4 ms·

If they say (development of) AGI is as dangerous as they say it is, it's on a level of WMD. And here you have unstable people and company working on it. Shouldn

• petre 3y ago
  If its proven to be dangerous, Congress with quickly regulate it. It's probably not that dangerous and all the attempts to picture it that way are likely fueled by greed, so that it's regulated out of small players' reach and subject it to export controls. The real threat is that big tech is going to control the most advanced AIs (already happening, MS is throwing billions at it) and everyone else will pay up to use the tech while also relinquishing control over their data and means of computation. It has happened with everything else becoming centralized: money, the Internet and basically most of your data.
• ilaksh 3y ago
  First of all, for many people, AGI just means general purpose rather than specific purpose AI. So there is a strong argument to make that that has been achieved with some of the the models. For other people, it's about how close it is to human performance and human diversity of tasks. In that case, at least GPT-4 is pretty close. There are clearly some types of things that it can't do even as well as your dog at the moment, but the list of those things has been shrinking with every release. If by AGI you mean, creating a fully alive digital simulation/emulation of human, I will give you that, it's probably not on that path. If you are incorrectly equating AGI and superintelligence, ASI is not the same thing.