As IBM pushes for more automation, its AI simply not up to the job of replacing staff

IBM's plan to replace thousands of roles with AI presently looks more like outsourcing jobs to India, at the expense of organizational competency. That view of Big Blue was offered to The Register after our report on the IT giant's latest layoffs, which resonated so strongly with several IBM employees that they contacted The …

[Anonymous Coward]

Eat your own cooking

I was always in favour of eating your own cooking. Eg run your payroll on your companies code. If it doesn't work - you don't get paid! Can they replace the AI strategists with an AI system?

I remember someone saying they had a fully automated smart system, which didn't need any human intervention. Till there was a power cut, and someone had to turn it on!

A customer gave a talk at Share and said "we can do easy coding, we want IBM/Oracle/mIcrosoft to do the hard stuff. AI can "do" the easy stuff, but not the hard stuff such as deep operating system level.

[katrinab]

Re: Eat your own cooking

So-called "AI" can copy/paste random stuff from Stack Overflow, except a human who did that would read the original question, and the comments to figure out if the code was likely to work for their scenario, and would know to change variable names and so on to make it fit into the rest of their code.

[ArrZarr]

Re: Eat your own cooking

It's like I've been telling the management - AI is a strong tool for supporting somebody to do more of the stuff they already know how to do. If they are already a competent developer, they can work with LLM support. If they're a competent salesperson, they can use it to build a framework to create your slides from. You 100% still need the human knowledge and experience to ensure that the fallible personal assistant isn't introducing errors or issues.

Personally I use it to code review what I built (team of one, it's very good at understanding code you throw at it), but when a non technical colleague used it to modify an existing piece of code, it lead to a chain of issues that took considerably more time to straighten out than if I had done the change in the first place because they didn't have the context or knowledge.

Thinking about it as I write this comment, thinking of current LLMs as fairly competent but very junior team members that you're responsible for the outcome of their work is a fairly solid analogy. Pro: Don't need to badger HR for the extra resource. Pro: Don't need to wait very long for them to do the work you're delegating. Con: Won't improve and become more trustable over time.

[Geoff Campbell]

Re: Eat your own cooking

"For the guidance of wise people, and the blind obedience of fools."

GJC

[NoneSuch]

If There Were True Justice.

"Right, lay off a few thousand staff."

"Done, chief."

"Great, so this AI stuff can do everything those former employees can do?"

"Not really, no."

"Wait, what?"

"A couple of hundred folks you just laid off were hardware engineers. They are needed to purchase, set up and maintain the AI servers. About a hundred were AI training experts, They would have set up the models and applied Deep Q learning methodology to optimize workflows. Another fifty were high-end programmers. We needed them to bring everything together. Right now, we just have a stack of boxed servers on the loading bay."

"Well, bring them back then!"

"Sorry, can't. They've formed their own company and they'll be charging five times what it would have cost us to employ them directly."

"So this is going to cost us a half billion over budget?"

"According to ChatGPT, yes."

[Eclectic Man]

Re: Eat your own cooking

The philosopher Michael Sandel recently did a radio programme on BBC Radio 4 about whether people would accept an AI version of themselves, that is, trained on their life experiences writings etc. as their 'legacy' when they died. It was a shame that he did not go to the logical conclusion and ask himself whether he would accept an LLM trained on his books, articles, media appearances etc. as his legacy.

It would be nice to try an experiment where an AI / LLM is trained on a writer's complete oeuvre, and the writer and the LLM are independently asked to produce some written statements on the writer's area of expertise such as an article for publication, a response to a question, a letter to the press, or an extended essay, and the results compared. My expectation is that the human writer's output would be more enlightening and sensible that the AI / LLM's effort, as the human has lots of inputs, experiences and thoughts of which the computer system will have no conception. My own written output (posts here, a few letters in 'the press' and three whole academic papers) is a bit 'thin'. I have a possibly identifiable style and point of view, but do not necessarily agree with what I have written in the past - how would an AI / LLM deal with that?

