Catch up on stories from the past week (and beyond) at the Slashdot story archive

 



Forgot your password?
typodupeerror
Programming

Google Research Chief Says Learning To Code 'as Important as Ever' (businessinsider.com) 58

Google's head of research Yossi Matias maintains that learning to code remains "as important as ever" despite AI's growing role in software development. While AI tools have reduced coding time for some developers -- and Alphabet CEO Sundar Pichai noting that AI now generates a quarter of all code, Matias stressed that human engineers still review and approve AI-generated code.

The Google executive, who also serves as a company VP, acknowledged that junior professionals have faced challenges gaining experience as AI handles entry-level tasks. Google has launched initiatives to support early-career employees through this transition. Matias compared coding literacy to basic mathematics, arguing it provides crucial understanding of technology regardless of career path.
This discussion has been archived. No new comments can be posted.

Google Research Chief Says Learning To Code 'as Important as Ever'

Comments Filter:
  • Code reviews (Score:5, Insightful)

    by sunderland56 ( 621843 ) on Monday November 11, 2024 @01:43PM (#64937251)

    Ask any software engineer whether they'd rather write code, or do code reviews.

    So Google's master plan is to have AI do the fun stuff, and humans do the drudge work? Good luck with that.

    • Re: (Score:3, Insightful)

      by Luckyo ( 1726890 )

      That's how reality has worked so far with all engineering jobs though. Google what mechanical engineers, electrical engineers etc did in 1970s. Then google what they did in 2000s.

      This has simply now come for software engineers. We are going from our equivalent of hand drawing everything from scratch to giving autoCAD directions and vetting the results.

    • Ask any software engineer whether they'd rather write code, or do code reviews.

      So Google's master plan is to have AI do the fun stuff, and humans do the drudge work? Good luck with that.

      This is the joy of automation. Take away the fun bits and leave humans to do the tedium and deal with the business bullshit. God, what a glorious future.

    • No, as a developer, I do *not* want to do the grunt work. This is the stuff AI takes off my hands, not the fun stuff (which is seeing the software take shape as I write it), but the annoying stuff

      As an example, I wanted to improve a search algorithm by using a Lowenstein distance calculation. I had no wish to actually write the function to perform the calculation, I was quite happy that Copilot could write it for me, and I plug it in to the larger system.

      This is the stuff AI will actually take on, the grunt

  • by Baron_Yam ( 643147 ) on Monday November 11, 2024 @01:51PM (#64937267)

    Coding is great for teaching project management, task ordering, how to dissect a problem into solvable components, etc. it's not the only way to learn those things.

    I think coding's valuable and should be encouraged, but it is not something we need to force universally.

    • by will4 ( 7250692 ) on Monday November 11, 2024 @04:14PM (#64937731)

      Agree 100%. We need to teach people basic life skills first instead of jumping to learning to code.

      - Compound interest and its long term effects of borrowing too much.
      - Cost benefit analysis of getting a college degree(s) , effect of borrowing money to do so (compound interest) and long term financial impacts
      - Balancing a checking account / financial account
      - Making and following a monthly budget
      - How to cook (something easy) from scratch - following instructions
      - Budgeting for food, making a grocery list, planning meals for a week
      - How to engage in a reasoned discussion without disruptive behavior and without using faulty or self-reported research
      - Critical reading, fact checking skills and how to identify agenda based news articles and agenda based scientific research
      - How to comprehend that a politician's media speech, statement is not the same as a law passed, policy implemented, or actual results
      - How to self-regulate behavior, be it compulsive shopping as recreation, gaming, social media use, binging entertainment, attention-seeking behavior, etc.
      - Knowledge of actual repressive forms of government and a relative comparison to your country's actual form of government

      • by lsllll ( 830002 )

        Damn, that's a good list!

