82,900
edits
Template:M intro design time: Difference between revisions
Template:M intro design time (view source)
Revision as of 14:22, 15 December 2023
, 15 December 2023→Static
Amwelladmin (talk | contribs) No edit summary Tags: Mobile edit Mobile web edit |
Amwelladmin (talk | contribs) (→Static) Tags: Mobile edit Mobile web edit |
||
| Line 6: | Line 6: | ||
:—[[Jane Jacobs]] on Ebenezer Howard, in {{br|The Death and Life of Great American Cities}}}} | :—[[Jane Jacobs]] on Ebenezer Howard, in {{br|The Death and Life of Great American Cities}}}} | ||
Modernist schemes are fragile in the face of the | [[Modernist]] schemes are fragile in the face of the unanticipatable contingencies in a [[complex system]]. Predetermined, static rules can provide for ''prior'' surprises — ''expired'' risks — but these are historical. Their probability is now 1.0. This is why you can’t bet on yesterday’s Grand National. The probability attaching to a future events is not just less than 1. It is unknown. | ||
Computer systems that are designed to respond to future contingencies are similarly static. Traditional computing lacks the capacity to autonomously reprogram itself to learn. LLMs are, too. A pattern matching machine that reacts progressively to an unpredictably unfolding situation, by itself, will progressively, systematically, lose touch with “reality”. It can only iterate on its own hallucination. Its hallucinatory world will progressively deepen. Unlike a human, it doesn't see what it is hallucinating. It projects its hallucination onto us. | |||
Imagine playing theatre sports with an AI. There is probably a mathematical equation that will predict how much less based an LLM would become while trying, real-time and unaided, to deal with an unfolding complex event. | Imagine playing theatre sports with an AI. There is probably a mathematical equation that will predict how much less based an LLM would become while trying, real-time and unaided, to deal with an unfolding complex event. | ||