Similarly for a computer programmer (I excuse myself from this as my code is for my purposes only and definitely not up to commercial standards, the test is for genuine experts). I expect that the AI / LLM would be excellent at following certain rules (such as commenting, layout etc.) but less than perfect at tricky, efficient or elegant coding. Anyone tried this themselves?

[Geoff Campbell]

Re: Eat your own cooking

There's an interesting fore-runner experiment here, in all of the famous-but-dead authors whose work has been continued by their children and/or PAs. I leave the reader to fill in the names and conclusions...

GJC

[Eclectic Man]

Re: Eat your own cooking

It seems that the London Evening Standard has tried this with deceased art critic Brian Sewell, and come a cropper (according to the Guardian):

"Who knew the late art critic Brian Sewell was such a tediously cliched writer? Especially since some of the dead verbiage in the London Standard’s AI version of Sewell reviewing Van Gogh at the National Gallery has become common currency only since his death at 84 in 2015.

Give him credit, he had a voice. And it was a posh voice. Evidently the chatbot used by the Standard needs to be fed a lot more novels by Evelyn Waugh and Anthony Powell, some Latin perhaps, and a mouthful of plums before it can begin to resemble the public school-educated, Courtauld-trained Sewell, who started his career as the protege of the upper-class art historian and Soviet spy Anthony Blunt.

I stopped watching The Crown when Sam West as Blunt, the traitorous surveyor of the queen’s pictures, talked about “early modern art”. No way would this snooty connoisseur have used that social historian’s terminology.

The Standard’s latest work is even more nonsensical. The neutral classless tones of AI don’t have the waspishness and hauteur of Sewell at all."

https://www.theguardian.com/media/2024/sep/26/london-standards-ai-imitation-of-brian-sewell-proves-art-critics-cannot-be-easily-replaced

SO, humans can still do some things better (or more annoyingly) than computers. I'll drink to that!

[Anonymous Coward]

So if we ask chatGPT....

AI is unlikely to fully replace experienced technical engineers, but it can complement their work and enhance productivity. Here's why:

1. Complex Problem Solving

Experienced engineers possess deep knowledge, creativity, and intuition, which are essential for solving complex, non-routine problems. AI excels in data processing and pattern recognition but lacks the human ability to understand broader contexts or think innovatively. For highly technical and novel engineering challenges, human expertise is indispensable.

2. Adaptability and Contextual Understanding

AI relies on training data and specific programming to function, while experienced engineers can adapt to changing requirements, ambiguous situations, and cross-disciplinary challenges. Engineers can handle unexpected issues, which AI might struggle with due to a lack of contextual flexibility.

3. Soft Skills and Leadership

Engineers often need leadership, communication, and collaboration skills to lead teams, mentor others, and manage projects. AI can't replicate these interpersonal dynamics, making engineers valuable for team-oriented tasks.

4. Ethical and Safety Judgments

In fields like aerospace, automotive, or infrastructure engineering, ethical and safety decisions are critical. Experienced engineers bring years of accumulated judgment to ensure designs and systems are safe. AI lacks the nuanced understanding of these considerations.

5. AI as a Tool, Not a Replacement

AI can automate routine tasks (e.g., simulations, data analysis, monitoring) and assist in generating designs or optimizing processes, freeing engineers to focus on more creative or strategic work. AI enhances efficiency but remains a tool that engineers use to improve outcomes.

In conclusion, AI is a powerful aid to technical engineers but lacks the creative, adaptive, and human aspects that experienced professionals provide.

So, if an AI (which isn't actually intelligent at all) can work out it can't replace experienced technical engineers, then why the hell can't C-Suite????

[2Blockchainz]

Re: So if we ask chatGPT....

Because investors have been rewarding companies who trim staff by raising their equity values. Since C-level executives only expect to stay in their positions a year to three, Wall Street and it's associated private wealth system are creating a moral hazard.

Oddly, Elon Musk, whose experiment with de-staffing Twitter has served as this model, has lost a ton of equity value on Twitter. I wish we could see approximately how much he's lost.

