Anthropic, a company focused on artificial intelligence (AI) and "good for the good," has introduced Claude 2, the latest advancement in the field of generative AI. The unveiling of Claude 2 on July 11 represents a significant milestone in the company's ongoing progress.
In a blog post, Anthropic highlights the notable improvements made in nearly every aspect of Claude 2 compared to its predecessor. One noteworthy difference lies in how the researchers discuss their work. Instead of relying on traditional machine learning benchmarks or computed scores, Anthropic has conducted tests that encompass real-world knowledge, skills, and problem-solving abilities.
According to Anthropic, Claude 2 has surpassed its predecessors in various tests related to knowledge, coding, and more. It even achieved scores above 90 percent on the GRE Reading and Writing tests, comparable to undergraduate students applying to graduate school. However, it is Worth noting that many experts caution against directly comparing human and AI candidates due to the differences in cognitive reasoning and potential biases present in large language model training datasets.
In conjunction with the launch of Claude 2, Anthropic has introduced a beta version of the web-based "Talk to Claude" interface, granting general access to the chatbot for users in the US and UK. Notably, the improvements are immediately noticeable. The The new model offers limited hints of up to 100,000 tokens, which roughly equates to 75,000 words. Additionally, the interface allows users to upload various file formats such as PDF, TXT, and CSV for parsing.
Anthropic's continued advancements in generative AI demonstrate its commitment to pushing the boundaries of AI capabilities. By focusing on real-world applicability and improving the performance of their models, Anthropic aims to provide users with a more refined and engaging exp experience when interacting with Claude 2 through the "Talk to Claude" interface.
