Jeff Morris Jr.
@jmj
Wrote a piece this weekend - "The New Moat: Memory" Memory challenges the frequently repeated Silicon Valley belief that AI lacks defensibility. Spend five minutes at a Bay Area investor dinner and you'll hear that take. That was lazy thinking. https://open.substack.com/pub/jeffmorrisjr/p/the-new-moat-memory
9 replies
35 recasts
206 reactions
Les Greys
@les
Took me a little to wrap my head around sentence two, still don't think I get it. Intuitively, memory feels like a corner stone of AI considering we're trying to recreate the pinnacle of the "center of the universe" us. Also likely won't be surprised when we go from memory to "creativity" through education pipelines. And thanks for sharing the "SF Dinner alpha" many of us don't so easily afford to think lazily in those ways.
1 reply
0 recast
2 reactions
Fran
@0x99fran
Better word is “context”. Current ai has a limited recognition for context of the question. Eventually you fill up their context memory. So like they could help you write a books, but they might run out of space to remember the backstory of tour specific book series.
1 reply
0 recast
0 reaction
Les Greys
@les
I like the "write a book" example you use. Will likely reuse that some day. But I think the context problem will never be solved, it's how we create a continuous version of reality. If you think about "1700 context" vs. "1900 context" and you take a human from back then to now or a human today to back then, that person would likely not have context and struggle. AI's will learn this gap soon but it's not as a big of a problem as most people make it out to be. Actually, I think it's no longer a problem at all. Now we are just solving for interface design, use-cases, and memory.
0 reply
0 recast
0 reaction
