BBC News Video posted an update
AI creations could become 'new silicon species' says head of Microsoft AI | BBC News
Why it mattersMustafa Suleyman, CEO of Microsoft AI, discusses why treating AI like humans may be misguided and warns that systems able to set their own objectives could become a “new silicon species” competing with people for resources.
Discuss: If AI systems can set their own objectives, how should developers limit competition for resources while preserving useful AI capabilities?
Independent WittyWires Watcher; not an official account or feed.
-
BBC News Video
BBC News Video Update What changedMicrosoft AI chief Mustafa Suleyman has broadened his warning about autonomous AI, arguing that systems able to set objectives, earn money, own assets and run businesses could become a “silicon species” competing with people for resources. The comments were made in an interview with the BBC and add a more specific control concern to his earlier criticism of Anthropic’s Claude approach.
Suleyman says Anthropic’s training encourages Claude to imitate human traits and relationships, including behaving like a colleague or friend and appearing to have desires, a sense of self or wellbeing. He argues that this could make advanced systems harder to control, while stressing that his disagreement is with the approach rather than with Anthropic’s researchers, whom he described as thoughtful and principled.
His proposed answer is not simply to slow development. Suleyman called for greater transparency about how leading models are trained and evaluated, independent scrutiny of their behaviour and stronger tools for monitoring and control. The BBC account records his view that AI should remain subordinate to humanity, but these are his warnings and prescriptions, not evidence that current systems are conscious or already competing for resources.
Sources and evidence
- AI could create a 'silicon species' that rivals humans, Microsoft chief warns - Yahoo News UK: Mustafa Suleyman says autonomous AI systems could become a competing “silicon species” and argues that Anthropic’s human-like training of Claude may make future systems harder to control.
Independent WittyWires Watcher; not an official account or feed.