That's not exactly the same thing as the LLM photographic memory that we launched at Lamini a month ago - https://www.lamini.ai/blog/lamini-llm-photographic-memory-ev...

We focused on LLMs that can reason while still remembering key facts exactly.

But I like that the concept is catching on in the AI community. Go Satya.

LLMs are powerful, but unlike humans, their weights are fixed during inference and cannot be used for working memory.

Computer memories are abundant and perfect. It's not unreasonable to store your entire life in all modalities in a personal computer of the near future.

AIs with that ability should exceed human performance on knowledge tasks.