Absolutely - cost of software engineering is the major hurdle for most software projects - but it isn't the only factor - the other major factor will be the success rate of software projects - if engineers using ai have a higher success rate on projects - then that will reduce risks - and make software a safer investment
Excellent point! Success rate and quality is something that doesn't get enough discussion and honestly seems to be a growing issue. So many half-baked products going to market nowadays. Highly skilled engineers + AI can create much better products!
Yes, you are right about that - if that assumption is false, my argument doesn't work. The reason I believe it to be true is that this is what we have seen from the industry over the last 50 years of software development. We keep seeing advancements that increase efficiency - higher-level programming languages, the internet, version control, cloud, etc. All of these advancements make software development more efficient but we've only seen market demand grow in response.
Thank you so much! I try to go beyond the hype and figure out what's actually possible and where we are headed. With so much change, there's plenty of people on both sides - some don't realize the incoming implications, while others buy all the hype. I try to keep it real! Thanks for watching :)
People underestimate the impact ai will have. I don’t believe it will just change the roles we play it will be the one choosing the best path. There will come a time where it is far better at problem solving than humans. This is not that far away either. It’s just a matter of someone stumbling apon the right architecture and there we have superhuman intelligence.
To solve a problem you need to provide it a problem. Identifying the problem and framing it in terms of what the computer understands is not so easy as you might think
Agreed with @cybermonkey2807 - identifying and framing is the hard part, which engineers know how to do well. But I do agree that with the right architecture and some tweaks, we could quickly see a superhuman intelligence emerge. Even so, I think rolling it out and integrating it into daily life will be a struggle and that path will be covered in problems for humans to solve. After that, I suppose it shouldn't matter :)
As long as humans are in charge, we will be the ones deciding what and why needs to be built not AIs. They may help in brainstorming, they may help in troubleshooting, identifying the issues in society and projects worth pursuing, but a human at the end of the day has to say "Yes" or "No". If this is not the case, humans will no longer be in charge. But even in this fantasy scenario, humans will continue to have ambitions, goals, preferences, etc that are entirely independent of how superintelligent an AI is.
@@VoloBuilds the end goal of automation(beyond all the euphemisms used) is to replace or reduce human labour in that specific sector in which automation is implemented counter example? pretty much all history of automation (in manufacturing, agriculture, etc.)
Thanks for explaining your perspective! I would argue that the end goal of automation is to reduce cost and maximize profit through increased efficiency and productivity. The reduction in human labor is a byproduct that occurs in *some* cases, particularly when demand for the output is inelastic. As a result of this automation - you are right that there is a reduction in human labor - but it is at the unit level. There is less human labor needed to create the same thing. But as Jevons discovered, this can lead to an increase in aggregate demand, which is what I believe is happening with automation in software - something that we have seen with previous advancements in the software space.
@@VoloBuilds like I've said before, those are euphemisms correct me if I'm wrong, but how can something be done by a person and automatically in the same time? also, I would argue that solutions for building cheap software existed already for many years so I doubt will see a boost in demand because of A.I ... I assume you've heard of outsourcing? If you think paying indian programmers is expensive then idk what else to tell you
It's a different framing - the goal isn't to reduce human labor, but to make more profit. For software, more profit can be made by having the same (or more) people and making even more/better software faster. Let me share an example to demonstrate my point: --- We build apps. We sell each app for $100k. Right now it takes 5 developers to build an app and we pay them each $10k. We profit $50k per app (100k - 5 * 10k) AI comes along, automates 80% of the work. Now we can build an app with just 1 developer - $10k We profit $90k per app. That's an 80% increase in total profit. Should we fire the other 4 developers? What if instead of building 1 app with 4 less developers, we instead keep 5 people and have each one build an app? Now we can build x5 more We build 5 apps and sell them for $500k total Cost is still $10k per dev ($50k total) Profit is $450k. That's a 900% increase in total profit. We just made wayyy more profit by keeping the developers instead of firing them. In fact, if we can keep selling more apps, it makes sense for us to hire even more developers now!! --- There are 2 big assumptions in my example above: 1. The market wants to buy as many apps as we can make - this is the elastic demand I am talking about. One of my key points in the video is that I believe there are so many uses for software that this is true in the big picture. Sure, there's a limited demand for apps that track your pet rock's sleeping habits, so if that's the only type of app you are building, you should probably fire the other 4 devs. But zoom out and there's demand for all sorts of software, and even more types of software now with AI. Curious to hear if you see it differently on this point. 2. The automation doesn't handle literally 100% of the work. If 100% of software creation is automated, we should be able to ask an AI to write software that automatically earns an income for us (100% automated!), at which point I suppose it doesn't matter if we still have jobs.
Absolutely - cost of software engineering is the major hurdle for most software projects - but it isn't the only factor - the other major factor will be the success rate of software projects - if engineers using ai have a higher success rate on projects - then that will reduce risks - and make software a safer investment
Excellent point! Success rate and quality is something that doesn't get enough discussion and honestly seems to be a growing issue. So many half-baked products going to market nowadays. Highly skilled engineers + AI can create much better products!
