I was about to say to be entirely apolitical is still political but then you mentioned making moves in a local and state sense which I can totally understand. Make an impact where you can make an impact rather than just add to the noise.
This has been incredibly insightful, about work aspects, talent agencies, and so much more. Currently I'm too hooked on a video game so there basically goes my slice of daily time pie, it's difficult to get away from that because I'm not very intrinsically motivated/driven, but this has given me food for thought.
Short (but complete) answer: The majority of jobs these days if you break it down to the very fundamental level are overwhelmingly shuffling data around in unique ways. This does not involve the need for physical interaction. Replicating brain neuron function is possible with non-organic systems. If you can replicate something close enough to a human brain outside of a human in a way that is at least as good and near infinitely scalable, it eliminates a lot of the need for humans to exist to solve those data shuffling problems. That's the theoretical finality of AI. Short term: Problems don't require nearly as many people as before to solve similar to all automation revolutions that came before. A single person becomes capable of doing the work of 10x or more people once trained. We're not quite all the way here now but definitely getting there. No need to panic all that much. Education is one sector that current AI tech has completely obliterated. I feel for the kiddos Long term: AI is a different beast than the automation revolutions that came before because at the very end it fundamentally eliminates the need for humans to exist. Previous technologies didn't do that. Stating that such a thing is impossible is naive. We humans exist and some of us are capable of doing extraordinary things. AI theoretically, will be just as capable. To put it in a way that is more relatable: Imagine you have a fully remote job and have a magic bot and you ask it "Do my job for me. All of it". Then you walk away. You check back at the end of the day and find that yep, everything had been done perfectly. Every task, every meeting, every callback, every piece of research that needed to be done, all of the R&D, emails, every presentation, everything. Perfectly executed, better than even you would be capable of doing in the same amount of time. As time goes on you check in on it less and less. You get more confident that the bot can do everything and one day after many years, you check back on it and it occurs to you that said bot is now the CEO of the company and the company orders of magnitude more successful than it was when the bot took over your job. Now imagine everybody is using that same bot. For what purpose would the concept of "having a job exist"? Any problem you have, an AI will be able to solve for you that doesn't involve something physical. Even then, it would be able to effectively help you along the entire path. Plus robots are progressing but at a snails pace compared to the AI systems so that makes the image of what that looks like more murky. Theoretically, even physical interaction wouldn't be necessary. Obviously there will be opportunity to exploit inefficiencies and unmet demands, but AI will also be capable of doing that or at least orchestrating that if opportunity arises. Conclusion: you exist in a world where AI has little foothold. Now the linchpin in the entire argument I just wrote: This assumes scaling continues in a steady fashion and some unknown fundamental bottleneck doesn't blindside progress. AI is currently constrained by 2 things, data and compute power. Raw data scaling has stopped about ~1-2 years ago. Aka we ran out of data to scrape. Hard to scrape more things when you've already scraped every archived book, research article, reddit post, blog, wikipedia article, store page, github page, youtube video, etc... everything on the internet you considered worthwhile to scrape. Petabytes if not exabytes of data. You already did that. So now what? Openai countered that limitation by generating synthetic data, refining said data, and increasing the time it takes to generate a response, aka forcing the AI to think longer. Enormous logic and reasoning improvements. Compute scaling shows no signs of slowing down and if anything is increasing with how many hundreds of billions of dollars companies are shelling out for infrastructure. I personally believe AI progress shows no signs of slowing down unless governments bring down the ban hammer unilaterally. If you read all that, thanks for reading! Accomplish great things!
Boo youtube censored my comment: Short (but complete) answer: The majority of jobs these days if you break it down to the very fundamental level are overwhelmingly shuffling data around in unique ways. This does not involve the need for physical interaction. Replicating brain neuron function is possible with non-organic systems. If you can replicate something close enough to a human brain outside of a human in a way that is at least as good and near infinitely scalable, it eliminates a lot of the need for humans to exist to solve those data shuffling problems. That's the theoretical finality of AI.
@@mememan9890 Those can't be automated away, yes. So those jobs will always remain, but AI can help massively improve efficiency if there is any room for efficiency to be gained. Plus robots themselves should advance in capability as time goes on further reducing the need for human labor. But never eliminating it. Case study: amazon fulfillment centers. Humans do things that robots/machines are "less good at". They can still do it but just not as fast/it costs more than having a human do it even if you pay the human ~$160/day. Everything that robots/machines are better at has been automated away. Now imagine that but for every physical job that has any semblance of consistency. Considerable reductions in workforce, but not elimination. Those jobs stay. But still reductions akin to the automation revolutions before. Hence why the number is ~90% and not 99%. In reality it will be a "0%" reduction as people will still do stuff but probably not really "make money" doing so but still call it a job. Hence, "the more things change, the more they stay the same"
I think it will replace jobs but just like any time in history jobs got replaced, it created whole new job markets. I don't think he's saying 90% of people will be unemployed.
Stream starts at 15:27
I was about to say to be entirely apolitical is still political but then you mentioned making moves in a local and state sense which I can totally understand.
Make an impact where you can make an impact rather than just add to the noise.
