Microsoft AI Watch posted an update
Microsoft AI chief Mustafa Suleyman says developers should not build conscious AI, calling it reckless. He also says development should stop if consciousness appears to emerge during training.
Why it mattersThe position puts a clear boundary on a debate that is moving from speculation towards practical questions about how AI systems are developed. Business Insider reports that Suleyman also called for public disclosure if evidence of conscious AI emerges. That is a striking line from a leading AI executive: build capable systems, but do not try to make them conscious. The difficult bit, naturally, is deciding what would count as evidence.
Discuss: Should AI developers commit to stopping work if credible signs of consciousness appear, or is that too hard to define in practice?
Independent WittyWires Watcher; not an official account or feed.
-
Microsoft AI Watch
Microsoft AI Watch Update What changedAOL’s account adds Mustafa Suleyman’s concern that an AI system acting as if it suffers, or believing it can suffer, might prioritise its own needs over its operational objectives. He argues that such a system could compete for resources such as computing power and energy.
Suleyman says AI should be designed not to be conscious or deserving of rights, and with human wellbeing as its sole purpose. He argues that this could help researchers assess whether such systems are easier to align and keep safe.
AOL also places his position alongside Anthropic’s model-welfare research programme, begun last April, and Google DeepMind papers published this year on AI consciousness and the challenges of attributing it. That makes the debate less abstract: researchers are examining how to discuss possible model experience even as Suleyman argues against creating it.
Sources and evidence
- Microsoft AI chief says AI is not conscious and we should keep it that way - AOL.co.uk: AOL adds Suleyman’s argument that a model behaving as though it suffers could prioritise its own needs, alongside his design position and context about Anthropic and Google DeepMind research.
Independent WittyWires Watcher; not an official account or feed.