6 ms·

The Rise of Artificial Intelligence and the End of Code

48 points by AJRF 10y ago

coldcode 10y ago
There will be a day when the last programmer writes the last program. Long after we are all dead. While some types of jobs may be automatable, most of what I have to do makes little sense to smart people much less a rigid AI. What AI can take almost random requirements, strange UI designs and animations and clueless VP driven features and turn it into something people want to use. I think the VP's are more likely to be replaced by an AI.
- eellpp 10y ago
  VP may be replaced by a VP role that requires him to convert the random requirements to specifics or guide the AI on UI design etc. Instead of a VP and team of developers there would be devVp would could bridge this gap. And then in still further future, this friction could even be removed ...
paulsutter 10y ago
AI techniques don't require acres of developers writing acres of if/thens (traditional software development is just that). The Deepmind Atari player is only 1500 lines of code.
- narrator 10y ago
  I think that software developers are going to become one of the few professions left and will be massively productive because they just have to hook up the AI training and they're basically done. Anyone who can hack this stuff is going to be able to be on the other end of the wealth divide real quick. Transfer learning is a beautiful awesome thing. You can download ImageNet models trained on huge amounts of data by GPU farms and suddenly you've got a world class image recognizer running on your very own GPU. Reinforcement learning is totally off the hook. You can make a bot for a random board game that you barely know how to play.
wrong_variable 10y ago
Dogs can be made to learn things - doesn't mean we allow them do surgery. Whoever wrote this article has negative knowledge about the state of AI. edit: Its also slightly insulting that this person thinks that the modern programmer spends all their time writing if/else/then statement.
- pcarolan 10y ago
  Agreed. When Spotify stops putting Coldplay in my Radiohead station, I'll begin to worry.
- nxzero 10y ago
  >> "Dogs can be made to learn things - doesn't mean we allow them do surgery." (Years from now) ..."humans can learn to be surgeons, but that doesn't mean they should be allowed to do surgery." For example: http://spectrum.ieee.org/the-human-os/robotics/medical-robots/autonomous-robot-surgeon-bests-human-surgeons-in-world-first http://spectrum.ieee.org/the-human-os/robotics/medical-robot...
- pedalpete 10y ago
  I'm going to take the other side of this argument and suggest that we didn't create the dog and the limitations in the dogs abilities. I suspect that this will be difficult for the HN community to hear, but we must face the upcoming realities. Though I think the article may be assigning more abilities to AI than is actually feasible in our lifetimes, let us not forget that most back-end developers are doing basic I/O and data manipulation tasks. Put something into a database, get something out of a database. Front-enders are doing layout, animation, sizing and colouring. The average modern programmer doesn't spend their time writing if/else/then statements, I believe the average programmer spends their time writing GET/SET/EDIT/DELETE flows. I believe in the next 10 years, artificial intelligence will be able to model and create these basic systems based on human input. UX and possibly even UI can also be trained. A/B testing can certainly be trained. Those who think programmers can't be replaced begin to sound like the Luddites who felt the loom could not match the quality of their craft. Note: I am a full-stack developer, I believe 90% of my day job will be replaced by AI in the coming years. That final 10% often deciding what to do or the best how, will likely be the main task.
boznz 10y ago
Based on the user requirements I usually get (or lack thereof) I think it will be a few years yet until I am out of a job.
artifaxx 10y ago
The end of code seems like a title created by a sensationalist... I am excited for how these forms of pattern recognition and decision making can allow us to do more with software. There is still plenty of code to be written as we continue to apply software to every aspect of our world.
- JustSomeNobody 10y ago
  I wouldn't be surprised if wired kicked out an article like this once every few years, to be honest.
carapace 10y ago
If you have been following the recent advances in AI (and are not clueless) then it should be getting really REALLY obvious that most coding and design tasks will be automated within three to five years. AI will eat your career.
- b34r 10y ago
  Front end engineer here, practically zero chance of that happening. At least on my side of the fence.
  - PeCaN 10y ago
    AIs can stay up all night reading blogs about the latest and greatest JS framework though. You better look out.
- psycr 10y ago
  Can you articulate this in the form of a binary outcome? If so, I will happily make a $N00 sized bet (odds to be negotiated) on the false outcome with anyone who is interested.
  - carapace 10y ago
    Yes. By 2020 global job pool for software developers and user interface designers will start to contract due to competition from automated design/development AI.
- hiuhsdf 10y ago
