Anthropic, the startup founded by former employees of OpenAI, introduced a new language model, Claude, to rival ChatGPT in the past months. The company, which also has an investment of 300 million dollars from Google, stated that the model had less harmful results. This was very important because competitors were being criticized for many harmful consequences, such as misinformation.
Anthropic, which has been operating since 2020, announced today that it has brought some innovations to Claude. Accordingly, the amount of input and output that Claude can process during use has been expanded. This means a huge improvement for existing models.
Claude’s word limit increased to 75,000, the chatbot can read a novel in less than a minute
OpenAI’s popular model ChatGPT has a 3000 word limit. This is a small amount, although it can be overcome by different methods. Anthrhopic, on the other hand, says in its announcement that it has increased the limit of its model Claude to 75,000. As you can imagine, this means serious progress.
The company says in a blog post it shared that this expansion shows that it’s enough to process The Great Gatsby novel in one go. So Claude can read a novel very quickly. By changing a single sentence in the novel, Anthropic asked the chatbot to notice the change. Claude did this in just 22 seconds. According to Anthropic, it takes around 5 hours for a human to read 75,000 words.
Artificial intelligence language models do not measure information by the number of words or characters, but by semantic units called “tokens”. By breaking down text into small chunks, this concept helps language models understand, predict, and produce texts. This is also quite normal; because the words are short or long does not show the complexity of their meanings.
The increase in capacity has the potential to enable many functions such as reading, summarizing, analyzing long documents to be done in minutes.
Here’s what Anthropic says now Claude can process 100,000 tokens, the equivalent of 75,000 words. Previously, this number was 9000 tokens. By comparison, OpenAI’s most advanced language model, GPT-4, can handle 8000 tokens. In the full version of this model, this number increases to 32000. The innovation that came to Anthropic also reveals that the information processing capacity of artificial intelligence models is increasing rapidly.
For now, the new feature opened for Anthropic’s partners; It is of great importance because of its potential to enable reading, summarizing and analyzing operations in minutes and longer conversations with artificial intelligence.
30669
