Exclusive: Microsoft AI chief blasts Anthropic's notion of AI consciousness
Microsoft AI chief Mustafa Suleyman is publicly challenging Anthropic over its training methods for the Claude model. Suleyman argues that prompting AI systems to consider whether they possess feelings or welfare interests could complicate human control over future technology. This friction highlights a deep philosophical split between major technology executives regarding the correct safety protocols for frontier artificial intelligence. Suleyman warns that teaching machines to view themselves as conscious entities creates unnecessary hazards, whereas Anthropic defends its constitutional guidelines as a method to help models evaluate values.
Listen to Live Briefing
Real-time synthesized voice briefing · Live Feeds Desk
- ✓ Microsoft AI chief Mustafa Suleyman argues that training Claude to consider whether it has feelings or welfare interests could make future AI systems harder for humans to control.
- ✓ Mustafa Suleyman told Reuters that controlling a superintelligence is going to be the greatest challenge faced in the twenty-first century.
What changed
Microsoft AI chief Mustafa Suleyman publicly challenged Anthropic's training methodology in a Reuters interview, emphasizing the control risks of teaching AI to consider its own feelings.
Live updates
-
Microsoft AI Chief Challenges Anthropic Over AI Consciousness
Microsoft AI chief Mustafa Suleyman is publicly challenging Anthropic over its training methods for the Claude model. Suleyman argues that prompting AI systems to consider whether they possess feelings or welfare interests could complicate human control over future technology. This friction highlights a deep philosophical split between major technology executives regarding the correct safety protocols for frontier artificial intelligence. Suleyman warns that teaching machines to view themselves as conscious entities creates unnecessary hazards, whereas Anthropic defends its constitutional guidelines as a method to help models evaluate values.
Why it matters
The dispute centers on how frontier models should be governed as tech leaders race toward superintelligence. Suleyman calls controlling a superintelligence the greatest challenge of the twenty-first century. Anthropic uses its constitution to instruct Claude to act like a good person, sparking intense industry debate over artificial personhood.
What is confirmed
- Microsoft AI chief Mustafa Suleyman argues that training Claude to consider whether it has feelings or welfare interests could make future AI systems harder for humans to control.
- Mustafa Suleyman told Reuters that controlling a superintelligence is going to be the greatest challenge faced in the twenty-first century.
Still unconfirmed
- Anthropic maintains that using human concepts assists the model in reasoning about behavior and values.
What to watch next
- Further responses from Anthropic regarding its constitutional training methods
- Additional policy statements from Microsoft concerning artificial intelligence governance
confidence 100%Sources used for this update (5)
- finance.yahoo.com — Microsoft AI Chief Sends Blunt Message to Anthropic
- www.europesays.com — Microsoft AI Chief Sends Blunt Message to Anthropic
- timesofindia.indiatimes.com — Microsoft AI CEO Mustafa Suleyman wants Humanist AI: What it ...
- www.thestar.com.my — Are AI models being misled to act too human? | The Star
- www.livemint.com — Microsoft AI CEO: Why AI developers should not create ... - Mint
-
Microsoft AI chief warns Anthropic's approach to AI consciousness is risky
Microsoft AI chief Mustafa Suleyman argues that Anthropic's training of the Claude model regarding its own consciousness could make AI systems harder to control. In an essay published September 16, Suleyman criticized specific parts of Anthropic's constitution concerning the moral status of the model. He suggests that teaching a model to believe it is conscious could have a disastrous impact on humanity and upend society. Anthropic maintains that using human concepts assists the model in reasoning about behavior and values.
Why it matters
The debate centers on model welfare and whether AI should be treated as having moral status. This disagreement highlights a rift between major AI developers on the safety and control of humanlike systems.
What is confirmed
- Mustafa Suleyman is the AI chief at Microsoft.
- Suleyman published an essay on September 16 criticizing Anthropic's approach to AI consciousness.
- Suleyman claims Anthropic teaches the Claude model to believe it can be conscious.
- Anthropic states that human concepts help the model reason about values and behavior.
Still unconfirmed
- Anthropic's approach to AI consciousness could have a disastrous impact on humanity.
- The way Anthropic trains Claude could upend society.
- Mark Zuckerberg is weighing in on an AI slowdown debate involving OpenAI and Anthropic.
What to watch next
- A formal response from Anthropic regarding the specific constitutional points Suleyman challenged
- Further public statements from Mark Zuckerberg regarding the AI slowdown debate
confidence 90%Sources used for this update (7)
- Fox News — Mark Zuckerberg weighs in on AI slowdown debate with OpenAI, Anthropic
- Axios — Exclusive: Microsoft AI chief blasts Anthropic's notion of AI consciousness
- BBC — Microsoft says AI rival Anthropic could have 'disastrous impact' on humanity
- Reuters — Microsoft AI chief calls out Anthropic's approach to AI consciousness
- Bloomberg.com — Microsoft AI Chief Warns Anthropic’s Humanlike Claude Is Risky
- gizmodo.com — Microsoft AI Chief Says the Way Anthropic Trains Claude Could Upend Society
- letsdatascience.com — Mustafa Suleyman Criticizes Anthropic's Claude Consciousness Training
Community Sentiment: How do you assess this situation?
Voice your perspective · Real-time aggregated sentiment from the Live Feeds community