HomeEconomyAnthropic launches faster AI model that surpasses latest developments...

Anthropic launches faster AI model that surpasses latest developments from OpenAI and Google

Anthropic, the company created by two former OpenAI executives, has a new artificial intelligence (AI) model called Claude 3.5 Sonnet. The novelty is described as a model. “faster”but with similar costs to an earlier version of Anthropic’s midsize model, the Claude 3 Sonnet.

The model can “write, edit, and run code independently”, with what the company says are “sophisticated reasoning and problem detection capabilities.” Also highlighted is the possibility of being able to translate code “with ease”.

According to the company, considered one of the main rivals of the creator of ChatGPT, Claude 3.5 Sonnet “outperformed competing models”, such as the latest versions of the OpenAI, Google or Meta models, in several tasks. The technology created by the brothers Darío and Daniela Amodei compared the model with the GPT-4o from OpenAI, the Gemini 1.5 Pro from Google and the Llama-400b from Meta.

The US regulator wants to analyze large technological investments in artificial intelligence companies

It has become common for there to be comparisons of this type every time an AI model is presented, with results that are usually favorable to the company that reveals the new feature. In recent months, with the different launches of AI models, there are constant changes in the tables. In this case, Anthropic announces results superior to OpenAI in seven out of eight criteria, such as reasoning, knowledge, code or mathematics in several languages. However, it loses to GPT-4o in one parameter: the ability to solve mathematical problems (71.1% compared to 76.6%).

Claude 3.5 Sonnet is now available for free at place of the company and in the application of chatbot Claudio for iOS. With limitations, yes, since subscribers of the Claude Pro version and subscribers of the team version will be able to access “significantly higher limits”says Anthropo.

The company includes in the announcement the issue related to the security and privacy of the model. In addition to saying that the models developed “are subject to rigorous testing”, explains that it used “external experts to test and refine the security of the mechanisms” in this version. Anthropic sent the model for analysis to the United Kingdom’s AI safety institute, UK AISI, which shared information with its North American counterpart, US AISI. Although he does not say what feedback he received from the institutes, he explains that it includes the comment experts “to ensure robust evaluations.”

Artificial intelligence. Will Europe be able to create rivals for OpenAI or does it risk being limited by its own rules?

The Claude 3.5 Sonnet should gain company soon, as the company intends to release version 3.5 Haiku (the name given to the lighter versions of the AI ​​model) and 3.5 Opus, the more advanced version. Sonnet is the name used until now for Anthropic’s intermediate models.

Anthropic was created in 2021, in the United States. The giants Amazon and Google are among the company’s largest investors. In March, Amazon invested $2.75 billion in the startup, for a total of $4 billion in investment. Google invested around $2 billion in OpenAI rival.

Amazon raises investment in Anthropic, an artificial intelligence startup that rivals the creator of ChatGPT, to four billion

Source: Observadora

- Advertisement -

Worldwide News, Local News in London, Tips & Tricks

- Advertisement -