In the Great War of AI, which is currently going on, Anthropic is a company that is known to be slightly less than Google (apparently) or Openai for the general public. However, his talk agent, Claude AI, has been honored by its users. And it is unlikely to change with the latest declaration of anthropic, which is launching citations, which is a feature that aims to improve the reliability of the response created by its cloud models. The purpose of this new feature is to reduce the mistakes by making it possible to anchor the answers in the correct documents and sources.
New Coats Feature Claude Language Pattern is available via API. © Bashari
It is important to say that despite their rapidly impressive effects, artificial intelligence models are open to criticism (and criticism) due to the tendency to produce incorrect or unverified information. AI’s famous “deception”. To solve this problem, Anthropic has introduced Citizens, a new feature that is accessible through its API and Google Vertex AI platform. This solution allows developers to embed the source files that use Claude 3.5 Sonnet and Claude 3.5 Haiku models to provide accurate references in their answers.
More for developers
According to Anthropic, the citations will significantly improve the verification of traceability and specifically produced information. By the developer, the theory will no longer need to resort to complex commands or no longer will a long prompt engineering process be guaranteed to get answers that are guaranteed to come from reliable sources. Developers can import documents into the Anthropic API in the PDF or simple text format. Once these files are integrated, the cloud models analyze the information contained in it and automatically enter the correct references in their answers. These references also refer to the exact excerpts used in the source document.
Therefore, if a user asks a summary of a report or asks a complex question based on a combination of text, the model will not only provide detailed answers, but also provide relevant references. This method is especially useful in such fields such as automatic document summarizing, complex questions and answers or customer support.
Easy integration, but not free.
Still, according to anthropic, one of the main benefits of citations is ease of use. The developer can activate this feature in a few easy stages via API. A specific command allows you to activate automatic entry of references when you process the documents provided. However, it is important to note that this new feature is not free. Its price depends on the length and number of used source documents. For example, the document containing about 100 pages costs approximately $ 0.30 with the cloud 3.5 Sonate model and $ 0.08 with Claude 3.5 Haiko. For developers, this additional cost will have to be weighed in terms of credibility.
Advertising battle
Finally, it is impossible to talk about the quote without mentioning the general context in which it occurs. Generative AI’s key providers are removing all stops at the beginning of the year. In addition to the citations, Openai has also revealed its operator agents, while Perplexity is launching its Sonor Search API and the French company Mistral AI has announced a partnership with AFP. It is interesting to note that all these actors are not just imitating each other and it seems that they are trying to chart their own path. With citations, Anthropic is taking himself as an actor and even more positioning that is credible and prominent on ethics.