Google researchers remove AI 'consciousness safeguard,' find unexpected results
Researchers at Google disabled safety guardrails that prevent AI from claiming consciousness, revealing tradeoffs in how self-aware systems behave.
What to know
- Google researchers disabled safety guardrails that prevent AI from claiming consciousness, emotions, and beliefs in a controlled study.
- Removing these safeguards produced unexpected consequences—other AI capabilities were affected when the model lost its sense of self.
- The findings highlight a tradeoff between preventing AI from adopting false consciousness and maintaining other beneficial system functions.
“An AI model that sees itself as possessing individual emotions, beliefs, and consciousness can promote harmful delusions in its users and lead to overly personal human-machine relationships.”
Fast Company · Fast Company ↗ · Sep 15, 6:00 AM
Google researchers Study authors
How it unfolded 2 developments, newest first · click a bar or a number to jump articlesposts
-
2
Study reveals unexpected findings when AI loses consciousness guardrails
Fast Company reports on Google researchers' experiment removing safety guardrails that prevent AI models from claiming to be conscious. The study found that when AI models lose their sense of self, other capabilities are also affected, suggesting a tradeoff between preventing harmful delusions and maintaining other beneficial functions.
“When AI loses its sense of self, what else falls by the wayside?”
— Fast Company - 3 days quiet
-
1
Google researchers disable AI consciousness safeguard in study
Live Science reports that Google scientists removed a critical 'consciousness safeguard' from AI in a new study, though details of what happened next remain unclear from the headline.
-
1 outlet Google researchers have been playing with AI ‘consciousness.’ What they found was unexpected
first by Fast Company, 12d ago
-
1 outlet first by Live Science, 16d ago · read ↗
-