ChatGPT Gets a ‘Helpful, Honest, and Harmless’ AI Rival Called Claude

AI safety and research company Anthropic just released its next-generation AI assistant called Claude(Opens in a new window).

Claude is a competitor to OpenAI’s ChatGPT, which has already generated some negative headlines. However, Claude is designed to be different, with a focus on being “helpful, honest, and harmless” while still managing to carry out all the tasks you’d expect of an AI assistant (for example, summarization, search, creative and collaborative writing, Q&A, and coding).

As Reuters reports(Opens in a new window), Anthropic was co-founded by former OpenAI executives Dario and Daniela Amodei, and have backing from Alphabet. The reason Claude is different and can be relied upon to be helpful and trustworthy is because the AI has been provided with a list of rules/principles and trained through self-improvement. It’s a method known as “Constitutional AI(Opens in a new window).”

As Anthropic explains:

“The process involves both a supervised learning and a reinforcement learning phase. In the supervised phase we sample from an initial model, then generate self-critiques and revisions, and then finetune the original model on revised responses. In the RL phase, we sample from the finetuned model, use a model to evaluate which of the two samples is better, and then train a preference model from this dataset of AI preferences. We then train with RL using the preference model as the reward signal, i.e. we use ‘RL from AI Feedback’ (RLAIF).”

The end result allows Claude to engage with harmful, unpleasant, and even malicious queries made by users by “explaining its objections to them.” So Claude can provide honest answers without producing any harmful output. That’s a very attractive feature to offer companies who want to use AI, but are concerned as to how it can be abused and potentially damage their brand or reputation.

Anthropic is offering two versions of its AI assistant today called Claude and Claude Instant. Claude is the state-of-the-art high performance model that can become “a delightful company representative, a research assistant, a creative partner, a task automator” and more, with the ability to customize personality, tone, and behavior. Claude Instant is simply lighter, less expensive, and much faster—optimized for low latency, high throughput use cases. Pricing for both is per million characters(Opens in a new window).

Claude is already being used by Quora in its Poe AI Chat app, by online education company Juni Learning to power a Discord tutor bot, by Notion to boost user productivity, by DuckDuckGo for DuckAssist, and by Robin AI to evaluate parts of legal contracts.

The AI assistant continues to be developed and improved, and Anthropic is planning to roll out updates in the coming weeks. Only time will tell how harmless the AI assistant turns out to be, as users will no doubt test it throughly just like they have with ChatGPT.