Will Chatbots Eat India's IT Industry? (economist.com) 61
Economist: What is the ideal job to outsource to AI? Today's AIs, in particular the Chatgpt-like generative sort, have a leaky memory, cannot handle physical objects and are worse than humans at interacting with humans. Where they excel is in manipulating numbers and symbols, especially within well-defined tasks such as writing bits of computer code. This happens to be the forte of giant existing outsourcing businesses -- India's information-technology companies. Seven of them, including the two biggest, Tata Consultancy Services (TCS) and Infosys, collectively laid off 75,000 employees last year. The firms say this reduction, equivalent to about 4% of their combined workforce, has nothing to do with ai and reflects the broader slowdown in the tech sector. In reality, they say, ai is an opportunity, not a threat.
Business services are critical to India's economy. The sector employs 5m people, or less than 1% of Indian workers, but contributes 7% of GDP and nearly a quarter of total exports. Simple services such as call centres account for a fifth of those foreign revenues. Three-fifths are generated by it services such as moving data to the computing cloud. The rest comes from sophisticated processes tailored for individual clients. Capital Economics, a research firm, calculates that an extreme case, in which ai wiped out the industry entirely and the resources were not reallocated, would knock nearly one percentage point off annual GDP growth over the next decade in India. In a likelier scenario of "a slow demise," the country would grow 0.3-0.4 percentage points less fast. The simplest jobs are the most vulnerable. Data from Upwork, a freelancing platform, shows that earnings for uncomplicated writing tasks like copy-editing fell by 5% between Chatgpt's launch in November 2022 and April 2023, relative to roles less affected by ai. In the year after Dall-e 2, an image-creation model, was launched in April 2022, wages for jobs like graphic design fell by 7-14%. Some companies are using AI to deal with simple customer-service requests and repetitive data-processing tasks. In April K. Krithivasan, chief executive of TCS, predicted that "maybe a year or so down the line" chatbots could do much of the work of a call-centre employee. In time, he mused, AI could foretell gripes and alleviate them before a customer ever picks up the phone.
Business services are critical to India's economy. The sector employs 5m people, or less than 1% of Indian workers, but contributes 7% of GDP and nearly a quarter of total exports. Simple services such as call centres account for a fifth of those foreign revenues. Three-fifths are generated by it services such as moving data to the computing cloud. The rest comes from sophisticated processes tailored for individual clients. Capital Economics, a research firm, calculates that an extreme case, in which ai wiped out the industry entirely and the resources were not reallocated, would knock nearly one percentage point off annual GDP growth over the next decade in India. In a likelier scenario of "a slow demise," the country would grow 0.3-0.4 percentage points less fast. The simplest jobs are the most vulnerable. Data from Upwork, a freelancing platform, shows that earnings for uncomplicated writing tasks like copy-editing fell by 5% between Chatgpt's launch in November 2022 and April 2023, relative to roles less affected by ai. In the year after Dall-e 2, an image-creation model, was launched in April 2022, wages for jobs like graphic design fell by 7-14%. Some companies are using AI to deal with simple customer-service requests and repetitive data-processing tasks. In April K. Krithivasan, chief executive of TCS, predicted that "maybe a year or so down the line" chatbots could do much of the work of a call-centre employee. In time, he mused, AI could foretell gripes and alleviate them before a customer ever picks up the phone.
Tata, WiPro, Cognizant, IBM bubbye, *WAVES* (Score:5, Insightful)
Alternatively, I'd also convert all H1B visas to Greencards and give the US tech industry a kick in the balls. They hate when they don't have an underclass they can pick on and threaten.
Re:Tata, WiPro, Cognizant, IBM bubbye, *WAVES* (Score:5, Insightful)
I dunno... Tier 1 phone techs are already pretty much worthless for anything more advanced than issuing an RMA/refund or escalating to a tier 2 tech. Tier 2 techs have, for at least a good half-decade, been in a race to the bottom and are now the ones who read you through the canned scripts that used to be the province of tier 1. You have to get escalated up to tier 3 these days to find anyone with half a brain and willing enough to use it to be useful.
As annoying as I find the voice chat bots we already have; they do indeed fulfill every purpose and aspect of the tier 1s and the work they are willing to do now. Today's tier 2... barely above the former tier 1 level of service... is so close to being equally basic and non-thinking that I bet AI could do >90% of the job right now. The only somewhat advanced thing that either would really need to do is effectively determine when to escalate.
Re: (Score:2)
Re: (Score:3)
Some can. Others are just flipping pages in a flip book asking questions and (badly) fitting your free-form response into the multiple choice answers to figure out which page to flip to next. They have no actual authority to make any decision and often don't even have a way to contact tier 2 other than tick a box on a support form and hope (really, not care if) someone in tier 2 notices. Some of them aren't even aware what company they're flipping for.
