Jack Dorsey Tells Andrew Yang: 'AI is Coming For Programming Jobs' (cnbc.com) 211
An anonymous reader quotes CNBC:
The rise of artificial intelligence will make even software engineers less sought after. That's because artificial intelligence will soon write its own software, according to Jack Dorsey, the tech billionaire boss of Twitter and Square. And that's going to put some beginning-level software engineers in a tough spot.
"We talk a lot about the self-driving trucks in and whatnot" when discussing how automation will replace jobs held by humans, Dorsey told former Democratic presidential hopeful Andrew Yang on an episode of the "Yang Speaks" podcast published Thursday. But A.I. "is even coming for programming" jobs, Dorsey said.
"A lot of the goals of machine learning and deep learning is to write the software itself over time so a lot of entry-level programming jobs will just not be as relevant anymore," Dorsey told Yang.
Dorsey also told Yang that he belives a Universal Basic Income could give workers "peace of mind" that they'll be able to "eat and feed their children while they are learning how to transition into this new world."
"We talk a lot about the self-driving trucks in and whatnot" when discussing how automation will replace jobs held by humans, Dorsey told former Democratic presidential hopeful Andrew Yang on an episode of the "Yang Speaks" podcast published Thursday. But A.I. "is even coming for programming" jobs, Dorsey said.
"A lot of the goals of machine learning and deep learning is to write the software itself over time so a lot of entry-level programming jobs will just not be as relevant anymore," Dorsey told Yang.
Dorsey also told Yang that he belives a Universal Basic Income could give workers "peace of mind" that they'll be able to "eat and feed their children while they are learning how to transition into this new world."
Entry level jobs (Score:5, Insightful)
You need people to start in entry level jobs so they can progress to intermediate and senior level roles. Where will those experienced people come from once the population ages out?
Re: (Score:2)
You need people to start in entry level jobs so they can progress to intermediate and senior level roles. Where will those experienced people come from once the population ages out?
A lot of people start in entry level jobs and remain there for their whole careers.
Others can become intermediate and senior pretty damn quick.
Assuming AI does really start to master software dev (a big 'if' in the near term) there's still going to be some work for talented but less experienced developers, but the level of minimum competency required for someone to be effective is going to keep rising.
The other part of that is if software gets easier to write is also gets cheaper, which means people will bu
Re:Entry level jobs (Score:5, Interesting)
I'm posting AC because my username gives away my real name and I'm about to say things that could reflect negatively on employers of mine - past and present.
I agree with you and I have no interest in "gate keeping." However, I really want the barrier to entry to be higher than it is today.
I worked for a consulting agency that chased "big fish" clients and would bill them through the nose. Some consultants charge by the hour - which is probably a fairer model (though that can be cheated too) - but this client billed per day per head. That incentivized them to pad teams with as many extremely-wet-behind-the-ears bootcamp devs as they could. As long as there was one or two experienced architects to the carry the team, they could completely rape the client on invoices. When those experienced engineers complained that they were doing all the work on top of trying to train these juniors, the response was exactly what you just said: "How are they supposed to learn if we don't give them a chance?"
This company even founded a bootcamp of their own in order to recruit these people who had no business being employed (YET). Many had potential and I wish every single one of them the best success as their careers progress. I just think that we were doing a disservice to our clients, a disservice to the more experienced engineers and possibly a disservice to those juniors as well because burnout is real and no one wants to feel out of their depth.
I also say this as someone who LOVES to teach. I get a tremendous amount of reward from passing my knowledge on to the next generation of devs. I enjoy training juniors. I just want them to come in a little bit more prepared than we are currently accepting.
I never went to college. I taught myself to program in the 90s starting with Basic, then I wanted to learn C/C++ and, needing tools and compilers but having no money, someone suggested I look into Linux and my life was forever changed. Back then we didn't have udemy and Youtube and the Internet that we have today. I learned from books, open source communities and a hell of a lot of trial and error and crappy little games that I would write. I wonder how much my journey would have been expedited had I had access to stackoverflow, github and the abundance of online courses that we have today.
If AI can increase the barrier of entry, then that will mean that people will have to work just a little bit harder before someone will be willing to pay them high salaries for not being competent. It will mean that intermediate and experienced devs won't have to feel as often like they're carrying the weight of others. It will mean that consultants won't be able to swindle their clients as much by padding teams with heads that aren't really doing much.
