An AI company has launched a ChatGPT rival that it claims can analyse entire books and maintain conversations for days.
The chatbot is an updated version of US firm Anthropic’s large language model, known as Claude 2. It was made available to the US and the UK public on Tuesday and will be expanded to more countries in the coming months. Users can also access the bot through Quora’s chatbot service, Poe.
AI chatbots mimic the way humans talk to provide text-based responses on a range of topics. US research firm OpenAI’s ChatGPT thrust the tech into the limelight after it amassed 100 million users within two months. A wave of rivals followed, including Microsoft’s Bing and Google’s Bard.
The chatbot’s meteoric success has also prompted warnings from experts and lawmakers over its alleged risks to privacy and society. Italy banned ChatGPT in March due to data safety concerns, and tech luminaries have called for advanced AI systems to be put on hold.
Anticipating some of those fears, Anthropic says Claude is built on trust and safety practices from major organisations and corporations. The chatbot was trained using the UN Declaration of Human Rights and Apple’s terms of service, among other documents.
“We’ve been iterating to improve the underlying safety of Claude 2, so that it is more harmless and harder to prompt to produce offensive or dangerous output,” Anthropic said in a blog post.
Anthropic’s chief executive Dario Amodei met Britain’s Prime Minister Rishi Sunak in May as part of a delegation of AI bosses. Mr Sunak and tech leaders discussed the risks AI poses, from disinformation and national security to existential threats. Amodei was among dozens of tech experts to sign a statement saying that AI could lead to the extinction of humanity and should therefore be treated with the same urgency as nuclear war.
In May, the San Francisco-based company increased the length of Claude’s input and output. This meant that it could digest hundreds of pages of technical documents or even a novel. It could also write lengthier documents, including memos, letters and short stories. In addition, Claude can fetch up-to-date information from the web — a feature that is limited to paying ChatGPT users.
Like its popular rival, the chatbot has some additional perks. Alongside its longer text-based responses, Anthropic has also improved Claude’s coding skills. The company says the bot scored 71.2% up from 56% on a popular Python coding test. It also scored 88% up from 85.2% on a large set of US primary school maths problems.
