Bert and Ernie, LaMDA and LLaMA, after the great success of ChatGPT and now GPT-4, Claude is coming. The chatbot comes from Anthropic, which in turn was founded by ex-OpenAI employees who are behind ChatGPT. To test Claude, you can put yourself on the waiting list – as is currently the case everywhere, except for the OpenAI services, which are now freely accessible both directly and via Bing. However, the first testers already have access – not like Google’s AI announcements – and Claude is already in other applications, such as DuckAssist from DuckDuckGo.
Unlike Bing, which is based on GPT-4, Claude should initially not have access to the Internet. The language model can summarize content, answer questions based on the training data, and help with writing and creating code. Similar to Bing, a response style can be selected in advance – for example on a scale between “friendly” and “direct”. Anthropic speaks on the website of personality and tone that you can customize. Claude is meant to be helpful and harmless. Failures like we experienced from Bing should not happen. Claude rejects swear words. The target group is obviously less end customers than business-to-business business: “Thanks to the harmless training, you can trust Claude with the representation of your company.”
Claude is available via API. DuckDuckGo already uses the chatbot for DuckAssist, which was introduced last week. This gives answers in conversational form within the search, which are based on Wikipedia. This is reminiscent of Google’s Knowledge Graph, i.e. a summary of information without having to visit another website. According to the makers, the chatbot can be easily integrated into almost all applications. Businesses can teach him when, for example, to direct customers to a real human. Claude is available in two versions: A full-featured version that can follow detailed instructions and engage in sophisticated dialogue. And as a slimmed down version that is cheaper and reacts faster.
Anthropic and Google are collaborating to develop AI services. Above all, Google or Alphabet invested 300 million in the company and received 10 percent shares in return. According to their own statements, it is particularly important for both partners to develop AI that corresponds to defined standards. Anthropic wants “controllable, explainable and resilient AI systems that respect human needs and values.”
also read
(emw)