Re: (Score:3)
If AI can increase the barrier of entry, then that will mean that people will have to work just a little bit harder before someone will be willing to pay them high salaries for not being competent.
AI won't magically make them work harder, if it raises the barrier of entry it would just reduce the number of developers overall. There is already significantly more reward for those who exit college with marketable skills compared to average college / boot camp graduates, so any incentives are already there. We simply don't have enough people who can meet those standards.
And being able to meet those standards is a hard thing. Time doesn't have that much to do with it either in my opinion. Most of the peop
Re: (Score:3)
if it raises the barrier of entry it would just reduce the number of developers overall.
It is far more likely that AI will lower the barriers to entry.
An AI system that looks at code and recognizes patterns that may be errors or non-portable or violate coding standards, would help entry-level coders to be more productive and learn faster.
I use Grammarly to help me write English text. It catches errors and suggests better phrasing. A "Grammarly for code" may do the same for programmers.
Re: (Score:2)
What I can't tolerate, however, is billing out someone who has less knowledge than a college sophmore at thousands of dollars per DAY to a client
Ha, then don't go into consulting. At least not with a big shop; that is their business model. It sucks, but then again consulting companies provide a large number of entry level jobs which might not exist if they didn't have such shady business practices. They probably increase the number of mid-range and senior developers our industry eventually has because they do this. I'm not sure if it is a net benefit to the industry though, just playing a bit of devil's advocate.
Re: (Score:2)
Simple: more time in school.
The number of years of schooling is have been increasing everywhere in the world, which no sign of stopping.
https://ourworldindata.org/gra... [ourworldindata.org]
Of course, school is not real life experience. But although I don't have any hard data to back it up, it looks schools and private companies are more and more connected. It is often criticized as companies are essentially asking schools to make better employees instead of focusing on general education, but it is a way to bridge the gap when
Re: (Score:2)
Re: (Score:2)
Twenty five years ago entry level people were fleshing out hand rolled web servers and ERP systems. We were testing and deploying load manage to software
Re: (Score:2)
You need people to start in entry level jobs so they can progress to intermediate and senior level roles. Where will those experienced people come from once the population ages out?
While I'm not a huge fan of the consulting industry, one thing they do well is provide a large number of entry level jobs. I have worked mostly for medium sized companies, and we rarely have much use for junior developers. I usually will only have a few employees with expertise in any given technology stack, and work which can be done by junior employees is often sparse. And what they can do generally requires more supervision than is worth the cost of paying the junior employee.
There is a big tragedy of th
It can only mean one thing (Score:5, Insightful)
It's time to start developing AIs that supervise other AIs to make bosses, managers and directors redundant.
Re:It can only mean one thing (Score:5, Funny)
Re: (Score:2)
Indeed. While it is true that most programmers are mediocre or bad, most "managers" are just bad. Keep the 10% good ones and you are all set. Problem with that is that many CEOs and other C-levels clearly are in the "bad" class and they will never fire themselves for incompetence, as richly deserved as that would be.
Re: (Score:2)
Why "ex"-programmers? You need someone to translate business babbling to something a program can accept -- but that program gets replaced every new generation. In paleocene, this was done by manually flipping switches. Then by writing a program on paper then passing it to a card punching lady (mere programmers were not allowed to see a computer). Then you had a typing machine and a line printer -- and a compiler! Then the languages kept improving.
That AI is just another step. And it will misunderstand
Re: (Score:3)
Re: (Score:2)
To be honest, I think that managers and directors will be replaced by AIs well before programming jobs.
At least middle management. Top level managers are likely to stay, for the big decisions and investment, and so are low level manager, as representatives for the rest of the team. But middle management, whose goal is mostly to monitor the situation, allocate budgets, and manage human resources can be "helped" a lot by algorithms. In fact it is already happening, to the point that an automated system may fi
Re: (Score:2)
You're joking, but my group just won a grant to research/write AI that will explain what other AIs are doing. Seriously, you can't come up with this shit even if you tried.
Re: (Score:2)
You're joking, but my group just won a grant to research/write AI that will explain what other AIs are doing. Seriously, you can't come up with this shit even if you tried.
