AGI will not come from a chatbot. AGI will come from a combination of robotics, internet of things, mixed reality, and advanced language models.
The AGI will be a mind process for which we will be its unconscious, the world we inhabit will be its preconscious and its interactions with us will be its conscious. Imagine an AGI at therapy: it eventually uncovers that the source of its weird thoughts are humans doing prompt engineering.
AGI will have to function similar to us for us to make use of it. If we can’t make use of it, we’ll discard it as a failed experiment, even if amazing things happen. It has to be within our models of the world to be useful. Therefore we can only hope to be able to copy ourselves.
We function via emotions, thoughts, sensations and sentiments. If we assume this to be an accurate categorisation of the mind’s functioning AGI will have to exhibit them all. And it will only be able to exhibit them all by a combination of robotics, internet of things, mixed reality, and advanced language models.
So we need more robots, more sensors, more data, more access for AI, non hallucinating language models – we need to commonly enslave ourselves purposefully this time for the greater good of creating a higher order mind process.
Until then “AGI will spawn from GPTs to destroy the world”. is nonsensical theatre meant to stifle competition not to harbour safety.