Is Anthropic's revamped Claude AI chatbot better than OpenAI's ChatGPT
What's the story
From writing code and essays to passing difficult exams, chatbots have stunned us all.
Now, Anthropic—an AI start-up founded by former OpenAI engineers—has updated its Claude AI chatbot, giving it the power to read through an entire novel in less than a minute.
It makes one wonder if the Claude chatbot, with its improved abilities, fares better than ChatGPT. Let's find out.
Limitation
Claude's context window has been expanded to 75,000 words
Chatbots are trained on large amounts of information but the reference data that these AI systems can process when in use is narrow. Their 'context window,' which refers to the combination of text input and output, is limited.
Anthropic has now expanded the context window of Claude, increasing it to roughly 75,000 words.
ChatGPT pales in comparison, with just 3,000 words.
Testing
Claude can process large amounts of data in seconds
With its upgraded context window, Claude could process the entirety of the novel, The Great Gatsby, in one shot.
The company tested the chatbot by changing one line in the novel to say Mr. Carraway was "a software engineer that works on machine learning tooling at Anthropic."
When asked to spot the error, the system gave the correct answer, that too in 22 seconds.
Tokens
Claude can process significantly more tokens
AI language models process information in what are called tokens, that is the input data is mapped into a sequence of meaningful data.
Now, Claude's context window has been updated from 9,000 to 100,000 tokens.
In comparison, OpenAI's most advanced system GPT-4 can process about 8,000 tokens while a limited-release model of GPT-4 can tackle up to 32,000 tokens.
Detai
What else can Claude do?
Claude, which was released in March, can give summaries, search, help with "creative and collaborative writing," write code, and answer questions—pretty much what ChatGPT can do.
Per the company, Claude is "less likely to produce harmful outputs," and is also "easier to converse with."
The chatbot is claimed to be "more steerable," allowing you to get the "desired output with less effort."
Information
Claude cannot access the internet
The ultimate aim of Anthropic is to make Claude "helpful, honest, and harmless." However, what could be the major difference is that, unlike ChatGPT, Claude has no ability to access the internet, and is built to be "self-contained."
Comparison
Which is better?
When comparing brute power, the updated Claude eclipses ChatGPT. However, it is yet to go mainstream. The supercharged Claude is currently available only to Anthropic's business partners.
Further, ChatGPT can perform several tasks more efficiently, from generating code to following prompts and giving better explanations.
Claude, on the other hand, appears to fall short on some parameters, especially coding and explanations.