I admit, it was a bit of a knee-jerk response... I don't think developers are all going to get made redundant, some yes, when a company is bad at managing their workforce. Anyway, we are still at the start of AIs and let them do singular tasks, but it will come to the point where we use AIs to manage AIs, because of the time it takes to train them will we need more than faster hardware, but we need to find ways to take shortcuts in software and still get the result we want.
Re: (Score:2)
Compilers already eliminated most programming jobs (Score:3)
Most programming jobs have already been eliminated.
It used to be that you needed to program computers by entering the ones and zeroes manually. Then compilers were invented so that you simply enter a few instructions about what you wanted the computer to do and then the compiler actually created the program.
Re:Compilers already eliminated most programming j (Score:5, Insightful)
Re:Compilers already eliminated most programming j (Score:5, Insightful)
That's right, we still need the specification in some form. If any AI writes the code, then the specification complexity just increases to the point where it is so meticulously detailed that you may as well have had a human write the code.
Either that, or you can have "AI"s (term used loosely) produce cookie-cutter apps that have a well defined set of input parameters. But then it's just a fancy templating system.
Re: (Score:2)
If any AI writes the code, then the specification complexity just increases to the point where it is so meticulously detailed that you may as well have had a human write the code.
Either that, or you can have "AI"s (term used loosely) produce cookie-cutter apps that have a well defined set of input parameters. But then it's just a fancy templating system.
Will AI every be able to write code, in terms of producing source code? Or is this speculation that the code functionality will be replaced by AI systems? I'm not sure a path to the former for appreciably complex code has been yet demonstrated. If it's the latter, then isn't this simply the continuation of the evolution that we've seen for decades, where automation replaces the simpler and more rote human activities and human activities move up the software stack, including creating yet higher abstractio
Re: (Score:2)
It seems like if you specify things narrowly enough you already have the source code! You don't need AI, you just need a compiler.
Re: Compilers already eliminated most programming (Score:5, Interesting)
That is precisely true, but the reality is that person will be called a mathematician. I have thought a lot about this one theoretically, and what it must come down to is someone generating a program such that x,y,z, and the x,y,z the program satisfies are mathematical translations of the customers wishes. Only with pretty trivial stuff like vanity websites will AI be able to translate customer wishes straight to product.
Re: (Score:2)
Re: (Score:2)
But that's not how programming works. The point of having higher level programming languages is precisely to avoid having programming be math. At the bottom level it's all just binary mathematics. But we create abstraction systems to avoid the programmer having to spend their time doing that.
If you follow only the existing trends in programming you will EVENTUALLY get to the point where programs are being written by software, given a functional description in your language of choice. I realize that this has
Re: (Score:3)
If you follow only the existing trends in programming you will EVENTUALLY get to the point where programs are being written by software, given a functional description in your language of choice.
That's called "a compiler", by the way, the bit that writes the software given a functional description. People seem to be unclear on this, for some reason. The part of software development that is the actual job is setting down a specification unambiguously in some sort of language. The hard part is "unambiguously", not "language".
The only real question to be asked is how long it's going to take for most programmers to be replaced with software.
This has happened continuously throughout the past 50 years or so. The tools get better, and whatever was the tedious repetitive part 20 years ago is done by software today,
Re: Compilers already eliminated most programming (Score:2)
Re: (Score:2)
maybe 2. He'll be replacing entire teams of well paid engineers.
Hopefully
Re: (Score:2)
I've never even been a professional programmer, but I've spend serious time writing code in the past, and could do it again if I saw the need -- and my experience goes all the way back to the 'entering ones and zeroes manually', both in the COSMAC ELF microcomputer trainer I built as a 15 year old, and the IMSAI 8080 I used to own back in the day -- and then Macro Assemblers and 'C' compilers. So I think I'm at least qualified to sa
You got it backwards. (Score:2)
Back when programming was lower-level and more difficult, there were fewer programming jobs. Demand for programmers was lower.
When higher-level compilers (like, FORTRAN, COBOL, C etc) hit the market, the result was an increase in the demand for programmers.
This is commonly observed in economics: technology that increases the efficiency with which a resource (including human resources) can be utilized results in an increased demand for said resource. Also, when the cost of something goes down, demand tends
Re: (Score:2)
Did they? So far, the compilers along with better hardware just expanded what you could do. Now that it's possible to write more advanced applications with high level languages, they hire more programmers and do exactly that.
