illuminateme
Power Member
- Feb 24, 2016
- 774
- 369
I just did, launched a new saas without begging devs or fundraise, in 24hrs, 3 months ago. The world is in denial while we are figuring out how to maximize its potential.
No, Chat GPT/AI won't replace Software Engineers
- Thread starter Juhku
- Start date
BlackHatSoda
Regular Member
- Oct 6, 2009
- 393
- 329
If you code for a living and think ChatGPT will replace all developers, you're either bad at your job or don't understand how these LLM work.
Eventually (at least a decade) I think you'll start to see complex multimodal ML models trained specifically in every aspect of software development, operating under an umbrella model which can coordinate the models specialized in UI design, front-end coding, back-end coding, database architecture, QA, graphics design, UX design, systems architecture, security hardening, etc.
There are already simple examples of this now. ChatGPT branches off coding to its Codex model for example.
These will be able to replace a lot of low-level devs and designers but they still won't be able to do everything. It's going to take a while before human expert level models are developed, optimized, kept up to date, and hosted at an affordable price for the hundreds of languages, development specialties, and technologies that make up the world of software development.
Right now, we're nowhere close to that. Don't believe the hype about "sparks of AGI" coming out of OpenAI. They're for-profit, of course they're going to BS about their products to get PR, just like every other company does.
Even when these complex multimodal AI software development systems get to the point where they can generate fairly complex software, they're still not going to be able to innovate new technologies the way humans can, it's just not what these models do or how they operate.
You need a very fundamentally different type of AI to get to where you're replacing humans doing jobs that require novel, dynamic, off-the-cuff thinking and decision making.
You can get lucky with some things by throwing enough hardware at it and letting some neural network brute force its way to a solution, but not all solutions can be brute forced in a practical way unless you have unlimited resources and time, which nobody has.
Even throwing quantum computation into the mix which might be able to speed up some aspects of ML, you're still at least a decade off from seeing real applications coming out of a digital/quantum hybrid computer architecture (even though such systems already exist, though not far out from prototypes).
All that being said, the real issue is the billions of other jobs that cheap specialized AI, cheap robotics and other automation will replace in the next 20-30 years.
Keep in mind that as these billions of jobs disappear (yes, a few new ones will be created but nothing compared to what is lost), people in the developing world will add close to 3 billion more people over that same 30 year time frame, there's going to be almost nothing for them to do.
You're going to have at least half the population unemployed and taxing the hell out of those who still have jobs to put the rest on perpetual welfare won't do anything but lead to A LOT of people who are very unhappy that they have no chance of upward mobility and have a lot of time on their hands to try and take what they want from those who have more than them.
Eventually (at least a decade) I think you'll start to see complex multimodal ML models trained specifically in every aspect of software development, operating under an umbrella model which can coordinate the models specialized in UI design, front-end coding, back-end coding, database architecture, QA, graphics design, UX design, systems architecture, security hardening, etc.
There are already simple examples of this now. ChatGPT branches off coding to its Codex model for example.
These will be able to replace a lot of low-level devs and designers but they still won't be able to do everything. It's going to take a while before human expert level models are developed, optimized, kept up to date, and hosted at an affordable price for the hundreds of languages, development specialties, and technologies that make up the world of software development.
Right now, we're nowhere close to that. Don't believe the hype about "sparks of AGI" coming out of OpenAI. They're for-profit, of course they're going to BS about their products to get PR, just like every other company does.
Even when these complex multimodal AI software development systems get to the point where they can generate fairly complex software, they're still not going to be able to innovate new technologies the way humans can, it's just not what these models do or how they operate.
You need a very fundamentally different type of AI to get to where you're replacing humans doing jobs that require novel, dynamic, off-the-cuff thinking and decision making.
You can get lucky with some things by throwing enough hardware at it and letting some neural network brute force its way to a solution, but not all solutions can be brute forced in a practical way unless you have unlimited resources and time, which nobody has.
Even throwing quantum computation into the mix which might be able to speed up some aspects of ML, you're still at least a decade off from seeing real applications coming out of a digital/quantum hybrid computer architecture (even though such systems already exist, though not far out from prototypes).
All that being said, the real issue is the billions of other jobs that cheap specialized AI, cheap robotics and other automation will replace in the next 20-30 years.
Keep in mind that as these billions of jobs disappear (yes, a few new ones will be created but nothing compared to what is lost), people in the developing world will add close to 3 billion more people over that same 30 year time frame, there's going to be almost nothing for them to do.
You're going to have at least half the population unemployed and taxing the hell out of those who still have jobs to put the rest on perpetual welfare won't do anything but lead to A LOT of people who are very unhappy that they have no chance of upward mobility and have a lot of time on their hands to try and take what they want from those who have more than them.
- Mar 22, 2013
- 6,272
- 10,985
AI is still in the early stage. Now it is GPT-4. What about GPT-10?
The real deal here is that millions of people are using ChatGPT / OpenAI every single day. Right now they collect a lot of data, while even earning money. Means they can improve their AI even faster and better.
Also, the same thing will happen with everything else: Only the specialists and experts will survive. If you are a guy who has little to medium skills, AI is going to take your job. But if you do a job in which you are extremely good, no one is going to replace you.
The real deal here is that millions of people are using ChatGPT / OpenAI every single day. Right now they collect a lot of data, while even earning money. Means they can improve their AI even faster and better.
Also, the same thing will happen with everything else: Only the specialists and experts will survive. If you are a guy who has little to medium skills, AI is going to take your job. But if you do a job in which you are extremely good, no one is going to replace you.
- May 27, 2016
- 11,083
- 2,516
The medium senior programmers will never be replaced. Maybe the entry level/junior programmers will have a hard time finding a job. But then again if you can't start somewhere the existing medium/senior programmers will be able to monopolize the market. At least these are my thoughts.
ProxyBird
Regular Member
- Jun 9, 2011
- 388
- 161
lol I'm with you man. I'm smart enough to know I have virtually no idea what to expect. Any prediction I make is probably so far from the truth if you showed me when the future happens I'd laugh at myself.