Ok, what everyone really wants to know is whether it is worth paying the $20 per month to get ChatGPT-4 rather than ChatGPT-3.5?
The question to ask yourself is, do you already use ChatGPT-3.5 quite a lot? Then, it’s worth it. It’s like buying an air fryer: if you hardly ever cooked before, an air fryer isn’t going to get you cooking.
I use ChatGPT-4 a lot. I am keeping an eye out for other options (the newest META LLM, LLaMA 70b, is apparently beating GPT-4 in coding tasks…), but for now ChatGPT-4 is top dog.
Read on to learn why this is the case, or if you are looking for a direct GPT-4 v GPT-3.5 comparison, head over here: ChatGPT-4 vs ChatGPT-3.5.
Ease of Use
OpenAI, the makers of ChatGPT, have given us a lesson in simplicity. When creating a new product it is easy to get caught up in every design detail; colours, typography, animation: the sheen. Really, the user just wants something that works and solves a problem for them.
The difference between user interface (the sheen) – UI – and user experience: UX.
Open AI have prioritised UX. There are no, and I mean NO, bells and whistles here.
Very simple, no distractions, exactly what is needed.
What we aim for at AI Reviews Trust too.
Quality of Output
Herein lies the first jump up from GPT-3.5. For all LLMs, the most significant way to improve them is to give them more data to train on. This is also the limiting factor for anyone trying to make their own LLM. The code is relatively straight forward, but you need a few billion dollars to fund the computer power needed to train models to compete with GPT-4.
It was trained on an estimated 1.76 trillion parameters and is actually 8 models working together (from the decoder). This is a significant difference to GPT-3, trained on about 175B parameters.
Versatility and Functionality
sd