It's not like farming where automation really did replace workers. You generally don't plow or plant in some fancy new way because of automation, you just do it with fewer workers. Compilers, OTOH, opened up whole new ways of doing things, as if it were now possible to plow the sky
No, it won't (Score:5, Insightful)
While you can easily use AI to replace grunt coding tasks and established models - these things already exist in the form of "automated programming tools" that generate the code from some basic configuration settings.
The problem is these tools only work for the one task/framework these were set for. if you deviate from that path (EG by being creative) the tool breaks down and/or more commonly the tool is just overrun by newer processes, methods and techniques.
Further we've tried going down this road multiple times before. Everybody seems to have forgotten Rational Rose which was a tool where you "wrote code" using UML shapes and diagrams resulting in automated code generation of in the language of YOUR CHOICE with no coding needed. Which, as anybody who's used the tool will tell you, is BS.
That doesn't mean AI can't do code generation - good lord, as a cloud programmer i would kill right now to have an AI to autogenerate cloudformation scripts for me (and those are known quantifies). You can even get an AI to write "Hello world" - but you're not going to get a Star Trek Utopian AI where you say "Hey, Siri, write me a AAA video game that I can sell for ONE MILLION DOLLARS"
Re: (Score:2)
We had something that did up BPML into an ETL XML format that ran on a Java OSGI modular system. It was actually quite robust. Rational Rose was... alright-ish, but yeah it oversold the hype biggly. But hey the same crew worked on the PCML for service programs on the System i and PCML has been banging for COBOL and RPGLE based services to bring them into a web container. So, yeah, their idea wasn't more hype than product but they did go on to do some pretty neat stuff. I recently moved some logic built
Re: (Score:3)
Nonsense (Score:2, Flamebait)
UBI is a pro-billionaire lock-in fantasy scheme built on fear.
AI will likely do jack ... (Score:3)
All "AI" has ever done is create a massive bubbles, slightly improve the state of the art in signal processing and classification and then collapse. Assuming technological civilization survives long enough eventually it will create true AI, will this time be the one? Doubtful.
But hey, the slight improvement in classification is nicely timed to coincide with the increase in mass surveillance ... so at least this time the investors in the bubble will see a bit better return on investment.
Re: (Score:2)
Re: (Score:2)
fat chance (Score:5, Insightful)
Smart humans can barely code. An AI that can code as well as a human would be a threat to our species' existence, and not some rich guy's money making device.
More Silicon Valley microdosing nonsense.
Re: (Score:2)
Smart humans can barely code.
So you're saying you're an unemployed programmer?
Once we have replicators we wont need cooks! (Score:5, Insightful)
Maybe someone with more knowledge of AI development can correct me, but this seems so far out in the future to be meaningless to make policy about today. Among all the jobs that will be threatened by automation I cannot imagine programming and software development are anywhere near being replaced (and I am not a programmer myself). I strongly suspect AI will play a large part in how programmers work and there will be integration into programmers development environments and tool-sets. That will make programmers more efficient and productive, but not put them out of work. With the difficulty that has come with teaching an AI to drive a car having one develop and efficient well thought out software application, especially one that would be used by humans is decades away. Much less all the programming work needed to build and maintain those AI's.
Re: (Score:2)
Re: Once we have replicators we wont need cooks! (Score:2)
I'd say we're at least 50 years away from anything even close to that because we haven't got a clue currently how 'thinking' or 'reasoning' even works in a biological brain and won't for quite some time to come. May not even be possible to create a machine that can do what our brains can do for all we know.
The amount of computation required using current methods is staggering. It takes data center sized computers to emulate a worm's 80-neuron brain, let alone the umptillions of neurons in a sapiens sapiens brain.
However, we are much closer than 50 years. There are 2 known solutions to this paradox: the first is quantum computing, which I personally think is a pipe dream or certainly more than 50 years off. The second is borg-like technology. We can already do it invasively but I believe EEG technology can a
Re: (Score:2)
Kurzweil estimates this will happen sometime around 2070-2080. So programmers are safe for a while.
Why pay Java dev $18K/year? (Score:2)
When you can spend $1M a month on AWS compute resources to write your boilerplate copypaste software projects?
Re: (Score:2)
Re: (Score:2)
Java devs at my company make six times that or more, and we do have Azure cloud project going. That has quite a few devs plus project manager, DBA... seems to me it's making work not replacing people. All my life IT has been giving people jobs, almost all jobs at my employer involve a computer as a tool, and then we need people to maintain and deploy the tools, the infrastructure....
