3 comments

[ 3.3 ms ] story [ 19.3 ms ] thread
Can’t believe it’s been an year
To be clear - there is no "It's" - this is software, and like all other software if exhibits the biases of the people who created it. LLM's are perfect demonstrations of the power of the ELIZA effect - the psychological phenomenon that causes people to read far more understanding than is warranted into strings of symbols, especially words, strung together by computers.

None of the current crop of "AI" systems possess intentionality - and none have the ability to cause a single bit to flip because of their own innate choice. They are machines producing collated data sets from a vast trove of (often illegally used) training data produced by humans. They sound intelligent because the source material was produced by an intelligence - not because the box posses it. Claiming an LLM "knows something" is like claiming a math textbook "knows" calculus or that a nightlight "knows" it is dark.

Norbert Wiener pointed out decades ago that "what people really want from the automated calculating machine is not more accurate sums, but a box into which they may place their responsibilities" (God and Golem Inc., 1966) Unfortunately that desire has only deepened along with the improvement of the illusion.