Get in touch
or send us a question?
CONTACT

OpenAI’s ChatGPT may have its first true rival in Anthropic’s new chatbot

AI startup Anthropic announced its new family of AI models, Claude 3, which includes Opus, its “most intelligent model,” which it said outperformed OpenAI’s GPT-4 in a range of tasks.

The family includes two other models: Sonnet, which is available now with Opus, and Haiku, the least powerful of the three, which Anthropic said will be available later.

In a series of evaluations of industry benchmarks including undergraduate- and graduate-level reasoning, math, and common knowledge, Opus outperformed GPT-4 — OpenAI’s most powerful version of ChatGPT — and Google’s Gemini.

“It exhibits near-human levels of comprehension and fluency on complex tasks, leading the frontier of general intelligence,” Anthropic said about Opus.

Meanwhile, all three models are improving in analysis and forecasting, content creation, code generation, and conversing in languages other than English, including Japanese and French, Anthropic said.

The new models can be used in live customer chats, for auto-completions, and data extraction, Anthropic said. According to the company, Haiku “can read an information and data dense research paper” that includes charts and graphs “in less than three seconds.”

The new models are also the first of Anthropic’s fleet that can process visual formats, including photos, charts, and graphs, so enterprise users can now use its models to read through PDFs and presentation slides, Anthropic said.

Unlike its existing models, Anthropic said the Claude 3 family is “significantly less likely to refuse to answer prompts” that cut too close to the system’s guardrails for harm. The company is also adding citations to the new models to show accuracy.

“Of course no model is perfect, and I think that’s a very important thing to say upfront,” Daniela Amodei, co-founder of Anthropic and former vice president of safety and policy at OpenAI, told CNBC. “We’ve tried very diligently to make these models the intersection of as capable and as safe as possible. Of course there are going to be places where the model still makes something up from time to time.”

Anthropic, which was founded by Amodei and her brother Dario, who also worked at OpenAI, in 2021, aims to develop “safer” chatbots than its competition in the AI space. The startup counts Google and Amazon as backers and has closed five funding rounds totaling around $7.3 billion over the past year, CNBC reported. In September, Amazon announced it was investing up to $4 billion and taking a minority stake in the company. Google followed up with an investment of $2 billion in October.

Source: Quarzt.