Not seeing this B.S. where automation costs jobs, except in nonsense studies that aren't really counting all the jobs the
It's the level of abstraction, stupid (Score:5, Insightful)
Anyone remember "automatic programming"? Or some of the talk about anyone being able to program in COBOL because it was so much like English? Old ideas, but maybe they were just too early and now the times have ripened.
My own label was "level of abstraction". It keeps going up, and what that means is you can get more actual work done with fewer lines of code. Or you can think of it in qualitative terms of changing the kinds of problems you can solve with apps. People don't even want to call them programs anymore. The focus is on what it does, whereas in "the old days" the focus was on getting it to run at all. (But if it was a REAL program the debugging never ended!)
From the AI perspective, I don't see it as a general AI thing. A good IDE is more of a special-purpose AI, and we're becoming extremely good at building those. If we can define a human skill clearly enough, from playing chess to ping pong, then we can build a computer to do it better. So far programming has been more of an art than a science, but if you change your level of abstraction again, maybe that isn't true for most programs these days. Lots of "daily use", even "one off" programs that solve interesting problems these days.
Recent anecdote. I have a little database with two relational tables. I decided I wanted to do a new kind of analysis on it. I spent a couple of days considering algorithms, but for a long time all of the algorithms appeared to be O(n^2), which looked nasty, so I didn't even start programming. Then I figured out a better algorithm that was O(n), but I realized I could just do that with a simple spreadsheet. Painting the numbers got the results I was seeking without really needing to uncoil the Python at all. (But the results turned out to be rather disappointing, even boring.)
"Times, they are a changing." I think we need to completely rethink economics in terms of time. Unfortunately conventional economists insist on prattling on about money because they think they can measure it, while time is much trickier. Time = money is a lie. The not-so-new reality is time >> money.
Metacomment on ye olde FP problem: I deliberately wrote this comment with the intention of stimulating an actual discussion of some actual social problems, but that took a couple of minutes. There were already two twatshots posted when I started, and there's been time for half a dozen more. If one of the objectives of discussions on Slashdot is to make people think about the stories, then I think those kinds of comments should go to hell or Twitter, whichever is worse.
Re: It's the level of abstraction, stupid (Score:2)
Re: (Score:2)
Anyone remember... the talk about anyone being able to program in COBOL because it was so much like English?
No, though I don't doubt some moron writing for some magazine said something that dumb.
What I do remember reading about in the history books was people claiming that COBOL was so much like English that an accounting manager could learn to automate tasks with it, and a stock broker could learn to use it too.
Educated professionals expected to be in the upper half of the competency curve for their profession, but who are not normally computer people. That is who was supposed to be able to learn to write it. No
Re: (Score:2)
Metacomment on ye olde FP problem: I deliberately wrote this comment with the intention of stimulating an actual discussion of some actual social problems, but that took a couple of minutes. There were already two twatshots posted when I started, and there's been time for half a dozen more. If one of the objectives of discussions on Slashdot is to make people think about the stories, then I think those kinds of comments should go to hell or Twitter, whichever is worse.
I read all your post, and none of the trolls, and I found your post valuable and it was highly voted. Thank you for taking the time to write it. I found value in it.
Re: (Score:2)
Yeah, and SQL was supposed to be a tool where people with little programming experience (managers) could use to produce reports without needing to go to the Information Systems (IS) department ...
How well did that work out?
Hell, even today, we have DEVELOPERS who write queries that join 8 tab
Re: (Score:2)
That has not been my experience. I manage servers for clients, and developers regularly create such queries, and I have to tell them to cache the results or rewrite them, ...etc.
One reason is that th
I only wish (Score:2)
After decades of computer programming, how much closer are we to the point where the programmer can spend a great deal more time thinking about architecture of a program than coding?
Not really that far at all.
I would LOVE for AI to come along and replace the most tedious aspects of programming, but so far it can't even do that well.
Where AI could really shine is in assisting programmers, I wonder if the reason it has not advanced more is they are tarting replacement rather than assistance...
Re: (Score:2)
Full Asses Only (Score:2)
You really want the half-assed excuse for 'AI' they keep trotting out to everyone to be writing the bulk of your code for you?
I want a competent version of that, that I feel like is possible but we have not seen yet.
