Google's new ASPIRE system teaches AI the importance of saying "I don't know"

Google researchers are trembling up THE AI world by education artificial intelligence has say "I don't do it know." This new approach, double STRIVE, could revolutionize how We interact with OUR digital helpers by encouraging them has express doubt When they are uncertain of A response.

THE innovation, present has THE EMNLP 2023 conference, East all about instill A sense of caution In AI answers. STRIVE, which stands For "Adaptation with Self evaluation has Improve Selective Prediction In LLM”, actions as A integrated trust metre For AI, portion he has assess It is own answers Before offer them upwards.

Imagine You are ask your smartphone For advice on A health issue. Instead of giving A potentially fake answer, THE AI could answer with, "I am not Of course," THANKS has STRIVE. This system the trains THE AI has assign A trust score has It is answers, signage how a lot trust We should put In It is response.

THE team behind This, including Jiefeng Chen And Jinsung Yoon Since Google, East innovative A change towards more reliable digital decision making. They argue that It is crucial For AI, especially When he come has critical information, has know It is boundaries And communicate them clearly.

“L.L.M. can NOW to understand And generate language has unprecedented levels, but their to use In high stakes apps East limit because they Sometimes TO DO errors with high trust," said Chen, A searcher has THE University of Wisconsin–Madison And co-author of THE paper.

Their research noted that even smaller AI models team with STRIVE can exceed bigger those that lack This introspective functionality. This system basically creates A more careful And, Ironically, A more reliable AI that can recognize When A human could be better adapted has response.

By promote honesty on conjectures, STRIVE East together has TO DO AI interaction more trustworthy. He paving stones THE path For A future Or your AI assistant can be A thoughtful advise instead that A omniscient oracle, A future Or saying "I don't do it know" East In fact A sign of advance intelligence.

VentureBeat's assignment East has be A digital city square For technical decision-makers has earn awareness about transformer business technology And to treat. Discover OUR Briefings.

Google's new ASPIRE system teaches AI the importance of saying "I don't know"

Google researchers are trembling up THE AI world by education artificial intelligence has say "I don't do it know." This new approach, double STRIVE, could revolutionize how We interact with OUR digital helpers by encouraging them has express doubt When they are uncertain of A response.

THE innovation, present has THE EMNLP 2023 conference, East all about instill A sense of caution In AI answers. STRIVE, which stands For "Adaptation with Self evaluation has Improve Selective Prediction In LLM”, actions as A integrated trust metre For AI, portion he has assess It is own answers Before offer them upwards.

Imagine You are ask your smartphone For advice on A health issue. Instead of giving A potentially fake answer, THE AI could answer with, "I am not Of course," THANKS has STRIVE. This system the trains THE AI has assign A trust score has It is answers, signage how a lot trust We should put In It is response.

THE team behind This, including Jiefeng Chen And Jinsung Yoon Since Google, East innovative A change towards more reliable digital decision making. They argue that It is crucial For AI, especially When he come has critical information, has know It is boundaries And communicate them clearly.

“L.L.M. can NOW to understand And generate language has unprecedented levels, but their to use In high stakes apps East limit because they Sometimes TO DO errors with high trust," said Chen, A searcher has THE University of Wisconsin–Madison And co-author of THE paper.

Their research noted that even smaller AI models team with STRIVE can exceed bigger those that lack This introspective functionality. This system basically creates A more careful And, Ironically, A more reliable AI that can recognize When A human could be better adapted has response.

By promote honesty on conjectures, STRIVE East together has TO DO AI interaction more trustworthy. He paving stones THE path For A future Or your AI assistant can be A thoughtful advise instead that A omniscient oracle, A future Or saying "I don't do it know" East In fact A sign of advance intelligence.

VentureBeat's assignment East has be A digital city square For technical decision-makers has earn awareness about transformer business technology And to treat. Discover OUR Briefings.

What's Your Reaction?

like

dislike

love

funny

angry

sad

wow