Convergence and acceleration towards AGI (or Artificial Cognitive Entities)

Key Takeaways

hey everyone david shapiro here with a quick video um i wanted to talk about this sort of convergence and acceleration towards agi that i'm sensing what i mean by that is that you know i'm i'm deeply embedded in this space i'm finding all the right people that are working on this stuff i'm making youtube videos about it that's why you're here watching this and there are a few ideas that are emerging and there seems to be some some well i already use the word convergence multiple people are coming up with similar ideas so there seems to be this kind of scientific consensus building and one example that i have is this idea of composable architecture or composable cognition or symphony of thought um there's a guy on the open ai forum and twitter and youtube josh uh bachinski i think i'm saying his name right anyways he's got a uh twitter post about a demo of his his version of um an artificial cognitive entity that he calls cassandra and it works almost identically to what i have built and proposed with natural language cognitive architecture and as far as i can tell he came up with it independently of me although he did post that after i published my book so who knows maybe he was inspired by my book it's neither here nor there the the point is that there are multiple people coming to similar conclusions uh about how to achieve agi and then i just saw someone posted a really helpful comment um about google flamingo which google flamingo is basically a natural language cognitive architecture attached to a robot which is exactly what i proposed in um in natural language cognitive architecture is you just come up with the instructions with a large language model the large language model serves as the brain of the robot it's flexible it's adaptive and then you can have a robot improvise you can have it uh behave in an open-ended fashion um through iterative recursive loops of behavior uh so i think agi or ace is much closer than anyone realizes it's just a matter of commercializing what we've already got now a little bit more a tiny little bit more research to make these things a little bit more powerful and a little bit more flexible excuse me and also to make sure that they're reliable and that's why i'm really glad that i've published my book benevolent by design which talks about the core objective functions we should give these machines or the core heuristic imperatives that we should give them to ensure that they remain safe uh anyways it's just it's wild to be a part of this conversation and a part of this community um just kind of seeing this stuff come to life in real time uh yeah i'm just i'm blown away that's why i'm taking a walk for lunch just to reflect on you know my my contribution which you know i was just a little little drop in the bucket a little tiny push and just being part of this this transformation that we're going through more and more thought leaders are saying that artificial intelligence and automation are going to be far and away the most important technology we've ever created intelligent machines are the final invention of humanity um and you know we got to do it right and uh it just you know i'm starting to see that that while i hope i've had a positive contribution um it's still like i always knew that i couldn't do this on my own right um it's just way too big there's way too much research to do way too much money involved right you know i was never going to train gpt3 on my own um that sort of thing um and it's just good to to see this collective effort like there's you know i don't know how many of us there are working on this kind of stuff you know maybe ten thousand maybe a hundred thousand i don't know how many people are working on these problems globally but we are all working together and we're where it seems like there's a consensus emerging about how to achieve this and uh yeah so this this convergence and acceleration towards agi i think it's closer than anyone realizes so anyways thanks for watching hope this was helpful