Anthropic, a synthetic intelligence startup positioning itself because the builder of a safer sort of chatbot, has launched a brand new model of its AI bot, named Claude.
Anthropic mentioned that Claude 2 is accessible to anybody within the US or UK on-line at claude.ai, and companies can entry it through an utility programming interface. The brand new launch on Tuesday comes a number of months after Anthropic started providing an earlier model of Claude to companies that needed so as to add it to their merchandise. Beforehand, the bot was examined by a handful of firms together with Quora Inc., which constructed it into an app known as Poe that lets customers ask questions.
Like its predecessor, Claude 2 is constructed atop a big language mannequin and can be utilized for written duties like summarizing, looking, answering questions and coding. Each fashions can presently absorb giant chunks of textual content — a person can ask it to summarize a guide, as an illustration — although Claude 2 can generate longer responses than its predecessor. Responses can attain as much as about 3,000 phrases, in accordance with information offered by the corporate. Claude 2 will even supply extra correct responses on some matters, reminiscent of coding and grade-school-level math, the corporate mentioned.
Anthropic’s purpose has been for Claude to be much less prone than different chatbots to manipulation. The corporate goals to manage the technology of dangerous or offensive ends in response to prompts — points which have dogged rival chatbots reminiscent of OpenAI’s ChatGPT. The startup mentioned that its checks indicated Claude 2 was higher than the present mannequin at producing what Anthropic calls “innocent responses” to “dangerous prompts.”
Anthropic president and co-founder Daniela Amodei mentioned that the newer mannequin is supposed to “really feel, type of subjectively, a bit bit higher” than the older model. She added: “And we additionally know that it is nonetheless not good. It nonetheless has limitations.”
Anthropic, based mostly in San Francisco, was based in 2021 by a handful of former OpenAI staffers, together with Amodei and her brother, Dario Amodei. The startup and its management have urged warning concerning the rush to develop and launch AI techniques, and their potential to rework industries.
In Might, the corporate mentioned it raised $450 million in funding, bringing the corporate’s whole raised to date to greater than $1 billion.
Anthropic mentioned it plans to make Claude obtainable to folks in additional international locations within the months forward.