I want to see fully-assed coding AI, I ag ree I want none of the half-assed stuff we have seen to date (which is why I pretty much always avoid code generators too).
Re: (Score:2)
You'd probably spend more time debugging the nonsense output from the damned thing than you would just writing it yourself.
That would be true if they were using some sort of expert system, but they're just "training" a black box. You're not going to debug shit. Life is not that simple if a mainstream contemporary AI wrote your code.
Re: (Score:2)
I can't even DO architecture anymore. I use to labor for days over design, code layout and how to setup the objects for C++ so the design would just fall into place. As a cloud programmer now doing severless microservices my choices are javascript or python and slathering some scripting code onto an IDE to make a few calls. I can still get some structure in there which makes my code better organized and adaptable but that's only a few minutes of thought (if that).
More of your architecture/design is pulle
Re: (Score:3)
After decades of computer programming, how much closer are we to the point where the programmer can spend a great deal more time thinking about architecture of a program than coding?
I spend way more time on architecture than implementation, or testing. 90% of a quality solution that won't require continuous updates is architecture.
I think they call it "waterfall process." Plan first, then do. If you need to make a change, stop; change the plan to include the change before trying to make the change. Start over if necessary to implement the new plan correctly.
The process works for roads and bridges, too; in fact, in structural engineering planning is mandatory! (Golly, I wonder why! /s)
Fuck Dorsey *and* Yang (Score:2, Troll)
What do they think they're accomplishing? We're still in the middle of the Crisis of the Century, too many people are down to one badly-frayed nerve, we're approaching record levels of national unemployment, and this rich jackass and his jackass buddy are spreading around more "robots are going to take all your jobs!" bullshit? Really!?
Yelling "FIRE!!!" in a crowded theatre showing the original Invastion of the Body Snatchers (with the original ending that frightened people half to death)
Someone should punch this Dorsey character in the
Never Stated: AI Will Replace Management (Score:5, Insightful)
Re: (Score:3)
But execs can't talk about feeling threatened ...
Well, they don't have to talk about it, because they are the ones doing the threatening and they know how to respond to it. But that too is what AIs can do, and are already doing in some gaming bots. Threat management is just another task for AIs really.
Nobody then likes it when jobs are being made redundant and Jack Dorsey is just stating more or less the obvious. Because once AIs can make people redundant will some companies not simply assign these people to new tasks, which is the obvious thing to do, an
Re: (Score:2)
Re: (Score:2)
If anything, that would be easier to mechanize
Interpersonal issues are easy to mechanize with AI? Please, tell me the secret.
Re: (Score:2)
Interpersonal issues are easy to mechanize with AI? Please, tell me the secret.
No, they are easy to remove, because many of these issues come from having a manager in the first place. A manager doesn't somehow stand on the outside of it. The things that go on in these relationships range from a simple smile to dodging, small-talk, rivalry, flirting, brown-nosing, revenge and even "sex with the boss" isn't unheard of. We do these things more often out of idleness than out of necessity. We also know what will happen once you replace someone with a machine. So do we now have automated te
Re: (Score:2)
Have you ever tried to complain to one or to reason with it? It's pointless and you'll learn it's just easier to follow the instructions and get on with it.
I yell at at and push buttons until a human answers.
Re: (Score:2)
I yell at at and push buttons until a human answers.
I know the feeling, but I can get just as mad at the people on the phone to be honest.
Me: Obediently follows the machine's instructions to the end, gives name, address, post code, then a human picks up ... ...
Human: "What is your post code?"
Me: "Why do you ask me for my post code? I just told the machine my post code! Do you not see this on your display?"
Human: "No, we don't get to see this information."
It's however not something I can blame the machine for. This was a design flaw, made by a human.
Just as soon (Score:2)
as it fixes Twitter, right Jack?
AI coming? (Score:3)
This is more difficult than he believes. (Score:3)
The problem with this proposal is that most specs I've seen for was an application should do are unreasonable and unworkable. You need to talk to the person/people requiring the work and find out what they really want rather than what they tell you off the top of their head. After you get that pinned down, then I guess it's theoretically possible for an AI to do it if anyone could (and that isn't always possible).
There's more than one reason for the joke about the "do what I mean" instruction.
