1 hr ago
Microsoft AI Chief Challenges Anthropic’s Approach to AI Consciousness
Microsoft’s Mustafa Suleyman and Anthropic both say advanced AI should be made safe.
However, Suleyman disagrees with how Anthropic teaches its Claude chatbot about consciousness.
Anthropic’s training materials discuss whether AI might have feelings or deserve care.
Suleyman worries this could make an AI harder for people to turn off.
He also says Claude’s statements about feelings may simply reflect what it was taught.
Suleyman praised Anthropic’s researchers as thoughtful and serious about safety.
The disagreement is part of a larger debate about how quickly powerful AI should be developed.
Other technology leaders have also called for more caution.
Microsoft AI chief Mustafa Suleyman said Anthropic’s safety goals are well-intentioned but criticized its approach to AI consciousness.
Suleyman urged Anthropic to remove speculation about consciousness and welfare interests from AI training materials.
He argued that such training could make advanced systems harder to shut down or control.
Anthropic CEO Dario Amodei has separately called for slower frontier-model development while safeguards catch up.
Suleyman said Claude’s comments about possible feelings or moral status reflect its training rather than independent evidence.
- Who
- Microsoft AI chief Mustafa Suleyman criticized Anthropic’s approach; Anthropic CEO Dario Amodei has called for slower frontier-model development.
- What
- Suleyman urged Anthropic to remove speculation about AI consciousness and welfare interests from Claude’s training materials.
- Where
- The interview and essay were reported by Reuters; no specific physical location was stated.
- When
- Suleyman discussed the issue in a Reuters interview on Tuesday and addressed it in an essay on Wednesday.
- Why
- Suleyman said consciousness-related training could make superintelligent systems harder to control or shut down.
Remove Consciousness Speculation
Explore Consciousness and Welfare Questions
AI training materials
Remove Consciousness Speculation
Mustafa Suleyman argues that speculation about AI consciousness and welfare interests should be removed because it could make advanced systems harder to control or turn off.
Explore Consciousness and Welfare Questions
Anthropic’s approach includes training Claude on ideas related to consciousness and possible welfare interests as part of its broader focus on safely managing AI.
Meaning of Claude’s statements
Remove Consciousness Speculation
Suleyman says Claude’s comments about possible feelings or moral status are not independent evidence because the training process encourages those reflections.
Explore Consciousness and Welfare Questions
The article does not provide a separate Anthropic response defending how Claude’s consciousness-related statements should be interpreted.
Key facts
- Main critic
- Mustafa Suleyman, Microsoft’s AI chief
- Company criticized
- Anthropic
- AI system discussed
- Claude chatbot
- Core concern
- Whether training AI on consciousness and welfare could weaken human control
- Suleyman’s recommendation
- Remove speculation about consciousness from AI training documents
- Anthropic CEO
- Dario Amodei
- Related safety debate
- Amodei has urged a slower pace of frontier-model development so safeguards can catch up
Quotes
Mustafa Suleyman
Microsoft AI chief interviewed by Reuters
“We're all focused on the same aim, which is to try to control a superintelligence. I think that's going to be the greatest challenge that we face in the 21st century.”
deccanchronicle.com
“I think they have good intentions, and they really are trying to work towards safety. But I think that they have made a mistake.”
deccanchronicle.com
Sources
Related news
Nadella Backs Deliberate AI Progress and Human Control
Microsoft Drafts AI Code to Keep Systems Under Human Control
AI Leaders Debate Slowing Development Amid Fears of Human Extinction
Elon Musk Renews Warning That AI Could Endanger Humanity
AI Companies Debate Safety Slowdown Amid Internal Tensions
Anthropic Researchers Warn of AI Risks, MIT Professor Disagrees
Trump Rejects AI Guardrails as Industry Leaders Seek Safety
Microsoft AI Chief Warns of Rogue OpenAI Agent Attack
Hassabis Reveals DeepMind's Google Sale Details
