In 2022 we saw a breakthrough in the field of artificial intelligence and chatbots. OpenAI launched ChatGPT to the public, a chatbot capable of processing and offering responses in natural language with a precision that ended up surprising everyone. Based on the company’s GPT-4 language model, the tool He has been turning the industry upside down for months, to such an extent that many companies have also integrated chatbots into their tools and services. The clearest example today is Microsoft with Bing Chat, also based on OpenAI technology, or Google with Bard, which is already available in many countries except Europe.
The fact that OpenAI went ahead with its product has led to a multitude of companies also venturing into this sector, launching their own conversational chatbot. Anthropic was already surprising with its response to ChatGPT through ‘Claude’, an AI with a much more powerful language model than GPT-4. In fact, the company recently announced that its model is already capable of processing and analyzing up to 100,000 text tokensthree times more than the most advanced OpenAI model to date.
A great tool for processing and analyzing long texts
To get a clearer idea of what 100,000 text tokens means, basically Claude would now have the ability to process, analyze, and provide responses to 6 hours of audio, the entire text of the Great Gatsby, 240 pages of documentation on an API, or the complete 89-page corporate archive of Netflix, to give a series of examples that further clarify the doubts. In essence, 100,000 tokens is about 75,000 words.
Mentioning the Great Gatbsy has not been a coincidence. And it is that Anthropic itself has used this example to explain Claude’s improvements. As they explain, they have introduced the entire classic novel by Francis Scott Fitzgerald (which by the way has 75,000 text tokens) and have modified just one of its lines to say: “Mr. Carraway was a software engineer who worked on software tools.” machine learning for Anthropic”. They have then instructed Claude to spot the difference, to which the AI did this job in just 22 seconds.
Claude’s evolution has been surprising, as he went from his 9,000 capacity tokens to 100,000. Given the large volume of work you can tackle now, It will be a very useful tool, especially for work environments where large amounts of text need to be processed and analyzed. Anthropic ensures that the average time for a person to read about 100,000 text tokens is approximately 5 hours. To this we also have to add the time it takes to analyze and memorize the information. Claude can do this in less than a minute, making him a very valuable resource for those more tedious tasks.
Claude can not only be used as a way to process long texts, since it is also essentially a chatbot, so we can talk with him about all kinds of topics and get answers to complex questions. From Anthropic they assure that it is a much better tool than the proposals based on vector search.
From the company they assure that this improvement is already available in your API, so any application that makes use of it will have the great power of Claude at its disposal. Through its website we can request access to this tool, as well as test it in Slack.
More information | anthropic
In Genbeta | From the ‘Google rate’ to the ‘ChatGPT rate’: Spanish media want to charge for the use that AI makes of their content