        I'd like to slightly disagree with planning meals for a week. My modus operandi when it comes to shopping for groceries and planning for meals is go to the store and see what's on sale, and get those items and make something with them. For instance, it's been a whole month that I haven't had steak because rireye hasn't been on sale. Meanwhile last Friday they had ribs for $1.99/pound. So I cooked 3 slabs for 5 adults and half of it is still in the fridge. Total cost for the mea

        • by will4 ( 7250692 ) on Monday November 11, 2024 @05:07PM (#64937887)

          - A productive skill not using computers, electricity or expensive power tools. Working with hands, planning it out, having the patience to follow through
          - How to incorporate 10 minutes of basic exercise into your day without require electronics, entertainment, or other distractions - walking, stretching, free exercise
          - How to organize and pack a suitcase with 1 week's clothes for a visit to a place moderately cooler or hotter than where you live without buying new clothes
          - How to estimate distance, scale, time, cost, etc. without guessing and without resorting to calculations
          - How to give effective pirate map style directions to someone else to walk from point A to B (spatial relations use)
          - How to converse with someone else without trying to elicit a gotcha phrase from them
          - How to read instructions without visuals, without video, without audio understand them and use them to complete a task
          - How to fill out by hand the most basic 1040 EZ US tax form if you have one job, have $500 in bank interest
          - How to do a simple repetitious task while holding a civil conversation with other people doing the same task. For example, the old time cleaning the rocks and dirt from a bag of lentil beans with 2 other people.
          - How to organize, sort and file away monthly paper statements. Concept of directories, folders, files, ordering by date, etc. even if they are just electronic PDF statements.

          • by will4 ( 7250692 )

            The general idea is that most basic life skill tasks would be done without the person 'needing' to be stimulated via music headphones, video in the background, etc.

            Probably the largest life skill missing in people today, existing with periods of low stimulation.

      • If your talking about the young, I would say the first is flexibility and quickly learning unrelated skills. Contract is becoming king i.e. job stability is dying. Budgeting is great, but it may be next to impossible, if your always looking, or transferring jobs. It's easy if you 1. Make enough to live a comfortable or semi-comfortable lifestyle and 2. are in a stable position, but not if you make less than a "living wage" and/or are always transferring jobs it's not so simple. We've seen "permanent" jobs b
    • I think too many people do it because they think it pays well, which it certainly can, but when they're that kind of person they tend to suck at it and likely end up working for some kind of code mill, the pay is kind of meh (maybe $50-$70kish) and they'll probably hate life.

  • Wrong problem (Score:4, Insightful)

    by RobinH ( 124750 ) on Monday November 11, 2024 @02:02PM (#64937287) Homepage
    I'm not sure why people thing coders spend their time typing. Sure, there's a bit of that, but mostly they spend time thinking and designing. Typing is the relatively small but relaxing part of the job.
    • I rarely code these days, but when I did my best stuff happened while I was driving between sites. The tricky bit was remembering what I'd thought of until it was safe to implement or at least take notes.

      • My best work often occurred in beach parking lots. I'd take a stack of papers, blank pads, and pencils with me. Salt air, surf sound, and no yammering or ringing telephones, helped a lot. "Pictures with circles and arrows", "and a paragraph on the back" after returning to a keyboard.
        • You all sound like frauds. My best ideas have happened when I had my brain engaged and I was sitting in front of the system coding my ass off. Why? Because I've spent fucking man-years doing that. Practice makes perfect, not decompressing on the beach or some other time-wasting horseshit. One learns and improves by doing. I have the code, the finished applications, and the resume to prove it: not a bunch of yellow legal pads with dumb looking drawings.
      • by lsllll ( 830002 )

        The most elegant solutions to problems I have come up with have been either the result of rolling in bed, trying to fall asleep, or towards the morning when I've woken up with an epiphany. It's like the harder I work at a problem, the more elusive it is. Maybe it's the lack of distractions around sleep.

        • For decades I thought it would be fun to keep a notebook on my night table to scratch down those 'middle of the night' epiphanies. I never did.

          But now I have my phone there, charging. I've used it a few times to make notes sufficient to remind myself later and revive the idea for further work.

          I recommend it. I have no idea how many decent ideas I let wither and just forgot about them because I didn't take a simple note.

        • This happened to me once, around 35 years into the field. The project had a baffling concurrency defect that appeared to be in a specific library, and we were at wit's end determining a cause. I woke up around 3:30 AM with a nagging suspicion and realizing that I needed to know more about the underlying algorithm. I went downstairs and pulled a copy of the CLR Algorithms white book off the shelf, and a listing of the small but crucial open source library in question.

          It soon became apparent that the librar

    • by Luckyo ( 1726890 )

      I suspect this is more of a "different focus" thing. When most engineering fields got autoCAD in 1980s and early 1990s, it basically removed the need for those who knew how to draw and visualize things on paper really well. Instead of focused on actual thinking and designing. But engineering jobs had a lot of people who were great at drawing and poor at thinking and designing. Because the job used to need a lot of drawing and much less thinking and designing pre-autoCAD, and almost no drawing and a lot of t

      • by dfghjk ( 711126 )

        "And after automation of coding, it's the software architects that will be of paramount importance, while people who're really good at writing code will go the way of the people who were really good at drawing."

