Which AI chatbot is the best?
Comparing ChatGPT, Claude and PI and which one you should use for your tasks
Read time: 5 minutes | Subscribe to this newsletter
Hey there! Moritz here. Thank you for joining me on another post of The Prompt Warrior.
This week the AI wars started heating up again. Two big things happened:
The ChatGPT Code Interpreter was finally opened up to all Plus users
Anthropic launched Claude v2, a ChatGPT competitor
With the models continuously getting better and better, the natural question arises: Which model should you be using?
I want to break this question down in today’s post.
I’ll go into ChatGPT Code Interpreter use cases, what benefits Claude has over ChatGPT and finally also talk about Inflection’s Pi (this one might surprise you).
Before we begin, a big thank you to our sponsors who keep this newsletter free to the reader.
This week’s newsletter is sponsored by AIPRM.
With AIPRM, you can build a prompt library on ChatGPT for your team. Crush marketing, sales, SEO, support, and productivity writing. Choose from thousands of prompts or build your own. Stop staring at a blank ChatGPT window. There’s a prompt for that.
ChatGPT + Code Interpreter
I’ve stated in my recent tweet that calling the Code Interpreter just another GPT-4 plugin that does data analysis is not really doing it justice. Rather I consider it almost a GPT-4.5, a fundamentally upgraded version of GPT-4.
GPT-4 was already able to produce code, but it wasn't able to run it. The Code Interpreter can. This opens up a whole new world of possibilities (see the use cases below).
The Code Interpreter seems to have built in iterative abilities. It recognizes when it's made a mistake and corrects it by itself. This makes it closer to an agent rather than an assistant.
The Code Interpreter now takes files as input. This is the first step in opening up the multimodal functions of GPT-4, meaning that it will be able to understand not just text but also images, audio and video. While the multimodal functions are built into GPT-4, OpenAI has not made them widely accessible yet. Expect this to happen soon.
Here is a non-exhaustive list of some use cases of the Code Interpreter that you can try out. I will also be diving deeper into some of them in the coming weeks.
Read and analyze uploaded files (you can upload one file or a .zip file)
Analyze datasets and basically assist as a data scientist
Read out text from images
Manipulate images, such as applying filters
Transform images into videos
Create a simple video game from scratch
Generate QR codes
Analyzing data: One of Code Interpreter’s strongest use cases
Anthropic’s Claude v2
The release of Claude v2 made a big splash this week. It’s now turning out to be a serious contender for ChatGPT.
One of its standout features is its impressive 100k context window (= 75,000 words or 150 pages), which enables it to take into account larger amounts of text when generating output. In contrast, GPT-4 only considers 32k tokens.
What does this mean for you?
You can now train the model on your writing style in a much better way. (I’ll be diving deeper into this one in an upcoming post as well).
You can upload entire books or PDFs and let the AI model reference every part of it with complete accuracy.
The model won’t forget the beginning of your conversations the longer the conversation gets. (I have this problem with ChatGPT).
A supposedly friendlier tone than ChatGPT
A 3-5x more affordable API than GPT
However, it is worth noting that many people, including me, have noticed that Claude v2 tends to hallucinate quite a lot. So watch out for that and always double check.
Overall when it comes to comparing ChatGPT and Claude and asking yourself which model to use, it should be dependant on your task. For most work related tasks I am still using ChatGPT, especially now that it has the upgraded Code Interpreter functionalities. When it comes to tasks where larger contexts are needed, I might be switching to Claude.
Another interesting model I want to mention is Pi by Inflection. I recently tried it out and found it fascinating.
It might not yet have the same popularity as ChatGPT or Bard. But the company raised an insane 1.3B in funding recently so expect it to make more noise soon.
Its focus is not so much on work use cases and it differs from ChatGPT in a number of ways:
It’s much more conversational.
It feels like you are talking to someone with empathy. It’s almost human-like.
The interface is only one chat, not multiple chat threads.
This how my conversation with Pi started:
I loved how PI was being so positive and energetic in this chat (even though I wasn’t). It felt like talking to a friend.
You can select from a couple of different openers to start a conversation with Pi, such as journaling, thinking something through or helping you find motivation.
You can see that the chats revolve more around introspection, mindfulness and companionship. It aims to be a more personal assistant than GPT.
That’s it for today!
If you enjoyed this, please help spread the love by forwarding this newsletter to a friend or colleague.
How valuable was this email to you?
I hope to see you in the next one!