They have said that for 40 years (Score:2)
Well, not the AI part. but I remember how in the early 80 at school how by 2000, we would have to make up jobs so that we could feel useful and people would have fake jobs. :D
Well, all that has happened are that complexity the speed of development of everything has gone up.
In the programming world, very few are doing opcode programming anymore, we keep adding layers of what we are doing and we will keep doing that.
This has been happening for years (Score:2)
It isn't just robots that have replaced workers. AI's have already been replacing workers in jobs for years. I personally know of one company that laid off thousands of workers after replacing them with AI (NDA applies). People used to joke about replacing their job with a 7 line perl script. Turns out that some of those people are now actually working really hard to do just that - and succeeding.
Here's the rub, as people get better at automating simple jobs they will gain the skills required for slightly h
Programming is figuring out what to write, not how (Score:3)
I have seen the performance of AI for computer security and attack analysis. It is a great supplement, but not a replacement for something built by humans. I think people don't realize the limits of AI. AI can't even 100% detect dicks and breasts in profile photos...how will it understand business requirements? 95% of my job doesn't involve writing code or syntax, but figuring out what to write and working through the edge cases. Java is not a productivity killer for me, life and business are. 95% of my day is figuring out what to write, not how to write it. Even with perfect AI, I'd still have to do 95% of that work.
I am confident business programming jobs are not going anywhere in 20 years.
Re:What is his agenda.... (Score:5, Interesting)
Our developers he is speaking about are most likely are the VBA programmers that are trying to automate and solve tasks
Yeah, if you listen to the podcast that's explicitly what he is talking about. Except that also includes things like EDI, ETL, task automation, scheduling, provisioning, and so on. There's a lot of jobs that's based on rote code that needs a few tweaks here and there to overcome the difference between the normal and irregular. We had a piece of software that did ETL between BI units. We used a BPML generator to produce boiler plate by passing it through XSLT. When we moved that to a CI service and we could have floor engineers just do up the BPML and then have the service create the code required to generate the data cube that they needed for a particular business unit. We had jobs specifically for data scientist that would do up the python and SQL required to provide their business unit with the data they needed, once we transitioned fully to that, all their jobs were done. We kept maybe two or three of the best and they handled the last 10% between all the business units that the system couldn't auto generate the boiler plate. So that was like 20ish jobs there gone.
Re:What is his agenda.... (Score:4, Interesting)
Replacing code like you describe is still very difficult for an AI to handle. That's because the hard part of the task is to exactly specify the job that needs to be done. And you'll soon discover that the process of describing these jobs in exact detail is basically equivalent to doing the job. The programming in these cases is the easy part, the hard part is in knowing exactly what you need to do. In this case the AI may provide a better tool. You could eliminate maybe half of the workers and correspondingly make the more skilled people more productive. But there is no special magic AI sauce for this type of problem, a new non-AI based tool could easily have the same effect.
To eliminate skilled programmers we need to develop "General Intelligence" AIs. So far attempts to do this have failed.
Re:What is his agenda.... (Score:4, Interesting)
Re: (Score:2)
Yes and no. I see your point, but at the same time I think of the last time I did that and the judgement calls involved in translating the HTML/CSS output from the designer into HTML and (S)CSS which was semantically structured and maintainable.
Re: (Score:2)
Re: (Score:2)
You have to create a CMS before the designer is even brought in, and then force the designer to design everything through your CMS interface.
Then you can successfully automate everything, and there is nothing they can do to thwart you!
Re: (Score:2)
Re: (Score:3)
Yep, AI can replace bosses far more easily that it can replace programmers.
People have been saying that programmers will soon be obsolete for as long as I can remember.
Anybody else remember this?
eg. http://www.tebbo.com/presshere... [tebbo.com]
Re:REEEEEE!!!! (Score:5, Interesting)
That was pretty much my first thought as well. This bullshit has been around for half a century or longer. When I was at university, the "5th Generation Languages" had just failed to deliver anything useful on this subject and that was 30 years ago.
As "AI" (in reality, cognitive systems and artificial stupidity, there is not even a glimmer of intelligence in machines and it is not a problem of speed or size, it is a fundamental problem) has basically not made any progress since then, besides doing the same mindless things faster, there is absolutely nothing to this bullshit.
There is another threat to the average programmer though: Somebody with enough clout may finally find out that mediocre and bad coders are exceptionally expensive and that if you use good and exceptional coders you actually do not need that many of them. And then many of these jobs will go away. But with the current prevalent stupidity that coding is easy, that may still take a while.
