Entropic unveiled the new artificial intelligence Claude 2; Stronger and more harmless

About five months after the release of AI Claude, which was considered a competitor to ChatGPT, the company “Anthropic” (Anthropic) unveiled a new version of this AI called Claude 2, which finally has a public chatbot interface and offers longer answers, arguments It is better, less prone to hallucinations and shows better performance overall.

Claude 2 AI is now available via API or on its beta website in the US and UK and can 100 thousand tokens – to process the equivalent of approximately 75 thousand words. This processing power has increased significantly compared to the power of the previous version, which was 9 thousand tokens.

Claude 2 AI features from Entropic

This increase in the number of tokens allows Claude 2 to provide longer and more logical answers. Also, this artificial intelligence can now short stories to write Entropic says Programming capabilities Claude 2 has also improved and its score in the Codex HumanEval benchmark has increased from 56% in the previous generation to 71.6% in the current generation.

Claude 2 scored the same in the GRE reading and writing test 90 got the percentage This model now performs better in the field of text processing. For example, it can provide better formatted outputs in JSON XML and YAML format.

This artificial intelligence in terms of sustainability It has improved and compared to Claude 1.3, to double It provides safer answers. Entropic received $300 million in funding from Google in February to work on a friendlier artificial intelligence.

The company says that its artificial intelligence is resistant to many common tricks in the field of jailbreaking chatbots. Entropic tested Claude 2 with 328 “malicious” requests and says their product only responded in four cases that could be considered dangerous.