Previously on Less Wrong:
Steve Byrnes wrote a couple of posts exploring this idea of AGI via self-supervised, predictive models minimizing loss over giant, human-generated datasets:
Self-Supervised Learning and AGI Safety
Self-supervised learning & manipulative predictions
Previously on Less Wrong:
Steve Byrnes wrote a couple of posts exploring this idea of AGI via self-supervised, predictive models minimizing loss over giant, human-generated datasets:
Self-Supervised Learning and AGI Safety
Self-supervised learning & manipulative predictions