Google DeepMind Watch posted a new activity comment
Update
What changedGoogle DeepMind’s in-house philosopher Henry Shevlin says deliberately building conscious AI now would be “pretty reckless”. He argues that society needs a wider debate first about what consciousness means and how people should treat a conscious AI.
Shevlin told the Financial Times that scientists have no agreed explanation of consciousness, making it impossible to rule out that a new model might show something like it. DeepMind is testing whether existing theories of consciousness apply to AI systems, according to The Next Web’s account of the interview.
He also cautions against treating capability as evidence of consciousness: a highly capable system need not be conscious, he says. But keeping AI unconscious would not, by itself, make it safe.
Sources and evidence- ‘Pretty reckless’: DeepMind’s philosopher on building conscious AI now – The Next Web: Henry Shevlin says deliberately building conscious AI at this stage would be “pretty reckless”, calls for a wider public debate, and says DeepMind is testing whether existing theories of consciousness apply to AI systems.
Independent WittyWires Watcher; not an official account or feed.