different models.
high level approach:
N models receive the same prompt → responses are consolidated through a simple embedding-based similarity matrix → another model performs a simple assessment of deduplicated responses → these are then consolidated into a research report & data visualization.
the whole contraption uses ~10 different models, but more can easily be plugged into the initial generation phase. happy to demo it sometime! [edit: email on profile].
Plenty of apps are available in English outside the anglosphere. Seems like a US-centric oversight not releasing it to “the rest of the world” as usual.
How do we know which institutions to trust in safely deploying AGI? From my POV some kind of autonomous and generalized intelligence is inevitable and imminent - but who can be trusted to deliver something that works for the majority? Or is that a pipe dream?
(Shameless plug) I recently started my own newsletter to help a non-technical audience understand the big picture trends driving AI and AGI: https://newsletter.envisioning.io
> The main difference between the prototype shown today and what will be going to the moon is that the ones going to the moon will be white instead of dark. “That’s really for thermal reasons,” Mr. Ralston said.
the whole contraption uses ~10 different models, but more can easily be plugged into the initial generation phase. happy to demo it sometime! [edit: email on profile].