>I can't understand the mindset of any AI researchers who stay at Google at this point.
>You're there (presumably) because you want to make an impact.
I can't speak to AI researchers specifically but I can speak to research more generally. Ideally you want to focus on somewhere you can make and impact, pursue the general line of research you find interesting/promising, AND have stability to at least live (or ideally enough pay to not care).
I've known enough highly talented researchers who almost always have to sacrifice the "making a difference" and "pursuing their pathway" to large degree just to survive. Ultimately what people are paying for to be researched is what matters, be it some business R&D division's head, federal agency biased directions, of whatever philanthropic connections some organizations are able to extract decide you should be doing. It's not just medicore researchers, it's what one might consider "world class" top of their field researchers. Everyone at some point makes sacrifices to pay the bills but there's this neverending gaslighting that occurs about what researchers want to do and what areas they choose to impact or some nonsense like that. That happens in very few idealized cases. E.g., tenured professorship at highly endowed institutions where they've already invested enough time doing the stuff they don't want to game the system to get to a point they can finally pursue their own paths and so on.
If you want to focus on an impact and perform research you need to be able to self-fund that and hope the area you work isn't capital intensive (AI specifically DNN, LLM, and highly data/compute intensive approaches is for the most part pretty capital intensive, hence one reason big tech pursued them to reduce competition--you can pursue paths theoretically but ultimately unless you find approaches that aren't as capital intensive it'll be awhile before you can experimentally test and iterate approaches).
Researchers just hope to get remotely close to the area they want to work in but by in large have to chase the money and lines of research behind it if they hope to remain a researcher. Maybe Google is different but I doubt it. This happens in other industries as well with people in R&D near the tops of their field working at market leaders still caving.
If by some miracle you happen to be near the top of your field and happen to guess or have some natural deep insight to the path forward where you're fortunate enough to make a huge breakthrough, you can often just spin it out yourself or take on some investment risk. Why work for Google if you have the practical path to say AGI in your hands? Try and do it on your own or find someone who doesn't own all your IP at the end of it. You work at big corp because there's ideally some stability balance with some semi-interesting research paths to you, probably not exactly what you want but as close as you can get.
I can't speak to AI researchers specifically but I can speak to research more generally. Ideally you want to focus on somewhere you can make and impact, pursue the general line of research you find interesting/promising, AND have stability to at least live (or ideally enough pay to not care).
I've known enough highly talented researchers who almost always have to sacrifice the "making a difference" and "pursuing their pathway" to large degree just to survive. Ultimately what people are paying for to be researched is what matters, be it some business R&D division's head, federal agency biased directions, of whatever philanthropic connections some organizations are able to extract decide you should be doing. It's not just medicore researchers, it's what one might consider "world class" top of their field researchers. Everyone at some point makes sacrifices to pay the bills but there's this neverending gaslighting that occurs about what researchers want to do and what areas they choose to impact or some nonsense like that. That happens in very few idealized cases. E.g., tenured professorship at highly endowed institutions where they've already invested enough time doing the stuff they don't want to game the system to get to a point they can finally pursue their own paths and so on.
If you want to focus on an impact and perform research you need to be able to self-fund that and hope the area you work isn't capital intensive (AI specifically DNN, LLM, and highly data/compute intensive approaches is for the most part pretty capital intensive, hence one reason big tech pursued them to reduce competition--you can pursue paths theoretically but ultimately unless you find approaches that aren't as capital intensive it'll be awhile before you can experimentally test and iterate approaches).
Researchers just hope to get remotely close to the area they want to work in but by in large have to chase the money and lines of research behind it if they hope to remain a researcher. Maybe Google is different but I doubt it. This happens in other industries as well with people in R&D near the tops of their field working at market leaders still caving.
If by some miracle you happen to be near the top of your field and happen to guess or have some natural deep insight to the path forward where you're fortunate enough to make a huge breakthrough, you can often just spin it out yourself or take on some investment risk. Why work for Google if you have the practical path to say AGI in your hands? Try and do it on your own or find someone who doesn't own all your IP at the end of it. You work at big corp because there's ideally some stability balance with some semi-interesting research paths to you, probably not exactly what you want but as close as you can get.