Tier 1 has been thoroughly enshitified.
I'm pretty sure A
Re: (Score:3)
The point of tier 1 support is to drive most of the callers away in disgust, and then filter out the remaining callers into those with trivial solutions (restart the device) leaving only a tiny number of callers who need real support from someone who understands the problem.
Re: (Score:3)
I dunno... Tier 1 phone techs are already pretty much worthless for anything more advanced than issuing an RMA/refund or escalating to a tier 2 tech. Tier 2 techs have, for at least a good half-decade, been in a race to the bottom and are now the ones who read you through the canned scripts that used to be the province of tier 1. You have to get escalated up to tier 3 these days to find anyone with half a brain and willing enough to use it to be useful.
As annoying as I find the voice chat bots we already have; they do indeed fulfill every purpose and aspect of the tier 1s and the work they are willing to do now. Today's tier 2... barely above the former tier 1 level of service... is so close to being equally basic and non-thinking that I bet AI could do >90% of the job right now. The only somewhat advanced thing that either would really need to do is effectively determine when to escalate.
They are annoying for a reason.
This is to deter anyone from calling by making every call a miserable experience.
Every RMA/refund costs money. If that process was painless, it would financially devastating to those companies.
Also, maybe you're not old enough but before outsourcing to India, the way to be annoying was to have insane hold times - like 45 minutes to 2 hours. Any sort of phone support from any big organization took the whole morning or afternoon.
Re: (Score:2)
These days in modern times, the hold times are still 45 minutes to 2 hours. Though sometimes you get through to tier 1 fast but then hold an hour for someone competent. Or more commonly you get a LOT of 5 to 10 minute holds, there's an answer and you're put on another hold after being transferred (sometimes in a circle), with a total of 45 minutes to two hours before you actually get to someone who can help.
Nothing has actually improved over time, though many things have gotten worse.
Re:Tata, WiPro, Cognizant, IBM bubbye, *WAVES* (Score:4, Funny)
At least the bots can speak English to there you can actually understand them....
Re: Tata, WiPro, Cognizant, IBM bubbye, *WAVES* (Score:1)
Re: (Score:3)
Beware when David Attenborough calls you about the extended car warranty.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
But a single low quality phone support person, at the moment, does not seem replaceable by an AI without less than one low quality support staff. AI is not free, not even low quality AI designed just to make the caller hang up. AI is also slow. Are they going to create massive server farms with thousands of employees just to shut down a mid-size call center?
I get the feeling that either the bean counters are screwing up the math badly, or else the intent of these announcements praising AI is an attempt t
Re: (Score:2)
Looks like it's time to reallocate some stock. The fat-cats don't care if they replace us with brown people or robots, so long as they can. It's a force of greed that worked for India before, but there is no loyalty there. The only thing is that I haven't seen good evidence that the bots are quite good enough to replace call center personnel, but the state-of-the-art does appear close enough to be believable. Take away call centers and remote support folks at the entry-tier and the rest of the reasons to "move to India" start to disintegrate also. I watched the offshoring wave obliterate the US tech companies that did well in the 1990's. I survived but many others got washed out to other industries or retired early and I'm not going to shed a bunch of tears if Indian-offshoring ends up back at square one.
Alternatively, I'd also convert all H1B visas to Greencards and give the US tech industry a kick in the balls. They hate when they don't have an underclass they can pick on and threaten.
A friend of mine got outsourced/replaced with someone from Eastern Europe.
He's brown in the USA and got replaced by a white person.
Sounds like from your perspective, my friend's incident evens it all out.
And, don't worry about my friend. He found a job in 1.5 months for higher position and pay and most importantly, 1/4 the commute.
Re: (Score:2)
Re: (Score:1)
IMHO, the H-1B program should never exist. If someone is that valuable, they should not be chained to one company. They should be treated as a truly valuable asset, assuming they are actually that, and not hired because they are cheaper. Instead of a special visa, they should get a permanent resident visa with a short path to citizenship, should they choose to go that route.
Offshoring needs to be looked at too. I have been told by people that if Bangalore knows about something, Moscow, Beijing, Pyongyan
Re: (Score:2)
It is indeed a bad situation, but I think one that companies like. An H1-B employee is essentially trapped; meaning they will work longer hours than the law allows for fear of being deported if they say no. Until that greencard comes through, they also don't dare leave the country, not even on a business trip, just in case they're not allowed back in to see their family. This turns around and hurts other workers, as the bosses will point the the pseudo-indentured workers and demand "why aren't you workin
Will chatbots be smart enough? (Score:2)
Generally, if I call a support line, it is for something that I cannot solve through a voice response tree or though the website. It requires information or understanding that I have not seen from current technologies. Mebbe chatbots will eventually get sufficiently Artificially Intelligent to understand and respond to exceptional cases. But until that miracle occurs, there better be a quick way to get to a Human Being who is (presumably) naturally somewhat intelligent and can respond to questions that h
Re:Will chatbots be smart enough? (Score:5, Insightful)
Re: (Score:2)
LLMs are way more manipulable, so i'd go in that direction even now.
