DeepSeek, the former domestic light, seems to have been a little cool lately A few months ago, it was hailed as a "disruptor" to challenge the giants of Silicon Valley, making huge waves with its very low cost, and it was in the limelight However, this craze comes and goes faster, in fact, the reason is very simple, that is, "unreliable"! An audit report by NewsGuard: DeepSeek has a staggering 83% accuracy "failure rate" when answering key information This means that the information it produces is false, misleading, or even fabricated out of thin air This is the classic "AI hallucination" DeepSeek actually reminds the entire AI community that the most important thing is to be smart, cheap, useless, and trustworthy That's why I've been following the @Mira_Network project lately Mira's idea was to do something more basic: to be a "disciplinary committee" + "fact-checking team" for all AI When any AI wants to give an answer, Mira will first break the answer into several key points, and then immediately scatter it to thousands of different small AIs in the network, so that they can independently verify it "back to back", and finally the minority obeys the majority Only the fact that the vast majority of AIs have nodded their heads to confirm will be released. Reduce the likelihood that the AI will be talking serious nonsense In the past, everyone was comparing who was "smarter". Now, DeepSeek uses its own "death" to tell us that it is useless to be smart, and reliability is the most important thing! #ai #mira @MiraNetworkCN #kaito #Yaps