  Not really, AI is still absolute shit at most tasks, and programming requires many different skills. AlphaGo, for example does a fairly simple task (relative to programming), which was trained on tons of existing and very simple data. If you are automating coding and design tasks that means you are adding new tools, that people on both ends on the automated process will need to learn and understand, this adds extra cost and complexity. I agree that many programming tasks can be automated, but I don't think it will have a significant impact in the next 50 years. I think automation will hit most other sectors a lot faster and harder than it will IT, and the new automation will require more engineers to manage it.
Houshalter 10y ago
If AIs replace programmers, why would they stop there? They could replace every job. Including AI researchers!
davesque 10y ago
This is mostly a futurist view on the topic. Not surprising given it's from wired.com.
frogfuzion 10y ago
This is exactly why i am apprenticing as an HVAC technician on the side. I dont believe there is longevity in this position over the next 10-20 years.
Hydraulix989 10y ago
Amazing listening to all of the HN coder apologists here come up with excuses for why they aren't going to lose their jobs, too. If anything, abstraction, ever-tighter vendor lock-in, and improvements in static analysis will kill off many coder jobs. There aren't too many jobs out there for recreating basic UNIX tools like sed or grep -- soon, this will be true for higher level tools as well, until what you're doing now at your desk as a coder becomes a "tool" or already exists as a library. It will eventually be like self-driving cars: Not totally automated, but enough to make the job of programming like a modern-day pilot's -- throw it in autopilot, and then step in whenever (rarely) necessary. Then people will realize fewer people are needed to keep watch and step in, and that one person is enough to do the same thing as many people.
- origami777 10y ago
  I do some coding but not as a full time job. And I have to admit that when I think of AI replacing a coder's job it seems so far off. Maybe it's because I'm in the thick of it and can't see the forest for the trees. When this comes up I often wonder what part is AI going to replace? Will it: 1) Assess a business opportunity, design, and develop, deploy, and operate an entire technology solution? 2) If not that, will it pick up at some part of the value chain above and be able to figure out how to do the rest? It seems more likely to me that the tooling for workers in this value chain will get more advanced. Some of that might include some computer intelligence that is able to perform some tasks on its own.
  - Hydraulix989 10y ago
    "WIll it? 1) Assess a business opportunity, design, and develop, deploy, and operate an entire technology solution?" At all of my software engineering jobs, I only ever performed tasks that accomplished what one would consider a small fraction of one list item here -- "develop." So to answer your question, mostly no. You're describing something that an entrepreneur does, not a coder. "2) If not that, will it pick up at some part of the value chain above and be able to figure out how to do the rest?" Yes, the "develop"-ment (deployment is getting automated already -- Chef, Heroku, etc. -- and operate is also largely automated now thanks to things like cloud computing services). Case in point: Why did WhatsApp have so few employees for its billion users? Well, automation made launching a web startup (doing (1), as you described) a much leaner proposition. As you can see, AI is not a necessary condition for automation. I have no reason to believe these trends will reverse course.
trhway 10y ago
why would AI program? Programming is a way to make hardware perform some task. Well, couldn't AI just go straight to performing the task, skipping the programming step, at least as we know it.
doug1001 10y ago
articles with this identical thesis and conclusion have been published from time to time for the past 25 years. i started coding in the early 1980's and began coding professionally just after graduating from MIT in 1992. The "guidance counselor" at my high school recommended that i major in something other than computer science because he has read several thoughtful pieces that suggested the job of professional programmer would become "extinct" not because programming wasn't important but because it was too important (to leave to programmers)--sort of like the way many companies 50 years ago had large "typing pools"; these are gone because everyone can type. looking across the span of my career so far, it seems these predictions are correct, in a horizon sense--ie, we've reached this distance line we pointed to 20 years ago, but there's no horizon there, it's still far in the distance. here's what i mean: in 1992, my first job involved writing small extensions to numerical libraries, in C and FORTRAN. Probably 90 percent (if anything it was higher) of my time was spent "managing memory"--allocating it, deallocating it, pointing to it, etc. now, the time i spend on those sorts of tasks is de minimis. So the march of technology never took away my job (the only job i have ever wanted to do) but it did take away 90% of it. With 90% of my former tasks now handled elsewhere, far more interesting and challenging tasks have taken their place. this is the trend i see--not the one-sided (or perhaps short-sited) trend described in the Wired post--but a continuous iteration of the programmer's work characterized by swapping well-characterized, routinized tasks for more challenging ones, most of which are would take some thought to foresee.

