Follow Slashdot stories on Twitter

 



Forgot your password?
typodupeerror
×
AI Technology

ChatGPT, Other AI Models To Disrupt Indian IT Firms (reuters.com) 47

Generative AI models such as ChatGPT will slow down market share gains and deflate pricing for Indian IT companies in the short term, analysts at J.P.Morgan said on Friday. From a report: As generative AI is implemented more broadly, consulting firms like Accenture and Deloitte and will gain market share over Indian IT firms like Infosys and Wipro in the near term, analysts at the brokerage said in a note to clients. Generative AI can be a "deflation driver" in the near term on legacy services as they compete on pricing, necessitate staff retraining and drive loss of competitiveness, they added.

"ChatGPT is likely to deflate legacy services the most and application services the least." Artificial intelligence company OpenAI's chatbot has dazzled amateurs and industry experts with its ability to spit out haikus, debug code and answer questions while imitating human speech, helping it attract a $10 billion investment from Microsoft earlier this month.

This discussion has been archived. No new comments can be posted.

ChatGPT, Other AI Models To Disrupt Indian IT Firms

Comments Filter:
  • Will India destabilize?
    • How would we know the difference? 2/3 of India is a big sprawling shanty town. I suppose this can turn it into 9/10 shanty town.

      Don't get me wrong, you have to start from somewhere; the US started as gold-hunter campsites and a place for Europe to dump their criminals and annoying zealots. (It hasn't actually changed much, it's just that these actions are formalized into corporate structures.)

      Anyhow, I believe bot-generated code has to be curated and tested by humans, so most the humans are not going away.

      • I suspect instead of needing 5 employees, these tools will allow companies to get away with 3 employees due to productivity gains.

    • memories.
      i remember when i became affected by indias gold rush of engineering jobs in america.
      indias response was for affected americans to just retrain for some other job.
      interesting times indeed.
      software engineering is now functionally obsolete.
      i hear that taco bell has openings for cooks and cleaners
      and taco bell offers to train you

      do not let the door hit you when you are let go

  • by gweihir ( 88907 ) on Friday February 10, 2023 @03:46PM (#63282885)

    Well, I have been saying that for decades, at least for most offshoring companies. So they probably now found that ChatGPT can sufficiently emulate the cluelessness and insanity of these "services" as to be not worse. As the ones that buy these services have a clue pretty much on the same level (or below) of ChatGPT, I recommend to next replace all outsourcing/offshoring "managers" with ChatGPT as well. It cannot make things worse! On the plus side, longer-term some crappy companies will go hopefully go bankrupt this way and stop bothering us.

    • I would be more kind with it... they found that most of what they outsourced was really low-value, low-skill work that could easily be automated, and would have years ago if it wasn't cheaper to have someone in India doing it either manually or with minimal automation.

      Today I did something in Grafana in 15 minutes that used to take me a month to do in Excel (back in 2002). Then, with little additional effort I could add visualizations to the data that essentially did the job it took a day for a VP to do in

      • by gweihir ( 88907 )

        Well, yes, that effect too. Low-skill IT work is going the way of the dodo and that is definitely a good thing.

    • ChatPHB can probably replace the MBA's, I don't find them very bright, and BS-ing has well-established patterns. Unlike technology, the shady side of human nature doesn't change very fast, making BS an ideal candidate for pattern-mining AI bots. Maybe one PHB can do the job of several using ChatPHB to help. They'd become a BS curator.

      (PHB is a Dilbert comic strip reference to Dilbert's clueless buzzword-spewing boss.)

      • by gweihir ( 88907 )

        Having just have a run-in with a no-insight, no-skill bean counter, I completely agree. Fortunately these people can be outmaneuvered if you are in a high-skill position but they are still a real nuisance and constantly decrease product quality and long-term outcome and force you to invest time into finding ways to actually get your job done. People with a massive negative overall contribution to society.

    • The improvement in AI is on a geometric curve. Maybe now it's doing generic filler, but it won't be long before it can do YOUR job. You're not as special as you think.
      • by gweihir ( 88907 )

        Hahahaha, no. It is actually more like a heavy-tail and it can basically already do everything it will ever amount to. Training it more is subject to diminishing returns and keeping it from going off into the deep end is getting harder and harder.

        And no, no Artificial Idiocy of the general types we have today will never be able to do any of my jobs. I do have an unfair advantage though: I know what these morons could do 30 years back and I do some really highly skilled stuff.

        • This is what it can do with the current settings and the current amount of compute. It can make original poems, write snippets of code to suit plain English descriptions, write fluff essays, accurately summarize papers without losing the important bits, propose obvious new ideas, generate legal opinions with supporting links, write book reports, write short stories, and write letters. Do you really think this is qualitatively different from doing all of the above but doing it well?

          Don't say no one warned y
          • by gweihir ( 88907 )

            Nope. It can do none of these things. All it can do is convincingly fake these skills with somewhat high accuracy, but which still makes for pretty bad reliability. You still need to have to an actual expert check every of the rather simple results it can produce, because some are just bad and many are subtly off. You also need that expert to integrate the solution fragments it can deliver. Besides, these machines are not able to do any of the real work I do and what I do comes with an assurance that I actu

            • "The question of whether a computer can think is no more interesting than the question of whether a submarine can swim."
              -Dijkstra

              It can do all of those things. The fact is you, and I mean YOU, personally, are unable to tell the difference between a poem a human wrote and one this machine wrote. Here's a haiku, written by either myself or ChatGPT. Can you guess which?