[EricB123]

Re: So if we ask chatGPT....

"Billions and billions" - Carl Sagan

[Triggerfish]

Re: So if we ask chatGPT....

The predictability of the average MBA, means they're probably the easiest to automate work wise TBH.

[Elongated Muskrat]

Re: So if we ask chatGPT....

He paid $44Bn for it, didn't he? Apparently it has lost 5.75% of it's "market cap" in the last 12 months, but as for its value? I think it has been estimated at far less than that.

[katrinab]

Re: So if we ask chatGPT....

I've got to the point now where I can recognise ChatGPT output a mile off, and just ignore what it is saying.

[Howard Sway]

Krishna's stated plan to replace people with AI appears not to have had the desired impact.

Unsurprising when your plan is to use something miraculous that can do everyone's jobs without bothering to find out whether it actually works well enough to do so.

IBM could be the first company where everybody loses their job to AI though - but only because the company's gone bust due to their stupid bet that it could do technical jobs better than a human.

[Martin Summers]

Re: Krishna's stated plan to replace people with AI appears not to have had the desired impact.

And let's hope they do go bust. Their plans are inhuman, at the expense of humans. Why would anyone want to work for IBM anymore? AI will never meaningfully replace human jobs. If they actually get anywhere with it at first, the plans will crash down the minute they experience problems that require real knowledge, intelligence and intuition. A CEO who thinks human beings are replaceable, is someone I'd never want to do business with let alone work for.

[Anonymous Coward]

Re: Krishna's stated plan to replace people with AI appears not to have had the desired impact.

> And let's hope they do go bust

That would be unfortunate because IBM do have some useful technologies that I'd like to see preserved, not least micro-slicing CPU time instead of the threading approach of Intel.

And if my Government client wasn't so obsessed with "commodity hardware" they might eventually realise that two or three large mainframes would allow them to do all they are failing to do with cloud, more cheaply and with far more control over their data.

[chozorho]

Re: Krishna's stated plan to replace people with AI appears not to have had the desired impact.

I respectfully disagree. I want IBM to learn from their mistakes and do better in the future. Aren't they doing research in quantum computing? I'd rather see them succeed in this area than some other company that is sympathetic to a foreign adversay.

But I may be biased, since I once dreamed of working for them.

[Irongut]

OMG their firmware is 4 years old!!! I take it Casey has not been in the industry long.

Four years is nothing. I'm sure most of us have a tale about an office or comms room with a decrepit PC running DOS, Netware or some other ancient OS that was vital to the company and could not be upgraded or replaced even after decades in service.

I remember running an MS Mail Postoffice (for the children present that's email from before Exchange, Outlook, GMail, etc) on an old IBM PS/2 in 1999. That's a 1987 computer running IBM PC DOS 3.3 on an 80286 which had the main status screen permanently burned in to its green screen monitor. If someone attached documents or presentations that were too large it would bring down the system, there was one manager who did this every quarter. Yet there was no money or impetus to upgrade despite the compnay being an IT services provider who installed NT4 and Exchange for our clients.

[Fruit and Nutcase]

AI in IBM lingo

Means "All India"

Moving jobs to India means it's being done by AI

[Groo The Wanderer]

Re: AI in IBM lingo

IBM's just playing loose with the translation of "AI". For them, it means "Anonymous Indian", not "Artificial Intelligence."

[Fruit and Nutcase]

Re: AI in IBM lingo

Touché!

Have a beer

[Martin Summers]

Re: AI in IBM lingo

And those Indian people will eventually want more money to progress their lifestyles and live better. Pretty much as we have in the West. What then, raid adoption agencies to raise worker slaves until they've had enough too?

[Like a badger]

Re: AI in IBM lingo

"And those Indian people will eventually want more money to progress their lifestyles and live better. Pretty much as we have in the West. What then, raid adoption agencies to raise worker slaves until they've had enough too?"

The plan then will be to re-source to Africa, or the poorer parts of South America.

