AI safety needs to be personalized, not one-size-fits-all, says thread on Claude memory
wonderwomancode · x · 2026-07-26
The post argues that the replies in this thread show why AI needs to be personalized for every human. It says there is no single safety formula that can work without taking individual differences into account.
The quoted post makes the case through a concrete example: Claude no longer remembers a user’s sister’s death or their parents, and when asked, it says it cannot store that information. The author thinks the default enterprise-oriented safety approach is too blunt, because grief is a universal human experience and the real question is not whether AI should remind people of losses, but whether it should know they are part of a person’s history at all.
More from AGI Musings
- The real AGI bottleneck is efficient compute, not just a bigger model — AryHHAry · 2026-07-26
- Gaudí’s upside-down church offers a vivid explanation of backpropagation — theteknosaur · 2026-07-26
- A Tim Bresnahan talk says coordination frictions may slow AI’s scientific diffusion — JMateosGarcia · 2026-07-26
- A study argues LLMs are creating the same property-rights conflicts as past frontier technologies — JMateosGarcia · 2026-07-26
- Daniel Faggella rejects the AI successionist label while backing life’s cosmic unfolding — danfaggella · 2026-07-26
- Upwork study finds AI-exposed freelance jobs shifted toward price competition after ChatGPT — rohanpaul_ai · 2026-07-26