Re: (Score:2)
Ture. At one place where I do academic teaching part of the IT has stopped to respond to support requests completely, unless you escalate rather aggressively. If IT organizations are crumbling (and many seem to be), no or fake support makes not a lot of difference.
Re: (Score:2)
It is intentionally undermining itself also. They demand to only need to know skills to pass a Microsoft certification, then are baffled that they get fired and replaced by someone cheaper with the same certificate? They demand sameness and monotony, only Microsoft applications allowed, and becoming clones of each other. Maybe it's not the bottom tier IT people demanding this, but as you go higher up the food chain it is the attitude that I've been seeing grow for the last several decades.
Re: (Score:2)
Well, Microsoft has accumulated an incredible mountain of technological debt and cannot keep its cloud or its OS secure. Eventually this is all going to come crashing down and everybody will need to move to better tech or die. Many will die.
You are correct though, competently done IT looks different. Competently done IT has second-sources for any critical product and a replacement strategy for any outsourced service. Of course, competently done IT needs competent IT people and IT leadership. Those seem to b
Re: (Score:2)
At least I can understand AI, tier-1 India support, not so much.
Disclosure: I worked for a company that had off-shore resources, culturally and linguistically moving your tier-1 and even tier-2 support off-shore does nothing but piss your customers off. It's not just India, it's China and the Philippines that welcome any offshoring opportunities.
Re: (Score:2)
Mebbe chatbots will eventually get sufficiently Artificially Intelligent to understand and respond to exceptional cases.
Maybe. But not with any currently known technology. LLMs are pretty much peaking at this time and they will get _worse_ (due to model collapse). Maybe somebody will eventually find something better, but whether that takes 10, 100, 1000 years or longer is completely unknown. All that is known is that currently known approaches have zero insight and can only deal with what they found frequently in their training data.
Re: (Score:2)
I played with chatbots a lot, mainly for small scripting code generation.
Background: I own both Linux and Windows machines, and I admit to never had learned scripting the proper way. My scripting skills are very basic. Now, I could focus on a certain scripting environment (say, Bash) and learn it better, but I can't learn 5+ scripting language at the same time, especially since I use each very infrequently. Main purpose for script usage is home lab task automation, for example I needed a Powershell script w
Re: (Score:2)
The worst is the way so many vendors have shifted their customer support onto the phone or chat... human OR bot... versus just letting people file a proper ticket. I don't need some human being to read me a bunch of "thank you"s and "we're sorry you experienced this" platitudes. And I certainly don't appreciate waiting on the line and waisting time explaining my issue and/or go through basic troubleshooting with every single goddamned person I get bounced to while they try to escalate me to the right tech
Re: (Score:2)
I've done that, filing reports with Apple. The result: Silence. Once or twice, my problem got resolved/the bug got fixed in a subsequent product release. In other cases, the bugs still exist, and I have no freaking clue if they've (a) read the report; (b) been able to reproduce my result; (c) figured out what the problem is; let alone (d) actually tested and scheduled a fix.
Apple is not alone in this, but it's the one that frustrates me the most. I stopped dealing with CVS' on-line systems a couple yea
I'm hopeful that future AI (Score:3)
...will provide the best customer service ever.
It has the potential to read every knowledge base, trouble ticket and engineering document for a product and give accurate answers to any technical problem
Unfortunately, today's AI will be as bad as the usual poorly trained person reading from a dumbed-down script
Re: (Score:2)
Unfortunately, AI, as known today, does not have that potential. There are many people that assume these are early systems and there is a lot of improvement. They are not. These systems, incapable as they are, are end-products of a long optimization process and there is little that can be done to make them better. In fact, they will be getting worse, because less and less non-contaminated training data will be available, due to model collapse.
Re: (Score:2)
These systems, incapable as they are, are end-products of a long optimization process and there is little that can be done to make them better.
That might not be true. There are still a lot of ancient algorithms in AI that might be improved with large datasets. In particular I think combining a knowledge base (like Cyc [wikipedia.org]) with an LLM might be particularly fruitful.
Re: (Score:2)
Re: (Score:2)
That is, in and of itself helpful, but that is not the AI applying reason to the problem. It doesn't understand any of it. It just knows that in the past, for a very similar seeming problem ABC, you said XYZ and that made it "resolved", so it says XYZ.
