Anthropic said its generative artificial intelligence, Claude, can read a full-size novel (that’s an average of 75,000 words) in less than 60 seconds. This means the AI can completely read Ernest Ermingway’s A Farewell to Arms which has 72,240 words in less than a minute. A human being will finish the book in 4-5 hours in comparison. Claude is able to do this because of a memory boost from 9,000 to 100,000 tokens.
To put this in perspective, OpenAI’s ChatGPT chatbot that the public can access has about 4,000 tokens but its upgraded GPT-4 version can process tokens. Tokens can be thought of as pieces of words that the AI Claude or any chatbot can read and analyze in any book. To determine Claude’s accuracy, Anthropic tested it with The Great Gatsby and intentionally edited a sentence to contain wrong information.
Within 22 seconds, Claude detected and pointed out the error.
To demonstrate that Claude is very good at what it does, it can write a book’s summary by itself while ChatGPT will collate a summary from what humans posted online. This suggests that Claude is likely better at AI-assisted writing than ChatGPT – and that it can possibly trump Microsoft’s Bing and Google’s Bard.
The newest version of Claude is not available to the public yet and only businesses partnering with Anthropic can access the API. The cost of using the latest model is still unknown. Anthropic said the newer Claude can help corporate users to read and summarize long research papers, financial statements, legislation, and legal papers. The AI will read an entire document or book in seconds and detect errors instantly too.