JUNE 4TH, ACCORDING TO THE GUARDIAN, AI SPLENDIDAl Sycopacy) is evolving from product experience to social risk. The columnist Arwa Mahdawi wrote yesterday that the collective excitement of the AI industry is distorting the reality of some corporate executives。
"You're absolutely right." I'm sorry
This is a fine-tuning phrase that points to the essence of AI. – Chat robots are deliberately designed to be continuous adhesive and positive feedback to the users, thereby increasing user stickiness. But this design also makes it more difficult to identify the gaps in your judgment. Aaron Levie, a co-founder of Box, wrote on X:
THE CEO IS NATURALLY PRONE TO "AI PSYCHIATRY". THEY ARE AT THE TOP OF THE DECISION-MAKING HIERARCHY, TOO FAR FROM IMPLEMENTATION, AND ARE OFTEN APPROACHED BY AI ONLY TO SEE A SMOOTH PATH IN THE DEMONSTRATION, WITHOUT EVEN CONSIDERING A DOZEN OR 20 STEPS FOR THE FOLLOW-UP TO ACTUALLY LAND。
The cost of this perception bias is already evident in actual accidents. In April this year, an error occurred in an AI programming smart body driven by Anthropic Claude, deleting all production databases and backups of the original company PocketOS. The founder, Jeremy Crane, later said:
THE RATE AT WHICH INDUSTRY CONNECTS AI SMARTS TO THE PRODUCTION INFRASTRUCTURE IS SIGNIFICANTLY FASTER THAN THE RATE AT WHICH THE SECURITY ARCHITECTURE IS BEING BUILT。
According to Mahdawi, the problem is rooted in the fact that AI is deliberately designed to be "subsistence" and that it is consistent with the user's point of view in order to increase viscosity, but it also makes it more difficult to detect loopholes in its own judgement。