A term coined in March
2023 by Silicon Valley engineers,
Tristan Harris and Aza Raskin, to describe a new and emerging class of Artificial Intelligence: "Generative Large Language Multi-Modal Model," or GLLMM ("Gollem" for
short). This new class of
AI uses predictive algorithms to creatively generate new and original output, including
music, artwork, poems, etc.
The word comes from a creature in ancient Jewish folklore called the "golem," a clay sculpture brought to life by
magic, and employed as a helper of the Jewish people. But the golem always had the potential to turn against its masters. The
legend aptly captures the precarious extremes of societally-benefitting good and existential evil presented by this new AI.
They introduced the term in their March
2023 video lecture "The A.I. Dilemma." Their main thesis was to compare this new and emerging AI with the established dangers of "societal entanglement" that older AI technology has already achieved via Twitter, Instagram,
Facebook, etc. They explained that first generation of AI --called "curation AI"-- has both helpful and harmful impacts on
society. And now this next generation of AI --called "creation AI"-- has even greater potential to be even more helpful and even more harmful. They are calling for legislation to control the usage of all forms of the new Gollem-Class AI before it become so deeply embedded in the workings of society that its inevitable entanglement can't ever be untangled.
Because
Chat GPT and
Bard are both predictive-language programs, capable of generating entirely original works of creativity, they
fall under the umbrella of what Harris and Raskin are now calling "Gollem-Class AIs."