Great video!
Thanks for watching! Glad you enjoyed it!
This assumes that demand for software is elastic like it is for steam engines. I don't think that's true.
Yes, you are right about that - if that assumption is false, my argument doesn't work. The reason I believe it to be true is that this is what we have seen from the industry over the last 50 years of software development. We keep seeing advancements that increase efficiency - higher-level programming languages, the internet, version control, cloud, etc. All of these advancements make software development more efficient but we've only seen market demand grow in response.
Great video, very well articulated and grounded in reality, instead of some futuristic vision of a godlike AI utopia.
Thank you so much! I try to go beyond the hype and figure out what's actually possible and where we are headed. With so much change, there's plenty of people on both sides - some don't realize the incoming implications, while others buy all the hype. I try to keep it real! Thanks for watching :)
Devin's Paradox
Devin's kinda overhyped (and apparently faked to some degree) - my thoughts on Devin here: th-cam.com/video/kVQzJ2K2Wpw/w-d-xo.html
wow this is the best for AI
People underestimate the impact ai will have. I don’t believe it will just change the roles we play it will be the one choosing the best path. There will come a time where it is far better at problem solving than humans. This is not that far away either. It’s just a matter of someone stumbling apon the right architecture and there we have superhuman intelligence.
Wrong.
To solve a problem you need to provide it a problem. Identifying the problem and framing it in terms of what the computer understands is not so easy as you might think
Agreed with @cybermonkey2807 - identifying and framing is the hard part, which engineers know how to do well. But I do agree that with the right architecture and some tweaks, we could quickly see a superhuman intelligence emerge. Even so, I think rolling it out and integrating it into daily life will be a struggle and that path will be covered in problems for humans to solve. After that, I suppose it shouldn't matter :)
As long as humans are in charge, we will be the ones deciding what and why needs to be built not AIs. They may help in brainstorming, they may help in troubleshooting, identifying the issues in society and projects worth pursuing, but a human at the end of the day has to say "Yes" or "No".
If this is not the case, humans will no longer be in charge. But even in this fantasy scenario, humans will continue to have ambitions, goals, preferences, etc that are entirely independent of how superintelligent an AI is.
Cherry picked example (cope)
Could you elaborate? I had several examples so trying to figure out what you're referring to and what counter examples you have
@@VoloBuilds the end goal of automation(beyond all the euphemisms used) is to replace or reduce human labour in that specific sector in which automation is implemented
counter example? pretty much all history of automation (in manufacturing, agriculture, etc.)
Thanks for explaining your perspective! I would argue that the end goal of automation is to reduce cost and maximize profit through increased efficiency and productivity. The reduction in human labor is a byproduct that occurs in *some* cases, particularly when demand for the output is inelastic.
As a result of this automation - you are right that there is a reduction in human labor - but it is at the unit level. There is less human labor needed to create the same thing. But as Jevons discovered, this can lead to an increase in aggregate demand, which is what I believe is happening with automation in software - something that we have seen with previous advancements in the software space.
@@VoloBuilds like I've said before, those are euphemisms
correct me if I'm wrong, but how can something be done by a person and automatically in the same time?
also, I would argue that solutions for building cheap software existed already for many years so I doubt will see a boost in demand because of A.I ... I assume you've heard of outsourcing? If you think paying indian programmers is expensive then idk what else to tell you
It's a different framing - the goal isn't to reduce human labor, but to make more profit. For software, more profit can be made by having the same (or more) people and making even more/better software faster.
Let me share an example to demonstrate my point:
---
We build apps. We sell each app for $100k.
Right now it takes 5 developers to build an app and we pay them each $10k.
We profit $50k per app (100k - 5 * 10k)
AI comes along, automates 80% of the work.
Now we can build an app with just 1 developer - $10k
We profit $90k per app. That's an 80% increase in total profit.
Should we fire the other 4 developers?
What if instead of building 1 app with 4 less developers, we instead keep 5 people and have each one build an app?
Now we can build x5 more
We build 5 apps and sell them for $500k total
Cost is still $10k per dev ($50k total)
Profit is $450k. That's a 900% increase in total profit.
We just made wayyy more profit by keeping the developers instead of firing them. In fact, if we can keep selling more apps, it makes sense for us to hire even more developers now!!
---
There are 2 big assumptions in my example above:
1. The market wants to buy as many apps as we can make - this is the elastic demand I am talking about. One of my key points in the video is that I believe there are so many uses for software that this is true in the big picture. Sure, there's a limited demand for apps that track your pet rock's sleeping habits, so if that's the only type of app you are building, you should probably fire the other 4 devs. But zoom out and there's demand for all sorts of software, and even more types of software now with AI. Curious to hear if you see it differently on this point.
2. The automation doesn't handle literally 100% of the work. If 100% of software creation is automated, we should be able to ask an AI to write software that automatically earns an income for us (100% automated!), at which point I suppose it doesn't matter if we still have jobs.
Cope