        A false and ridiculous dichotomy based on the assumption that "coding" is simple enough to be automated. Design and coding are coupled, "software architect" and "really good at writing code" are not orthogonal activities, quite the opposite.

        The vast majority of coding is fixed function, limited reso

        • by RobinH ( 124750 )

          The vast majority of coding is fixed function, limited resource embedded programming.

          Source please?

        • "And after automation of coding, it's the software architects that will be of paramount importance, while people who're really good at writing code will go the way of the people who were really good at drawing."

          A false and ridiculous dichotomy based on the assumption that "coding" is simple enough to be automated. Design and coding are coupled, "software architect" and "really good at writing code" are not orthogonal activities, quite the opposite.

          The vast majority of coding is fixed function, limited resource embedded programming. People who make specious comments like this one do not even understand what programmers do.

          I'd like to know far more since I'm not sure where the time and cost savings are supposed to come from involving humans. I generally spend somewhat more time figuring out somebody else's code than I do thinking up my own (but that depends on prone to code obfuscation the original coder was). On top of that I spend about as much time testing my own code as I do writing it. It seems far more efficient and cost saving to just fire all the humans and train AIs to review the code. That should be (a) trivially ea

          • That should be [...]

            So you wrote all that tripe without even knowing what is. Just use the tools (they are mostly free) and find out for yourself how viable AI code reviews are. Right now the state of the art is that AI cannot write more than about 30 lines of code without major bugs or going in some completely fucked up direction that'll need refactoring. This is the case with ChatGPT, Claude, CoPilot, and Grok. If you tried to use the for code reviews they might be helpful, but not as much as an experienced coder because whe

        • A false and ridiculous dichotomy

          Quite right. It also sounds like he's lying. I have several traditional engineers in my family who are experts with AutoCAD and, while automation is better and AutoLISP can help, it's not some panacea that eliminated all design skills. He sounds like he's just making shit up to me.

        • by Luckyo ( 1726890 )

          This is why I suggested people look up how people used to build things like airplanes, power plants and so on from hand made drawings. Ever tried routing thousands of kilometers of pipes of various diameters, grades, pressures, vibration levels and so on? Every power plant needs this done. Every single one. And a single minor mistake in connecting can result in entire power plant needing massive corrective work of tearing out hundreds of kilometers of piping and replacing it with newly routed one. And this

      • I suspect this is more of a "different focus" thing. When most engineering fields got autoCAD in 1980s and early 1990s, it basically removed the need for those who knew how to draw and visualize things on paper really well. Instead of focused on actual thinking and designing. But engineering jobs had a lot of people who were great at drawing and poor at thinking and designing. Because the job used to need a lot of drawing and much less thinking and designing pre-autoCAD, and almost no drawing and a lot of thinking and designing post-autoCAD.

        Same thing will apply here. Today there are a lot of people who are really good at coding. There are far less people who can do proper software architecting at large scale, because writing the code is so fundamentally important. Just as knowing how to draw your design on paper was so fundamentally important for engineering pre-autoCAD and so much fewer people did the actual large scale design work.

        And after automation of coding, it's the software architects that will be of paramount importance, while people who're really good at writing code will go the way of the people who were really good at drawing. Either re-learn to the new focus or change profession.

        I don't think this is gonna work the way you think it is. The real dream is to kill-off software engineering as a profession and have the entire process automated from the moment some manager day-dreams pie-in-the-sky bullshit to the moment that day dream is sorta/kinda running like he imagined, bugs and all. They'll spin up another AI to deal with bug checking / testing. The upper management has been promised that software development will be as simple as sitting at a prompt and saying, "I need gobbledegoo

        • You're right. He's dreaming. It's a dream of every manager since programming was something they had to start paying for. I've seen it with IDEs, CASE method, Indian offshoring, etc... Every 10 years there is some technology or phenomenon that's supposed to make programming easy for everyone and provide a silver bullet to kill the need for heavily trained expert coders. It's never worked and I don't expect AI to change that. I use it all the time and it's still terrible at writing code that works past about
        • by Luckyo ( 1726890 )

          You can only believe this if you shove your head in the sand and pretend that autoCAD isn't real, and hasn't done exactly what I described decades ago.