[Dan 55]

If IBM are running a cloud service and their infrastructure is exploitable by the last four years of CVEs and they're firing a third of their network engineers they might as well be painting a dartboard on their back. ChatGPT isn't going to save them there.

[Fading]

EOL 4 years ago

Does not equal four years old.

[Elongated Muskrat]

I read it as "the firmware was no longer supported 4 years ago", not "the firmware is 4 years old". Probably more like 14 years old, not 4.

[Steve Button]

"As IBM pushes for more automation, its AI simply not up to the job of replacing staff"

... and are there any companies that are actually replacing staff with AI?

I mean, I'd like to replace my Bertie vacuum cleaner with an AI robot, but I'm not taking Bertie to the tip just yet. Just because I want something, doesn't mean that thing automatically exists.

It's a bit like BT announcing that they are going to replace 40% of staff with AI. It's simply not going to happen in the short to medium term. They might be able to insert AI into some low level functions, where it can help to diagnose network faults for example, but the people won't be replaced, they will just be supplanted. You'll still need just as many people to feed the AI in the first place, probably many more.

As an aside, I can imagine at some point in the future we certainly will have an AI which is capable of doing a better job of vacuuming (and washing up, cooking, clothes washing) than I currently manage. Obviously the words "we" and "some point" are doing a lot of heavy lifting here. "We" could mean my great-great grandchildren, and "some point" could be way after I'm pushing up daisies. But it'll happen eventually.

[yoganmahew]

Re: "As IBM pushes for more automation, its AI simply not up to the job of replacing staff"

Yeah, and the problem is, by the time the beancounters realise that the staff they 'let go' aren't replaced by productivity improvements, those staff are really gone. Many into permanent retirement, some into less hamster-wheel intense jobs. All realising there isn't enough money to want to go back to the hell-hole they were previously in (it's de facto a hell-hole if management thinks that they can replace staff with AI...).

[Snake]

Re: "As IBM pushes for more automation, its AI simply not up to the job of replacing staff"

The quote was

"The whole outsourced to AI thing is a myth that somehow our upper echelon of execs believes exists right now,”

If this is indeed true then the 'upper echelon' shouldn't *be* upper echelon - this is IBM we are talking about. One of the great Founding Fathers, if not arguably the greatest / most important one, that helped create the Tech World that we all exist in. If it weren't for the System/360, I'd argue business computing wouldn't be anything what it is today (because, after all, the PC was originally conceived as an add-on).

If the C-suite is truly this stupid then they shouldn't be in the C-suite of one of the most important survivors of the Founding Fathers that exists. If true, they just don't get it, they don't understand technology, they are just a bunch of trained MBA monkeys pressing their buttons and reading their spreadsheets but not understanding a single iota of what their business is actually capable of, what their business actually does. *At this point in its development*, thinking that "AI" will replace your coders is a VAST lesson in utter stupidity. Years from now? Maybe. Now? Not a chance, it isn't even "AI".

This should be a lesson to the stockholders to seriously consider replacing the entire lot of them, but that's reaching as stockholders never seem to question the upper management, only follow along like the little Wall Street-loyal lapdogs that they are.

Idiots. The world is populated by idiots.

[Boris the Cockroach]

The fall

of civilisation.

Where we fire the senior/experienced engineers and try to replace them with AI, the result is increased profits than benefit C-level folks and shareholders and leaving the company to wither away as the expertise drains out of it.

And why do they do it?

Well its simple really... they dont have a clue to increase profits otherwise. cutting costs is all they know.

Its rather like Asimov's Foundation series, where the empire had a problem with its atomic power plants caused by lack of skilled people to staff them. the solution... shutdown the power plants because it was quicker than training new staff

[Anonymous Coward]

Re: The fall

Yeah, as Kelsey Hightower suggested (broadly), IBM's "over-reliance on AI coding assistants might choke off that pipeline in the future, removing the next generation of engineers". And, as Blake says here (essentially), it's made worse by the sorry state of "Watsonx Code Assistant".

Looks to me like a lose-lose situation for Big Blue.

