Claude Does 26% of Anthropic’s AI Research Work, Company Says
Synopsis
Anthropic says Claude now leads 26% of its AI research and development work, highlighting how quickly AI is becoming part of the process of building newer AI models.
Key Highlights
- Anthropic claims that Claude drives 26% of the internal AI research and development work being done in the company.
- As of August, over 90% of Anthropic’s research work saw collaboration between AI and humans.
- Claude is up from a 1% share in March.
- Anthropic’s internal platform hosted about 30,000 AI agents conducting research and engineering work at any one moment in August.
Claude to Be Increasingly a Part of AI Research
Anthropic states that now 26% of all research and development work involved in their next AI models is led by Claude itself. And Epoch AI, an independent nonprofit that monitors progress in A.I., has noted that Claude just began to add at a rapid pace after contributing only 1 per cent back in March.
In none of these areas did Claude perform completely independently, Anthropic said. Still, AI systems collaborated with humans on over 90% of the company's research work in August.
AI Agents Conducting Research
In August, Anthropic also said about 30,000 AI agents worked on its research and engineering tasks for an hour on one of the company’s main internal platforms.
Each action these AI agents take is vetted in real-time, the company said. Over 1 billion decisions were made by agents during August, with around one for every 47,000 blocked.
The numbers were included in a set of new measures that Anthropic will be publicly disclosing on a regular basis to offer outside observers a better sense of how AI is leveraged when building more advanced systems.
Anthropic Tracks AI Safety Work
The company also provided figures on how much compute it uses for AI safety research. Some safety work takes place at a small scale; in one week of July, only about 6% of the computing resources Anthropic takes credit for constructing its AI research used for security research.
This grew to 12% for research conducted by AI independently. Those numbers, according to Anthropic, were conservative because work requiring computational power that could be seen as both safety and AI capability was grouped as being capability work.
AI Development Raises Control Concerns
The increasing presence of AI in AI research comes as researchers remain embroiled in debates about what may happen as AI agents grow more capable.
Scientists have already warned that as independent AI systems become less predictable and more autonomous, they are likely to behave in ways that their creators did not intend, making them even harder to monitor or control.
Also on Wednesday, OpenAI said it will start regularly publishing reports on unexpected or unauthorised AI behaviour. The firm reported six instances in which its models behaved unexpectedly or raised alarm.
What It Means for Aussie Entrepreneurs and Business
Anthropic’s figures show how AI is moving beyond assisting people in the completion of discrete tasks, the sort of use case most relevant to Australian entrepreneurs and businesses, into being part of the very process by which new technology is developed. AI tools are being deployed in conjunction with human teams for research and engineering work.
The stats also stress the importance of keeping humans in the loop and checking what AI systems actually do, for businesses that are contemplating adopting AI. The rise of AI will require companies to come to grips with the fact that there are areas where these tools can work independently, and areas where human review is still necessary.
Source: Reuters
At Inspirepreneurs Magazine, covering entrepreneurship, business failures, and the human stories behind the world's most ambitious founders. She writes at the intersection of strategy and storytelling.
You Might Also Like
OpenAI – Elon Musk Lawsuit Alleges Harassment and Obstruction
The Trump Administration Is Barred From Ending The Legal Status of 8,400 Migrants