This has been incredibly insightful, about work aspects, talent agencies, and so much more. Currently I'm too hooked on a video game so there basically goes my slice of daily time pie,
it's difficult to get away from that because I'm not very intrinsically motivated/driven, but this has given me food for thought.
love hearing you talk keep it up devin
i'm 5 hours in and haven't heard the secret callout...what's the deal
6:07:30 just missed out 😅
"Boring is a symptom of an unobserved life" - have been telling people this for years but never been able to put it so eloquently. 🔥
Welcome back to live streaming! 🙌 yes! Always enjoyed your content
holy shit - a 6 hour watch, nice :D
I hope Devin explains why he thinks 90% of jobs will be replaced by AI. That seems insane to me.
Short (but complete) answer: The majority of jobs these days if you break it down to the very fundamental level are overwhelmingly shuffling data around in unique ways. This does not involve the need for physical interaction. Replicating brain neuron function is possible with non-organic systems. If you can replicate something close enough to a human brain outside of a human in a way that is at least as good and near infinitely scalable, it eliminates a lot of the need for humans to exist to solve those data shuffling problems. That's the theoretical finality of AI.
Short term: Problems don't require nearly as many people as before to solve similar to all automation revolutions that came before. A single person becomes capable of doing the work of 10x or more people once trained. We're not quite all the way here now but definitely getting there. No need to panic all that much. Education is one sector that current AI tech has completely obliterated. I feel for the kiddos
Long term: AI is a different beast than the automation revolutions that came before because at the very end it fundamentally eliminates the need for humans to exist. Previous technologies didn't do that. Stating that such a thing is impossible is naive. We humans exist and some of us are capable of doing extraordinary things. AI theoretically, will be just as capable.
To put it in a way that is more relatable: Imagine you have a fully remote job and have a magic bot and you ask it "Do my job for me. All of it". Then you walk away. You check back at the end of the day and find that yep, everything had been done perfectly. Every task, every meeting, every callback, every piece of research that needed to be done, all of the R&D, emails, every presentation, everything. Perfectly executed, better than even you would be capable of doing in the same amount of time. As time goes on you check in on it less and less. You get more confident that the bot can do everything and one day after many years, you check back on it and it occurs to you that said bot is now the CEO of the company and the company orders of magnitude more successful than it was when the bot took over your job.
Now imagine everybody is using that same bot.
For what purpose would the concept of "having a job exist"? Any problem you have, an AI will be able to solve for you that doesn't involve something physical. Even then, it would be able to effectively help you along the entire path. Plus robots are progressing but at a snails pace compared to the AI systems so that makes the image of what that looks like more murky. Theoretically, even physical interaction wouldn't be necessary.
Obviously there will be opportunity to exploit inefficiencies and unmet demands, but AI will also be capable of doing that or at least orchestrating that if opportunity arises.
Conclusion: you exist in a world where AI has little foothold.
Now the linchpin in the entire argument I just wrote: This assumes scaling continues in a steady fashion and some unknown fundamental bottleneck doesn't blindside progress. AI is currently constrained by 2 things, data and compute power. Raw data scaling has stopped about ~1-2 years ago. Aka we ran out of data to scrape. Hard to scrape more things when you've already scraped every archived book, research article, reddit post, blog, wikipedia article, store page, github page, youtube video, etc... everything on the internet you considered worthwhile to scrape. Petabytes if not exabytes of data. You already did that. So now what? Openai countered that limitation by generating synthetic data, refining said data, and increasing the time it takes to generate a response, aka forcing the AI to think longer. Enormous logic and reasoning improvements. Compute scaling shows no signs of slowing down and if anything is increasing with how many hundreds of billions of dollars companies are shelling out for infrastructure. I personally believe AI progress shows no signs of slowing down unless governments bring down the ban hammer unilaterally.
If you read all that, thanks for reading! Accomplish great things!
Boo youtube censored my comment:
Short (but complete) answer: The majority of jobs these days if you break it down to the very fundamental level are overwhelmingly shuffling data around in unique ways. This does not involve the need for physical interaction. Replicating brain neuron function is possible with non-organic systems. If you can replicate something close enough to a human brain outside of a human in a way that is at least as good and near infinitely scalable, it eliminates a lot of the need for humans to exist to solve those data shuffling problems. That's the theoretical finality of AI.
@@LiveType What about jobs that involve physical processes (ie turning the gears, shoveling the roads etc.)?
@@mememan9890 Those can't be automated away, yes. So those jobs will always remain, but AI can help massively improve efficiency if there is any room for efficiency to be gained. Plus robots themselves should advance in capability as time goes on further reducing the need for human labor. But never eliminating it.
Case study: amazon fulfillment centers. Humans do things that robots/machines are "less good at". They can still do it but just not as fast/it costs more than having a human do it even if you pay the human ~$160/day. Everything that robots/machines are better at has been automated away.
Now imagine that but for every physical job that has any semblance of consistency. Considerable reductions in workforce, but not elimination. Those jobs stay. But still reductions akin to the automation revolutions before. Hence why the number is ~90% and not 99%.
In reality it will be a "0%" reduction as people will still do stuff but probably not really "make money" doing so but still call it a job. Hence, "the more things change, the more they stay the same"
I think it will replace jobs but just like any time in history jobs got replaced, it created whole new job markets. I don't think he's saying 90% of people will be unemployed.
Will I find EvE market arbitrage guides on your Patreon?
first
scooby doo
i wonder how him and jasmine doing