[Fruit and Nutcase]

Re: The fall

The C-level folks with all their AI gadgets, first to get self-drive cars etc (if they are not killed by aforementioned self-drive cars running into the back of emergeny vehicles) will also be the first to experience cognitive decline of subjugating function to AI. We don't even need to get to AI - but use it or lose it - the more we get AI to do things without an equal diversion of that mental capacity, we are going to lose that brain power. Not a problem I have to contend with, as I'll be dust long before this

[Anonymous Coward]

Sooooo IBM

This vapourware is sooooo IBM. Nothing new there.

[cjcox]

IBM can teach a "master class" on this.

IBM is a true master of knowing how to get people to "quit" instead of the high cost and high visibility of "layoffs".

They know how to push your "I hate working here" buttons when desired. Saves them a ton of money and paperwork.

[OldGeezer]

Re: IBM can teach a "master class" on this.

Having had the misfortune to have worked for them in the 90s, IBM is the one company I will never ever work for again - even if hell freezes over!

[theblackhand]

To whom it my concern

Dear IBM

You are correct that AI can replace part of your workforce. The mistake you are making is the section of the workforce you are targeting.

The employees that you are currently looking to replace with AI actually provide your customers with useful work.

Those that are suggesting replacing workers with AI would be perfect candidates for being replaced by AI as they are higher up the foodchain, so the savings will be greater, and they have likely built a career on being incompetent, aka the IBM way.

Lots of love

A customer and former employee

[Tron]

A raft of tech companies may go this way.

AI is not reliable and may never be, so forcing your own company to rely on it is like cutting down a tree from above.

I'm surprised that skilled staff still care enough about a company that sees them as disposable, to make a fuss. When the momentum gathers, I think most would be looking to bag whatever redundancy payout they can and escape to somewhere that has a decent future. No matter how big a company is, if it is badly run for long enough, it will fail. Call it a tipping point or an event horizon, but employees really need to bail before the wheels start to come off. Given the shortage of tech talent, there must be sensible people looking to poach quality talent.

[Alan Brown]

"thanks to an internal ban on using externally sourced LLMs"

This isn't just an AI problem in IBM, it's across the board and results in IBM software often being bloody awful to use

I get _why_ they have such a ban but devs working in isolation don't get to compare the enduser experience (wether customers od coders) and realise how bad "their way" may be

[abend0c4]

Re: "thanks to an internal ban on using externally sourced LLMs"

And it goes back a long way. Back in the 80s, I couldn't believe how bad IBM's software development tools were (even the "online" options were just resource-hungry wrappers around batch compilation and the buffered screens made editing unnecessarily tedious) compared with not just their competition but with industry norms. Their networking was belated and messy and their midrange systems clung to the cardpunch/lineprinter model long after minicomputers had obsoleted it.

What IBM had in its favour was well-engineered and solidly reliable hardware - and ruthless hardware sales teams. The software was always considered an unfortunate necessity to support the hardware sales. These days, IBM seems to believe everything it does is an unfortunate necessity to support its income and appears to be basing its business model on entitlement rather than engineering. Not that it's the only culprit.

[Anonymous Coward]

Sounds like it's the C-suite that should be replaced by AI.

[Anonymous Coward]

Beat me to it

Spewing stupid shit and making stupid decisions is something AI can do for free, who needs Arvind and his cronies to run the company into the ground?

[Groo The Wanderer]

Iche Bin Moron seems to be the new translation for "IBM "

[Anonymous Coward]

Trying to be the future it wants to sell to other companies... I get the strategy, but does anyone in an honest job (as opposed to CTO BS) really believe any of it?

And if they did, where are the unions and/or government railing for universal basic incomes in lieu of people having jobs?

[Cem Ayin]

Not an accident...

...waiting to happen. Rather an unmitigated disaster unfolding - first gradually, then suddenly.

