If Google *canβt* effectively leverage their advantages, Iβll hate them forever; it would just be inexcusable big company nonsense!!! I donβt have a dog in the fight, but Christ would it cement their reputation as having just totally lost it at the strategic level.
I see no evidence we are near AGI nor that Google is leading the quest towards it. I consider the assumption that LLMs make AGI inevitable and close mostly a myth of the current hype cycle.
We don't know what it will take to make AGI. It might be that we have all of the tech and just need more compute/data. It might be that we need innovations so fundamental that nothing we do today even resembles what will power AGI.
My guess is that it's somewhere in between. I would be extremely surprised if AGI is achieved without some sort of language comprehension. I would not be surprised if current LLMs somewhat resemble the language comprehension modules of AGI.
Overall, for this post, I'm looking at technologies that I think are "fairly plausibly a subset of what it will take to build AGI". LLMs fall into that. I do not think that just scaling LLMs will get us to AGI, though.
Very interesting...guess it depends on ones opinion of Google X and the whole "celebrating failure culture," one can def say they have had a lotta duds, will be fun to see it play out no matter the winner!
If Google *canβt* effectively leverage their advantages, Iβll hate them forever; it would just be inexcusable big company nonsense!!! I donβt have a dog in the fight, but Christ would it cement their reputation as having just totally lost it at the strategic level.
I see no evidence we are near AGI nor that Google is leading the quest towards it. I consider the assumption that LLMs make AGI inevitable and close mostly a myth of the current hype cycle.
We don't know what it will take to make AGI. It might be that we have all of the tech and just need more compute/data. It might be that we need innovations so fundamental that nothing we do today even resembles what will power AGI.
My guess is that it's somewhere in between. I would be extremely surprised if AGI is achieved without some sort of language comprehension. I would not be surprised if current LLMs somewhat resemble the language comprehension modules of AGI.
Overall, for this post, I'm looking at technologies that I think are "fairly plausibly a subset of what it will take to build AGI". LLMs fall into that. I do not think that just scaling LLMs will get us to AGI, though.
I might be wrong but I thought it was China winning this race, with their Wu Dao 2.0?
Very interesting...guess it depends on ones opinion of Google X and the whole "celebrating failure culture," one can def say they have had a lotta duds, will be fun to see it play out no matter the winner!