In the process of learning that response, it did not form a concept of X, Y, or Z or how any of them relate to A,B, or C.
Re: (Score:2)
Re: (Score:2)
> Unfortunately, AI, as known today, does not have that potential. There are many people that assume these are early systems and there is a lot of improvement.
If you are talking about ChatGPT, I agree with you. ChatGPT development is mostly about adding layers to it to filter out wrong answers.
But Google AI projects in general, dig deep into the AI processes to make fundamental changes, for example like Evoformer in AlphaFold2 was replaced with Pairformer in AlphaFold3. This is what real AI research shou
Re: (Score:2)
While I'm dreaming, it'll pick up my accent and speak to me in my own vernacular. But you're right, they'll screw it up.
Not really (Score:2)
Chatbots and robot customer service (tm) can barely figure out what I want when I explicitly tell them in plain language. Even with massive improve it's less efficient than a human. There is a push for AI in my industry and tech speciality and there are definitely use cases but people keep wanting to treat it like a silver bullet.
Maybe they can all retrain as plumbers (Score:2)
Re: (Score:2)
It's not bad advice. Less stress and you can make six figures, and robots aren't taking that job for at least 100 years.
Re: (Score:2)
"ackshually..."
Let's hope so (Score:2)
Unfortunately, Betteridge says no... (Score:2)
Re: (Score:2)
I think bodyshops like TATA will take a long time to adapt to writing code in ways that work well with LLMs as the sort of developer behaviors that foster good results is way outside the outsourcing culture of delivering the bare minimum required by the contract. But oh boy won't they try.
Disclaimer: I work for an Indian company (no, I am not Indian). Money's good, project is nice, and my team is not AI-threatened (ITOPS).
But yes, there's much turmoil around AI. And yes, they are trying very hard to use it.
AI? (Score:2)
I keep saying it, and I will continue to say it, what we have IS NOT AI. A slight advancement in expert systems. That is all. It's cool, sure. Useful, sure. But we need to STOP PRETENDING THAT IT'S THE AI of science fiction, or any futurist vision. We will be seeing a ton of "AI" projects imploding over the next couple of years because most are UNTENNABLE. See that AI generated video? Well, mostly done my hand after some "AI" randomized some of the images. See that Chatbot? Oh, it'll answer any qu
Re: (Score:2)
You're right, and you're wrong. Mostly because the definition of AI keeps changing.
If you were to show a "hey siri" demo to someone from the 1960's, they would call it AI. (Heck, Siri is better than Star Trek's Ship Computer (TOS) )
Same for a Tesla auto-summon for anyone from the 80's.
The Turing Test used to be the benchmark for crossing over into AI - and now we dismiss that accomplishment as "just a chatbot".
It's good that we're moving the line of "real AI" closer to AGI and ASI. We should be expecting m
Re: (Score:2)
Turing test was about measuring human like intelligence, what we call nowadays "stong AI" or "general AI". It was never about testing whether something is AI or not. AI as a word has been just "weak AI" for decades (For example we had a book called "Artificial Intelligence Programming" from 1987).
>> What we have will not solve any problems that expert systems already deal with
This claim is also wrong. AlphaFold2 has solved 200 million protein folding problems that expert systems or human experts team
The problem is automation. (Score:3)
We're running out of work, and there's nothing on the horizon to replace what we're automating. That was supposed to be a good thing, but we're not socially or emotionally prepared to deal with it.
Re: (Score:2)
That 70% who lost their middle class lifestyle (Score:2)
Or maybe you'll be homeless in your mid 50s eating out of cans and hoping the cops don't catch you and send you to a work camp. Life's funny like that.
IT Industry? (Score:3)
What IT Industry? Do you mean the scam caller industry. Yes, AI probably will take that over.
"worse than humans at interacting with humans" (Score:2)
The article says that AI is "worse than humans at interacting with humans." This assertion misses two important points.
First, the key point of first-level customer service is to redirect as many customers as possible away from escalation, thereby saving money. This can be done either by giving customers what they need or by frustrating the customers so much that they go way. Whether this is done by low-cost humans or low-cost computers is only important in terms of how many customers avoid escalation.
Sec
AI is redefined as... (Score:2)
AI is redefined as Artificial Indians?
Call center scripts... (Score:3)
...can be followed as easily by an AI as by a person. The crap-level call centers can easily be replaced. You may even be able to understand them.
The key question: how hard will it be to get a second-level tech on the line, if the AI can't solve your problem?
same thing for the register runners at McDonalds (Score:2)
automation is rightfully taking the minimally skilled positions away from those who have tried to make a career living off them. That's not what they were meant to be used for, oh well.