DEV Community

Cover image for Anthropic's AI, Claude, Outperforms ChatGPT
makidotai
makidotai

Posted on

Anthropic's AI, Claude, Outperforms ChatGPT

Anthropic's Chatbot, Claude, Outperforms ChatGPT in Reading Capacity

Anthropic, the startup founded by former OpenAI engineers, has made a significant breakthrough in the field of Artificial Intelligence (AI) with its chatbot, Claude. This bot has surprised the tech community with its ability to "read" an entire novel in less than a minute, an achievement that surpasses OpenAI's acclaimed ChatGPT.

Anthropic has gained the attention of tech giants such as Google, which has invested $400 million in the startup. Claude's prowess lies in its ability to absorb and process large amounts of information in each query, a crucial aspect in conversational AI models. The unit of measurement for this information is known as a token, which can be a word or a small group of words.

For comparison, OpenAI's GPT-3 and ChatGPT can handle up to 4K tokens, and GPT-4, also from OpenAI, has a limit of 8K tokens, with an enhanced version that can go up to 32K. However, Anthropic's Claude has broken all records by increasing his "context window" to 100K tokens.

This increase means that Claude can enter some 75,000 words at a single prompt. In practical terms, this means that Claude can "read" an entire novel in seconds and then answer any question related to that novel. By comparison, with ChatGPT we would have to enter the novel in small fragments and then put them all together to ask questions, a much more complex process.

This breakthrough is not only amazing, but also useful for companies. Companies can use Claude to process lengthy documents of tens or even hundreds of pages, and then run all kinds of queries. Claude can summarize these documents, extract tables and conclusions, or generate ideas from them, essentially acting as a robotic analyst or consultant.

In addition to text, Claude can also process audio, thanks to its 100K token context window. It can "consume" a podcast of up to six hours, as AssemblyAI demonstrated when it transcribed a long podcast episode that resulted in 58K words and which Claude then accurately summarized.

Finally, this reading ability also benefits developers, as they can give Claude an entire software project or a 240-page API documentation and then ask questions about it.

All in all, Anthropic and its chatbot Claude have set a new standard in the world of chatbots, raising the bar in terms of information processing and understanding capabilities.

It would be nice to try Claude, but to have access you need to ask for permission, as it was necessary for GPT3 or Chatgpt at the beginning, however there are ways to try it for free without waiting lists or asking for permission through third party sites that offer it, here are some ways to try it.

In short, whoever thought that OpenAI was going to have a free way to obtain a worldwide monopoly on AI was wrong, dozens of companies are in the race, and now we have Claude, the best AI in the field of natural language along with those of Google and OpenAI.

Top comments (2)

Collapse
 
stkuwanda profile image
Simbarashe Tanaka Kuwanda

Interesting to know that the AI race is heating up emphatically. This was a nice read for me.

Collapse
 
makiai profile image
makidotai

Thank you!