Error rate for American artificial intelligence products: 62 percent
A few weeks ago when DeepSeek, a low-cost Chinese alternative to ChatGPT and other U.S. large language models, was released, its critics warned that DeepSeek's fail rate for people seeking reliable news and information about contentious issues was 83 percent.Welcome to the party!The comparable error rate for American artificial intelligence products: 62 percent.
This isn't a technical revolution. It's an embarrassment. Those are ridiculously high rates of disinformation and misinformation for systems that − if you believe the media hype − are poised to revolutionize our lives.
No comments:
Post a Comment