          Best part? Doesn't matter if you believe it or not, or even believe it and manage to get it banned. PRC is going all in on automating this because of their upcoming population crash. If we don't get it done, they will. And whoever gets this done will be so much more effective, that struggle for supremacy will look a lot like USSR vs US in 1980s.

          Because that w

    • I don't know about you, but I thought coders type furiously all day, while getting a blowjob and having someone point a loaded gun to their head!
  • "We need an excess supply of students who want to become developers for us, so that we can hire them for low pay, and have them compete against each other in an AI-dominated workplace to increase the reliability of the tools that will eventually make redundant what few openings for humans remain."

    All the C-level execs are salivating at the prospect of making even more profit by cutting as many developers as they can. If nobody wants to become a developer because of this, if nobody wants to learn how to cod

    • Or course you think this. Only since AI has management seriously thought about reducing tech workforce.

      Well, ok, not since 2000.

    • by HiThere ( 15173 )

      You're probably right, but if so they're being short-sighted. By the time today's high school students graduate from college, coding won't be done...at least in the sense it is today. Actually, I stopped recommending that people train to become programmers about a decade ago. The writing was on the wall. We'll still need programmers for a few more years, but in decreasing numbers.

      When I started programming, every company with access to a computer had a staff of programmers. These days, nowhere near as

      • I'd advise younger folks to go into business, the trades, or become a doctor. As for people like me, there was never any choice. I was always going to code and making a living was just incidental. I do it because love doing it and that love is what gives me an edge over people who just go to school and learn it because someone told them it was a good idea. I have only been more sought after and better paid since AI entered the scene.
      • 25% is probably the tip of the iceberg. Real smart developers will find ways to use AI to become 5x or 10x more productive. But as they are smart they will not show that, they will just keep developing at the same productivity rate and use the extra spare time to productize their insights then after a year or so suddenly leave and start their own company.

    • "We need an excess supply of students who want to become developers for us, so that we can hire them for low pay, and have them compete against each other in an AI-dominated workplace to increase the reliability of the tools that will eventually make redundant what few openings for humans remain."

      All the C-level execs are salivating at the prospect of making even more profit by cutting as many developers as they can. If nobody wants to become a developer because of this, if nobody wants to learn how to code anymore because AI has taken their jobs, they have only themselves to blame. Supply and demand work both ways but greedy capitalist fuckwads always plead and whine when the same system that made them billionaires end up working against their interests.

      The AI prophets are pushing HARD to make sure all current gen developers are using AI tools. Management believes the hype, that we have to use AI all the time or we're falling behind. Right now, using AI puts us behind. But, it's all good, because we're training our replacements. One day, or so the AI prophets have promised, we'll be able to use the AI software development cycle sans-humans, and all those wasted man-hours will have paid off for the corporations! IMAGINE NO MORE HUMANS IN THE SOFTWARE DEVELO

    • Maybe so, but if you're any good at what you do, you have very little to worry about. The average student in a music or art class never really learns to sing or paint, but those classes are still valuable in providing exposure. The average student that takes a coding class will not ever be good at coding, but the exposure might still do them some good.

  • Just get one AI to review the other AI's code.
    And fire everybody.
    Will happen, quality and value to customers are standing in the way of profits.
    • by allo ( 1728082 )

      It's called agents. One agent writes, one critiques, one corrects. They may be different LLM or the same, but they do different roles.
      The success of this varies, but it is better than having just one role.

  • growing role? (Score:4, Insightful)

    by dfghjk ( 711126 ) on Monday November 11, 2024 @02:24PM (#64937347)

    "...despite AI's growing role in software development."

    Who says there's a growing role? AI companies?

    • Pichai was telling us last week that about 25% of the code produced now in Google, is generated AI. Says article posted here on /.
      Soon, 50%, then 100%.
      • I suspect he's utterly full of shit, too. He's probably counting CSS as "code" and fabricating the rest. Google has been around for over 20 years. You believe they went back and re-coded everything so they could pack in 30% "AI code" just so he could say that? No, he's a CEO and he just made that shit up on his own. I call bullshit. Yes, we all use AI sometimes to help spot a problem or review a snippet. I also use it sometimes to help build a working model or document existing code. However, it's nowhere n
        • just for the record, I don't believe very much at all... and definitely not from any management people.
          The rush to be sitting with the cool kids in AI will surely end in tears.. but still seems to earn salespeople their quarterly bonus at this time.
           
