This is perhaps absurd/ off/ wrong/ whatever, but you could just think how Elon Musk's plan for the Tesla Bot, could work like this: you ask it to like I think Elon said "put that nut on that wheel using that wrench", then it has primed those features in its brain nut, wheel, wrench, and so it has to start looking for those objects (seeks to recognize them externally), which seems like we have the technology to do that when you think how good DALL-E is at prediction based on recognized context, or how good Google's similar image search is (it recognizes uploaded images too). Then once it seems them, it has to of course do it in the right order. It must grab the nut, and is again going to need to recognize the "turning 'whateverwasprimedhereobject' using 'primedtool' ". It is fun to think about. I hope they "put it together" (using existing tools. It can take some work in this case but maybe we are closer this time. They have to make it cheap but very useful, and both those requirements are really hard, but you can almost feel them in reach).
There's a lot of thingies to make that work, and you shave that list of things needed down if it is more general purpose and learns those things. Even if it is just an experiment to make just one of them, it would be interesting to try, maybe. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tf6dddebe1e89183a-Mb7d0ef24433a137b1c58d244 Delivery options: https://agi.topicbox.com/groups/agi/subscription
