Chatbots that use artificial intelligence (AI) are getting increasingly effective as the use of this technology spreads. One such instance is Claude, a chatbot created by Google-backed AI startup Anthropic. The business, which was established by former OpenAI engineers, asserted that their chatbot can now read about 75,000 words in a matter of seconds.
Anthropic claimed to have enlarged the context window of Claude from 9K to 100K tokens, or roughly 75,000 words, in a statement about the enhanced functionality of Claude.
The fundamental text or code building blocks that a large language model (LLM) AI utilises to process and produce language are known as tokens. Characters, words, subwords, and other text or code fragments can all be used as tokens.
Businesses can now send hundreds of pages of content for Claude to read and assess thanks to an expanded token context window. Additionally, it implies that discussions with Claude can last for several hours or perhaps days.
According to a report from The Verge, ChatGPT has a token restriction of approximately 3,000 words. Additionally, it stated that a limited-release edition of OpenAI’s GPT-4 can handle up to 32,000 tokens and processes about 8,000 tokens.
The Great Gatsby was read by Claude in 22 seconds
The complete text of The Great Gatsby novel, according to Anthropic, was put into Claude-Instant, which handles 72,000 tokens, and one line was edited. The model gave the right response in 22 seconds when asked to identify the differences between the original text and the updated text.
According to a study published in Science Direct, the average human can read 100,000 tokens of text in roughly 5 hours. Additionally, it takes humans longer to process, retain, and think critically about that data.
Anthropic recently released the guidelines for their “Constitutional AI” training approach, which aims to allay worries about the decision-making, transparency, and safety of AI systems.