  • by MpVpRb ( 1423381 ) on Monday November 11, 2024 @03:05PM (#64937501)

    ...talent is real and it takes a special kind of mind to be good at programming.
    While I agree that learning the basics of software is a useful way to train the mind, many untalented students will be very disappointed when they try to find work.
    Part of being a successful programmer is constantly learning. Knowledge gets stale in this field very fast.
    In order to succeed in the future world of software, it will be necessary to learn the new tools as they are developed.
    And no, AI will not magically convert vague statements in natural language into excellent code. Natural language is not even close to being able to express the complexity in software. This is why programming languages were invented. For a good example of using natural language to express complex concepts, look at the law. For thousands of years, smart people have tried to precisely write down laws and other smart people have succeeded in finding loopholes and unexpected edge cases.

  • Too many students required to learn coding let ChatGPT and the like do the thinking for them. That way they never get the exercise needed. And as soon as they enter a range of tasks were ChatGPT does not cut it anymore or requires close supervision, these students are completely lost. Not good. Obviously, there are still some that get it and do just fine. But many are struggling more now than before. Yes, I have seen this first-hand in a class I teach.

  • I mean although this has been promised for decades now, it's actually still possible that you will want to use a computer. If you cannot program your computer, you are severely limited with what you can do with a computer. This may not have been a problem in the past, but today manufacturers of computer system often have to shame in exploiting their users for their own capital gains.

    If you can't program your computer you are then doomed to work for the company by wasting your time doom-scrolling in the hope

  • BIG difference. Anyone can learn to read code, but actual system design and implementation is true coding.

    It's like learning to read vs learning to write an essay with a coherent collection of thoughts.

    Or is 'learning to code' just the new data entry clerk?
  • to learn how to think and reason.

    • That's all fine and good but only actually coding is going to significantly help you get better. Sure there are foundational skills, but they aren't going to take you to the level you need to be to be a professional in most cases. My advice is to practice, code, and grind. Don't waste time trying to have AI do everything for you and just use it for the areas it's good for. Stay coding. Stay on a project or three all the time. Don't listen to assholes telling you this or that will make you a better coder or
  • There is nothing of value that Google has launched in the past decade. Just money grabs, ways to make the web more annoying, and clones of stuff others were doing.

    Would we even notice or care if Google stopped delivering anything?

    • I don't know. They might not be coming out with amazing new stuff, but that's not the same as doing nothing.

      GMail's spam filter is still an order of magnitude better than any other.
      Google Maps is still significantly superior to any other navigation software.
      Android continues to be refined and improved.

      These things don't happen when a company just sits on their hands.

      • by Njovich ( 553857 )

        Killing the competition using buyouts and unfair market practices is hardly innovation. Like who were even the contenders here?

        Gmail: Proton. See this: https://proton.me/blog/search-... [proton.me]

        Google Maps: Waze which they bought and then left to bleed out. Also what is improved in the past 10 years there? Seems to me just more advertisements on it now.

        Android: Sorry what exactly is improved in Android? Like I legit couldn't name one thing that's improved in android in the past 10 years other than that it's harder n

        • IMO Proton Mail sells a false sense of security. There is no such thing as "secure" email, unless you count those janky web-based things doctor's offices use, and even those are dubious. Because Proton interacts with "ordinary" email accounts, all bets are off.

          Google Navigation has continuously improved accuracy, adding things like navigation through large shopping center parking lots, speed limits, and crowd-sourced business hours. Not revolutionary, but these refinements make life better for users.

          Android

          • by Njovich ( 553857 )

            IMO Proton Mail sells a false sense of security. There is no such thing as "secure" email

            Ah the old 'there is no perfect security so no security layer ever makes sense'. The 1980s called and they want it back. Anyway, the whole point was that Google used illegal methods to depress their competition, and there are many other examples as well. Google is the gateway to the web and anything that they don't like they might suppress. And absolutely, in many ways you can do security things with email that Google does not. And it's fine that they don't. But when a competitor rises and they tried to kil

  • AI coding is a like a zombie movie, gone wrong.
  • "Important as ever" is a good way to put it. I.e. not really all that important for the masses.
  • company that benefits from cheaper labor advocates for more people to be trained to do said job.... while constantly pushing for software to replace said workers.

"Security is mostly a superstition. It does not exist in nature... Life is either a daring adventure or nothing." -- Helen Keller

Working...