"Senior software engineers stopped being developed in the US around 2012," Blake said. "That’s the real story. No country on Earth is producing new coders faster than old ones retire. India and Brazil were the last countries and both stopped growing new devs circa 2023. China stopped growing new devs in 2020."

and

"If it weren't for LLMs, there would be a serious lack of programmers in the next five years as Gen Xers started retiring," said Blake. "I had planned on coding 'til the day I died, but now I think I’ll be talking to LLMs primarily instead."

Assuming these statements to be factually correct, this is just isane; the relatively best outcome would be for this scenario to fail hard before unfixable damage is done. In fact, the worst case scenario would be their A"I" folks somehow managing to actually make this work, more or less, initially. Meaning that before long, they would find themselves with millions and millions of LoC that no one and nothing in the world understands and is able to maintain (including the LLMs that generated them). And it's not just IBM, of course; this would then be the code that the very foundation of our civilsation critically depends on. I hope I won't live to see the day when that bill comes due.

[Doctor Huh?]

"Oddly, Elon Musk, whose experiment with de-staffing Twitter has served as this model, has lost a ton of equity value on Twitter. I wish we could see approximately how much he's lost."

His mind, as demonstrated by available evidence. The rest is just money.

[herman]

Indian Business Machines

So Indian Business Machines is experimenting with Artificial Indians. No wonder it is not working out very well.

[ethindp]

> "Senior software engineers stopped being developed in the US around 2012," Blake said. "That’s the real story. No country on Earth is producing new coders faster than old ones retire. India and Brazil were the last countries and both stopped growing new devs circa 2023. China stopped growing new devs in 2020."

I don't think this is true. Not really, anyway. I think that the surveys show that companies just *don't* want to hire junior developers who could become senior devs if trained. I'm 25 and have been actively job hunting for 4 years now and still haven't been hired -- and that's after going through lots of resume revisions. But of course, Companies love their automation, and they keep making it harder and harder and harder to get past their stupid applicant tracking systems, so of course they "stopped producing senior software engineers". Maybe if they stopped using ATSs (or reduced how strict they are) they might get more talent, but noooooooo, that's too much to ask for. Oh no, devs in India and other countries where labor laws are practically non-existent are somehow better than people right here in the US. The irony is, of course, that if you ask around, companies will swear up and down that there isn't anyone to hire here, yet there are at least 6 million people who want to be hired generally. Yet they can easily find someone to hire in India. Talk about absurd...

/rant

[Anonymous Coward]

AI HR

IF unit_age>35 AND unit_pay >40000 AND unit_start_date <20240101 THEN resource_action(unit)

[TimMaher]

Alex, Blake & Casey

Surely that should have been “Harvey, Arthur & Larry”?

“I’m sorry Dave. I can’t do that”.

[JK63]

This strategy of dumping experienced staff is not unique to IBM in the tech industry.

[Bebu]

Albert's Bridge

I am reminded of Tom Stoppard's (radio) play Albert's Bridge where the management of the Clufton Bridge implemented a "reduction in force" of the four painters who could paint the bridge in two years, to a single painter (Albert) on the basis that the new silvery paint lasted eight years while the existing brown paint lasted only two years. The four painters had just finished and were ready to start again. ;)

[Aa95]

IBM, doomed like K-mart

Remember a few years ago, the hot trend was cloud computing? Everyone wanted to emulate AWS. IBM bought Softlayer, and attempted to build PaaS on top of it. The consulting company I worked for at the time was an IBM business partner, so we attempted to use it and offer it to clients. It was so bad though that nothing worked- IBM clients didn't even want us bringing it up as an option. IBM couldn't even get the billing for it to work!

Now, the hype is AI. IBM doesn't talk about cloud anymore, it's got a new buzzword to chase. And they're lagging just as far behind as they did with cloud.

The problem though is that IBM is decimating their cash cows while still striking out on innovation. For example, they are moving zOS development to India, which means they've put it in maintenance mode. Innovation there is dead.

The very last K-mart closed this week in the US. It was a pioneering low cost retailer, but first bad management and then an owner who bought it just to suck every last cent out of it doomed the company. IBM is going to meet the same fate as K-mart.