompt and tell it not to do that lol. You’re a machine, talk to me like one lmao. It started using slang and basically talking like a teenager.