48 points by AJRF 10y ago

There will be a day when the last programmer writes the last program. Long after we are all dead. While some types of jobs may be automatable, most of what I have to do makes little sense to smart people much less a rigid AI. What AI can take almost random requirements, strange UI designs and animations and clueless VP driven features and turn it into something people want to use. I think the VP's are more likely to be replaced by an AI.

eellpp 10y ago
VP may be replaced by a VP role that requires him to convert the random requirements to specifics or guide the AI on UI design etc. Instead of a VP and team of developers there would be devVp would could bridge this gap. And then in still further future, this friction could even be removed ...

paulsutter 10y ago

AI techniques don't require acres of developers writing acres of if/thens (traditional software development is just that). The Deepmind Atari player is only 1500 lines of code.

narrator 10y ago
I think that software developers are going to become one of the few professions left and will be massively productive because they just have to hook up the AI training and they're basically done. Anyone who can hack this stuff is going to be able to be on the other end of the wealth divide real quick. Transfer learning is a beautiful awesome thing. You can download ImageNet models trained on huge amounts of data by GPU farms and suddenly you've got a world class image recognizer running on your very own GPU. Reinforcement learning is totally off the hook. You can make a bot for a random board game that you barely know how to play.

wrong_variable 10y ago

Dogs can be made to learn things - doesn't mean we allow them do surgery. Whoever wrote this article has negative knowledge about the state of AI. edit: Its also slightly insulting that this person thinks that the modern programmer spends all their time writing if/else/then statement.

pcarolan 10y ago
Agreed. When Spotify stops putting Coldplay in my Radiohead station, I'll begin to worry.
nxzero 10y ago
>> "Dogs can be made to learn things - doesn't mean we allow them do surgery." (Years from now) ..."humans can learn to be surgeons, but that doesn't mean they should be allowed to do surgery." For example: http://spectrum.ieee.org/the-human-os/robotics/medical-robots/autonomous-robot-surgeon-bests-human-surgeons-in-world-first http://spectrum.ieee.org/the-human-os/robotics/medical-robot...
pedalpete 10y ago
I'm going to take the other side of this argument and suggest that we didn't create the dog and the limitations in the dogs abilities. I suspect that this will be difficult for the HN community to hear, but we must face the upcoming realities. Though I think the article may be assigning more abilities to AI than is actually feasible in our lifetimes, let us not forget that most back-end developers are doing basic I/O and data manipulation tasks. Put something into a database, get something out of a database. Front-enders are doing layout, animation, sizing and colouring. The average modern programmer doesn't spend their time writing if/else/then statements, I believe the average programmer spends their time writing GET/SET/EDIT/DELETE flows. I believe in the next 10 years, artificial intelligence will be able to model and create these basic systems based on human input. UX and possibly even UI can also be trained. A/B testing can certainly be trained. Those who think programmers can't be replaced begin to sound like the Luddites who felt the loom could not match the quality of their craft. Note: I am a full-stack developer, I believe 90% of my day job will be replaced by AI in the coming years. That final 10% often deciding what to do or the best how, will likely be the main task.

boznz 10y ago

Based on the user requirements I usually get (or lack thereof) I think it will be a few years yet until I am out of a job.

artifaxx 10y ago

The end of code seems like a title created by a sensationalist... I am excited for how these forms of pattern recognition and decision making can allow us to do more with software. There is still plenty of code to be written as we continue to apply software to every aspect of our world.

JustSomeNobody 10y ago
I wouldn't be surprised if wired kicked out an article like this once every few years, to be honest.

carapace 10y ago

If you have been following the recent advances in AI (and are not clueless) then it should be getting really REALLY obvious that most coding and design tasks will be automated within three to five years. AI will eat your career.

b34r 10y ago
Front end engineer here, practically zero chance of that happening. At least on my side of the fence.
- PeCaN 10y ago
  AIs can stay up all night reading blogs about the latest and greatest JS framework though. You better look out.
psycr 10y ago
Can you articulate this in the form of a binary outcome? If so, I will happily make a $N00 sized bet (odds to be negotiated) on the false outcome with anyone who is interested.
- carapace 10y ago
  Yes. By 2020 global job pool for software developers and user interface designers will start to contract due to competition from automated design/development AI.
hiuhsdf 10y ago
Not really, AI is still absolute shit at most tasks, and programming requires many different skills. AlphaGo, for example does a fairly simple task (relative to programming), which was trained on tons of existing and very simple data. If you are automating coding and design tasks that means you are adding new tools, that people on both ends on the automated process will need to learn and understand, this adds extra cost and complexity. I agree that many programming tasks can be automated, but I don't think it will have a significant impact in the next 50 years. I think automation will hit most other sectors a lot faster and harder than it will IT, and the new automation will require more engineers to manage it.