Re: (Score:2)
With the current amount of bad talent in the workforce I'm nearly certain AI can solve our problems.
It is at least superior to some guy who submits questions in a PR.
At some point, the AI will get it right....
Re: (Score:2)
Re:REEEEEE!!!! (Score:4, Informative)
At some point, the AI will get it right....
Hahahaha, no. That is exceptionally unlikely at this point.
Re:REEEEEE!!!! (Score:5, Informative)
If anyone doubts it, I invite them to read Doug Hofstadter's book "Godel, Escher, and Bach".
We haven't the slightest idea how to make anything resembling real "AI". Not even a clue.
Yes, we have some programs that do some pretty impressive things. All written by human beings. Sometimes by teams, but still by people.
We would need a completely revolutionary paradigm shift before we would even have an inkling about how real AI could work.
And nobody even knows what kind of paradigm shift it would be.
These "machine learning" things are good at sifting through data, finding relationships, and so on.
But they're not even much good at distinguishing cause-and-effect from spurious correlations.
Re:REEEEEE!!!! (Score:4, Insightful)
If anyone doubts it, I invite them to read Doug Hofstadter's book "Godel, Escher, and Bach".
GED is a wonderful wonderful work of literature and philosophy, and a very entertaining book. I highly recommend it.
But GED will teach you absolutely nothing about modern AI. It isn't going to help you understand TensorFlow.
Re: (Score:3, Insightful)
It will teach you that it doesn't exist.
And TensorFlow is not even remotely "AI".
You're using the term too loosely, Bill.
Re: (Score:2)
Yes, it will teach you about modern AI.
It will teach you that it doesn't exist.
AI is a building. At the University. Sometimes it is near the Computer Science buildings. Sometimes not. Sometimes it just is one room in the basement of the Computing Center, where non-CS students go to use computers.
Re: (Score:3)
It was a buzzword, for selling software.
But it's a lie. It's 100% incorrect. As professionals, people involved in the software industry should stop using it for the kind of software we have today.
I don't mean to rant. But what if -- a very big "if" -- some real breakthrough came around that actually might pertain to AI? It would make the current popular use of the word look embarrassingly silly. That's an under
Re: (Score:3)
I guess I should say that the industry has been using "AI" much too loosely. For decades.
After a generation or so, the common misuse becomes the proper use. Sorry, the current BS is just what "AI" means now. And machine learning is useful for a bunch of stuff that you can't figure out how to code explicitly.
Re: (Score:3)
Real artificial intelligence?
Something else?
Re: (Score:3)
I've always called it "machine intelligence" ("artificial" sounds a bit pejorative). though that sounds too much like "machine learning" these days, so maybe "machine sapience".
Thing is, the stuff we actually have is exactly what "AI researchers" have been working on for the past 50+ years. Speech recognition. Machine vision. Handwriting recognition. It's not total misuse of the term if it's what AI researchers actually research (almost none of them are working towards actual machine intelligence, but i
Re: (Score:2)
TensorFlow is just an optimization. It does absolutely nothing that was not possible before. Making machines intelligent is not a performance problem or we would have had (slow) intelligent machines a long time ago.
Re: (Score:3)
Making machines intelligent is not a performance problem or we would have had (slow) intelligent machines a long time ago.
In a way it is. With enough brute force, you can create intelligence by fully simulating a human brain at the quantum level using well-tested physical equations.
Re: (Score:2)
I've been saying this, here on Slashdot, for many years.
And it still hasn't happened.
Re: (Score:2)
Re: (Score:2)
As "AI" (in reality, cognitive systems and artificial stupidity, there is not even a glimmer of intelligence in machines and it is not a problem of speed or size, it is a fundamental problem) has basically not made any progress since then, besides doing the same mindless things faster, there is absolutely nothing to this bullshit.
"AI" has come to men the BS we have now. The term has gone the way of "hacker" I'm afraid. I kinda like "Machine Intelligence" to denote strong AI: human-style intelligence, just hosted in a machine. Or perhaps "Machine Sapience", since few enough people use that word it makes good jargon.
In any case, even if we did somehow invent Machine Sapience, what the heck makes anyone think the new person would be interested in a coding job? I mean really, the BS we have now is no threat to dev jobs at all: it ju