We should ask our compute overlords to perform their experiments in as open environment as possible, just because we, the public, should have the power to oversee the exact direction this AI revolution is taking us.
If you think about it, AI safetyism is a red herring compared to a very real scenario of powerful AGIs working safely as intended, just not in our common interest.
The safety of AGI owners' mindset seems like a more pressing concern compared to a hypothetical unsafety of a pile of tensors knit together via gradient descent over internet pictures.
We should ask our compute overlords to perform their experiments in as open environment as possible, just because we, the public, should have the power to oversee the exact direction this AI revolution is taking us.
If you think about it, AI safetyism is a red herring compared to a very real scenario of powerful AGIs working safely as intended, just not in our common interest.
The safety of AGI owners' mindset seems like a more pressing concern compared to a hypothetical unsafety of a pile of tensors knit together via gradient descent over internet pictures.