              Grass grows in autumn
              Fragility will prevail
              To live is to die

              You might guess right, having pretty much a 50/50 chance. But if I wrote it, it'
        • There’s all sorts of expert system work that hasn’t made it into the current AI/ML frenzy and I don’t think it will eliminate engineering positions any time soon but I do think in the near future we’ll see a lot of expert systems in place that basically allow a single engineer to do a whole teams worth of work and faster without the teamwork drag.

          • by gweihir ( 88907 )

            Maybe. Of course being able to do it with a smaller team is always a huge win much better than just linear in the reduction of the headcount. And any larger team has some people that do simple stuff and many people that are just average in skill, insight and experience. Hence enabling single (exceptional) engineers to do whole projects alone again (remember this is the classic mode: single engineers or very small teams) would be a massive win. The main question will be whether it is possible to add a workin

            • My experience with copilot is that if you keep your functions small and write good comments it will write anything remotely common though you might need to switch some things around. I’m seeing papers about other lifecycle phases being done with AI/ML and I’m doing some 90s era reading right now and with the constraints they were up against that there is a lot not yet bothered with as far as building out expert systems for things. The hard stuff is going to take a lot of human toil for awhile

      • by Njovich ( 553857 )

        The improvement in AI is on a geometric curve

        A flat line is a geometric curve too. For most people making this type of statements their n=1 (GPT3), so given one datapoint you can make any type of graph through it.

  • by rsilvergun ( 571051 ) on Friday February 10, 2023 @03:47PM (#63282891)
    to say a lot of people are going to lose their jobs to automation. It's phrased this way so as to not cause panic. Panic that is probably deserved.

    It's one thing to outsource jobs. That's not what's happening here. These are jobs being destroyed. And ask any American who lost their job to outsourcing how "retraining" worked out for them. Ask Ohio and Michigan how that turned out? There's a reason we call it the "Rust Belt" and not the "gleaming center of trade" belt.
    • ask any American who lost their job to outsourcing how "retraining" worked out for them.

      These programs take the collective work of humanity and privatize the gain from them. It’s easy to hide behind the fact there is no higher reasoning and you still need a human, but even the simple ones we have now are already vastly boosting workplace productivity. Workers have seen pay stay flat with inflation since the 70s while productivity gains mean average salaries should be around double. Instead, all of these gains are funneled to a few people. At some point services like these that take f

      • In the last 40 years the top 1% pocketed 50 trillion dollars. Meanwhile our national debt hit 31 trillion.

        If somebody spent the last 40 years running up your credit card bill would you pay it?

        So far for the majority of Americans it seems like we're going to. Meanwhile those 1 percenters sitting on all our money are laughing at a suckers
    • by Tablizer ( 95088 )

      "Bots are taking all your jerbs!"

    • by dddux ( 3656447 )

      On a positive side, they will also stop procreating so much. No more baby slaves.

  • ChatGPT gonna take over the Indian scam call centres' jobs then?

    • Let's hope it can be used against them. Get the AI to keep them on line all day long wasting time.

    • Re:How's that? (Score:5, Interesting)

      by sound+vision ( 884283 ) on Friday February 10, 2023 @04:19PM (#63282981) Journal

      Scams are going to be one of the most lucrative applications of this technology. Not so much the call center scams, but those done over a textual medium. Most of the rapport-building process - which is most of the entire process - can be done automated.

      Social media, text messaging, and instant messaging are about to become a whole lot less viable, until these services add some kind of ID verification requirement. They're bad enough as it is, but once scammer labor gets replaced with scambots, the floodgates will really be open.

      • You don't need a requirement you just need to learn how to evaluate a source. Twitter might be a bit of a problem because they seem to want to sell verification is a service for a monthly fee. But that just means you stop using Twitter as a source of information. At least unless it's an account from a person you absolutely know in real life and you're able to make sure you're actually viewing that account and not an imposter.

        It's really not that big of a change given that most of the massive amounts of
  • On the other hand, in the age of ChatGPT offshoring firms might substantially improve the average quality of code produced WINK WINK.

    • by maitas ( 98290 )

      Yep. I think that things will go the other way around. Accenture and Deloitte won't be able to justify a premium over Indian consulting since everyone will be using ChatGPT like tools to code.

  • Self-driving cars initially showed lots of promise and potential, but making them practical under real-world conditions is proving to be a huge sticking point. The first profitable batch will probably be taxis and shuttles that run on pre-vetted routes.

    I suspect these bots replacing white collar jobs will be similar. It might happen, but the road to practicality will be much longer than expected, gradually easing into narrow niches and slowly expanding out from there. Thus, save a bit more retirement than

  • They have less insight into AI than slashdotters.
    This is about making certain stock prices move in ways that are
    most profitable to JP Morgan.

  • Comment removed based on user account deletion
  • "Show bobs and vagene." Okay, maybe that's more Replika than ChatGPT.

  • You cannot talk to a person in India from the US with paying quite a high toll rate. There are no landlines that connect there, it's an incredible satellite bounce to the other hemisphere...

    What we have here is the "Indian IT Guy Accent" that developed in the USA... as call centers disliked talking to stupid users at cut-rate prices in the late 90s/early 2000. Now, companies brag about "All US Tech Support" that speaks fluent Midwestern English.

    So, this is a Speech-To-Text then text-search then playback scr

    • So, this is a Speech-To-Text then text-search then playback script system. Not AI.

      I suspect the majority of them already are like this.

A Fortran compiler is the hobgoblin of little minis.

Working...