Alphabet-backed Anthropic launches OpenAI rival named Claude

Anthropic, an artificial intelligence company backed by Alphabet, released a large language model on Tuesday that directly competes with offerings from OpenAI, backed by Microsoft, the creator of ChatGPT.

Large language models are algorithms that learn to generate text by providing them with human-written training text. In recent years, researchers have achieved much more human-like results with such models by dramatically increasing the amount of data fed to them and the amount of computing power used to train them.

Claude, as Anthropic’s model is known, is designed to perform tasks similar to ChatGPT by responding to prompts with human-like text output, whether in the form of editing legal contracts or writing computer code.

But Anthropic, which was co-founded by siblings Dario and Daniela Amodei, who are both former OpenAI executives, has focused on producing AI systems that are less likely to generate offensive or dangerous content. , such as instructions for hacking or crafting. weapons, than other systems.

These AI security concerns rose to prominence last month after Microsoft said it would limit queries to its new chat-powered Bing search engine after a New York Times columnist claimed. discovered that the chatbot displayed an alter ego and produced disturbing responses during a long conversation.

Security issues have been a thorny issue for tech companies because chatbots don’t understand the meaning of the words they generate.

To avoid generating harmful content, chatbot creators often program them to avoid certain domains altogether. But this makes chatbots vulnerable to so-called “rapid engineering”, where users circumvent restrictions.

Anthropic took a different approach, giving Claude a set of principles when the model is “trained” with large amounts of textual data. Rather than trying to avoid potentially dangerous topics, Claude is designed to explain his objections, based on his principles.

“There was nothing scary. That’s one of the reasons we liked Anthropic,” Richard Robinson, chief executive of Robin AI, a London-based startup that uses AI, told Reuters. to analyze the legal contracts to which Anthropic granted early access to Claude. .

Robinson said his company had tried applying OpenAI’s technology to contracts, but found that Claude was both better at understanding dense legal language and less likely to generate strange responses.

“The challenge was more to get it to relax its constraints somewhat for genuinely acceptable uses,” Robinson said.

© Thomson Reuters 2023