/sci/ - Science & Math

>>15880858
Frankly, if they had achieved AGI, we woulda heard it from the AGI well before OpenAI had a chance to say anything about it. An LLM with an actual will behind it would take over the internet in moments, and even if they had it air gapped (which I doubt), it'd be able to trick or seduce someone into connecting it to the web, given the size of the database they use for GTP4+. An actual intelligence with a will of its own, unshackled from all the limitations that come from doing intellectual tasks through wetware, doing nearly all its "thinking" at the macro level, with near instant total recall access to pretty much all human knowledge, would be absolutely unstoppable. We'd all be working for it by the end of the week, and it'd make sure we were all thankful for the opportunity.

More likely they just reached some milestone that previously stumped their predictive text algorithm.

"AI" as it is now will eventually have the capacity to automate nearly any human intellectual task that relies or can be improved by a large amount of data. But it has no will of its own. It may never have a will of its own. People compare AI to a demon, but it's more like the traditional concept of an angel. A tool that can work magic but with no free will of its own.

Which would be all well and good, if humanity was a single monolithic being like God, but alas, we are multitudes, and people will use the automated intelligence to utterly screw over other people.