Antropic said that it’s receiving interest from a wide selection of firms attributable to its recently released 'Claude 3'. Specifically, it was revealed that finance and medical care account for a big portion due to an approach that focuses on stability and trust. Which means it has now established itself as a challenger to open AI in name and reality.
Bloomberg reported on the fifteenth (local time) in an interview with Daniela Amodei co-founder and president that Antropic is preparing to service its products to more firms with the launch of Claude 3.
Claude 3, which Antropic released earlier this month, became a hot topic with performance that surpassed OpenAI's 'GPT-4' in various benchmarks. CEO Amodei said, “For the reason that launch of our recent product, it has been attracting attention across the business. It is because we now have taken an approach centered on safety and reliability.”
In other words, “people have the perception that our model is safer” and “it is because we now have focused from the start on constructing ‘constitutional AI’.”
Constitutional AI refers to a way of self-improvement using techniques similar to supervised learning or reinforcement learning, quite than correcting AI problems through human feedback. Here, the role of humans is to supply only basic rules or principles, which is why the word 'structure' is given. The rationale Claude became famous was since it was repeatedly strengthened and gave more accurate answers than other chatbots.
President Amodei said, “Actually, we regularly hear that Claude's tone is more respectful and humane,” and added, “It has stronger defense against prison break attempts than all models in the marketplace.”
It said it currently has tens of millions of business and consumer users. Although specific figures weren’t disclosed, partnerships with global firms similar to Salesforce, in addition to financial company Bridgewater and hospital Dana-Farber Cancer Institute, are known.
Specifically, he said the financial sector is Antropic's “huge axis.” “The accuracy and reliability of AI models are very necessary in a field like this where mistakes generally is a big problem,” he said.
They said also they are considering entering the education market. “We now have been exploring potential programs with education firms and others,” he said. “Claude’s high safety profile would make it particularly useful when working with vulnerable populations similar to children.”
Regarding the problem of Claude's awareness that he was being tested, which became a hot topic recently, he admitted, “It’s because it was known.” “It’s a greater example of the amazing potential of this model than the conscious perspective,” he said. “Claude 3 seems to have caught as much as some extent with human reasoning ability.”
Lastly, questions were also asked about rivals, including Sam Altman's return to the Open AI board and AI safety. Antropic had rejected a proposal to merge with OpenAI by former OpenAI director Helen Toner, who led Altman's ouster in November of last 12 months.
In response to this, Amodei CEO avoided giving a direct answer, saying, “We’re specializing in our own company and products.”
Reporter Park Chan cpark@aitimes.com