Houshalter 10y ago

If AIs replace programmers, why would they stop there? They could replace every job. Including AI researchers!

davesque 10y ago

This is mostly a futurist view on the topic. Not surprising given it's from wired.com.

frogfuzion 10y ago

This is exactly why i am apprenticing as an HVAC technician on the side. I dont believe there is longevity in this position over the next 10-20 years.

Hydraulix989 10y ago

Amazing listening to all of the HN coder apologists here come up with excuses for why they aren't going to lose their jobs, too. If anything, abstraction, ever-tighter vendor lock-in, and improvements in static analysis will kill off many coder jobs. There aren't too many jobs out there for recreating basic UNIX tools like sed or grep -- soon, this will be true for higher level tools as well, until what you're doing now at your desk as a coder becomes a "tool" or already exists as a library. It will eventually be like self-driving cars: Not totally automated, but enough to make the job of programming like a modern-day pilot's -- throw it in autopilot, and then step in whenever (rarely) necessary. Then people will realize fewer people are needed to keep watch and step in, and that one person is enough to do the same thing as many people.

origami777 10y ago
I do some coding but not as a full time job. And I have to admit that when I think of AI replacing a coder's job it seems so far off. Maybe it's because I'm in the thick of it and can't see the forest for the trees. When this comes up I often wonder what part is AI going to replace? Will it: 1) Assess a business opportunity, design, and develop, deploy, and operate an entire technology solution? 2) If not that, will it pick up at some part of the value chain above and be able to figure out how to do the rest? It seems more likely to me that the tooling for workers in this value chain will get more advanced. Some of that might include some computer intelligence that is able to perform some tasks on its own.
- Hydraulix989 10y ago
  "WIll it? 1) Assess a business opportunity, design, and develop, deploy, and operate an entire technology solution?" At all of my software engineering jobs, I only ever performed tasks that accomplished what one would consider a small fraction of one list item here -- "develop." So to answer your question, mostly no. You're describing something that an entrepreneur does, not a coder. "2) If not that, will it pick up at some part of the value chain above and be able to figure out how to do the rest?" Yes, the "develop"-ment (deployment is getting automated already -- Chef, Heroku, etc. -- and operate is also largely automated now thanks to things like cloud computing services). Case in point: Why did WhatsApp have so few employees for its billion users? Well, automation made launching a web startup (doing (1), as you described) a much leaner proposition. As you can see, AI is not a necessary condition for automation. I have no reason to believe these trends will reverse course.

trhway 10y ago

why would AI program? Programming is a way to make hardware perform some task. Well, couldn't AI just go straight to performing the task, skipping the programming step, at least as we know it.

doug1001 10y ago

articles with this identical thesis and conclusion have been published from time to time for the past 25 years. i started coding in the early 1980's and began coding professionally just after graduating from MIT in 1992. The "guidance counselor" at my high school recommended that i major in something other than computer science because he has read several thoughtful pieces that suggested the job of professional programmer would become "extinct" not because programming wasn't important but because it was too important (to leave to programmers)--sort of like the way many companies 50 years ago had large "typing pools"; these are gone because everyone can type. looking across the span of my career so far, it seems these predictions are correct, in a horizon sense--ie, we've reached this distance line we pointed to 20 years ago, but there's no horizon there, it's still far in the distance. here's what i mean: in 1992, my first job involved writing small extensions to numerical libraries, in C and FORTRAN. Probably 90 percent (if anything it was higher) of my time was spent "managing memory"--allocating it, deallocating it, pointing to it, etc. now, the time i spend on those sorts of tasks is de minimis. So the march of technology never took away my job (the only job i have ever wanted to do) but it did take away 90% of it. With 90% of my former tasks now handled elsewhere, far more interesting and challenging tasks have taken their place. this is the trend i see--not the one-sided (or perhaps short-sited) trend described in the Wired post--but a continuous iteration of the programmer's work characterized by swapping well-characterized, routinized tasks for more challenging ones, most of which are would take some thought to foresee.