Execs admit AI makes them value human workers less Executives have leaned in to AI, only to stumble before reaching any return on their investment. "Most AI spending has under-delivered, leaving execs feeling like they’re burning cash," says employment biz G-P (Globalization Partners) in its third annual AI at Work Report. The report finds corporate leaders' enthusiasm for AI …
> 88 percent expressed concern that employees are using AI performatively rather than adding business value.
Gee, I wonder why that would be. It can’t be because managers, lacking common sense or any meaningful connection to the work, are telling people they better use it or else get laid off, miss bonuses, etc. One of my friends says he has his work’s AI write sonnets because one of their key KPIs is tokens used.
Vogon by any chance?
Ode to a Lump of Green Putty I Found in My Armpit One Midsummer Morning...
In the sultry swamp of sweaty pits,
A putrid glob did congeal,
A muculent mound of malodorous green,
Its sickly scent did it unseal.
The sun's blazing wrath did scorch the Earth,
Yet in my fetid hollow it hid,
This loathsome lump of emerald woe,
Like a festering secret amid.
Oh, revolting mass of viscous slime,
Your oozing form doth repulse,
A horrid blob, of rancid hue,
You make my stomach convulse.
With tender touch I did you pluck,
From your noisome lair of despair,
And marveled at your repugnant grace,
My hand smothered in your snare.
Foul whispers you shared, of nauseous fate,
Your pestilent charm I could not resist,
Your squelching embrace, an odious dance,
The world eclipsed in your miasmic mist.
In you I saw the end of all,
A putrescent doom impending,
But in my heart, a macabre delight,
For in my armpit you'd been attending.
So here I stand, reciting this ode,
To a small lump of green putty I found,
In the depths of my musky crevasse,
Midsummer morning, profound.
One of my friends says he has his work’s AI write sonnets because one of their key KPIs is tokens used.
I've put requests to Copilot "create me a picture of a family of feathered dinosaurs in a leafy forest; the feathers have fractal patterns on them, and the tree leaves are tiny and intricately veined; strong dappled sunlight illuminates the scene; the dinosaurs are bipedal carnivores". When it comes back I get the limited resolution image, so the follow up is "enlarge the lower quarter of that image to fill the full frame", which of course can go on for some time.
Not sure how my organisation accounts for token usage, but all I'm doing is toeing the line from our organisation's leader who tells us that AI is going to be transformative.
Actually, as large language models they are excellent tools and impressive for language manipulation purposes.
For my crimes, part of my job is currently "corporate writing", in the shape of my employer's annual report. That means compressing roughly a thousand man-years of complex and often technical effort into about 15,000 words of everyday language; factual accuracy is paramount. I have written contributions in all document formats (including meeting transcripts) from almost 100 people across the organisation, so my starting inputs are over 50,000 words, some verbose, some mere bullets. Knowing that "AI" isn't intelligent and can't be trusted, it is still an exceptionally useful tool for tidying up complex documents, for finding structure in random piles of human gibberish, offering shorter versions of long text, for homogenising language to a standard style. And then I have to read the entire lot and make sure that it's not AI slop. Then every single number has to be audited and signed off by a relevant qualified professional. With all that checking, does this save us time? Yes, because in previous years we still had to do the fact checking on raw meat sack output. It also means (whether people like it or not) a better quality output, with more errors caught, better proofreading (by the LLM). It isn't one prompt to produce an entire document, it's a series of selective prompts where that adds value for me and my colleagues. Some content is always human authored and human checked. Some passages don't need LLM attention; some I'll try LLM and find it adds no value; for others it is a godsend.
That won't fit most people's use cases, but as I say it's a language model, I'm using it in its sweet spot. That won't justify the trillions being pissed up the wall by desperate corporations and avaricious investors, nor the mendacious marketing by the companies who are trying to make money out of it. The biggest problem is of course gullibility and FOMO on the part of execs - whilst "only" 23% of execs have total confidence in AI, that's almost a quarter of execs putting their hand up to admit they are clueless arseholes. Based on the commentariat's wider experience does 23% seem right?
"AI skills" — Just possible that in sense that manglement thinks it means those skills cannot exist or at best is an oxymoron.
"lack of data literacy" — No idea what these geniuses think they mean by "data literacy" but analytics and data sciences have been a part of most disciplines for decades, so there isn't really a great shortage of skills. Just a shortage of those skills available for minimum wage compensation, I suspect.
My take pre·AI of most (all) enterprises is that they are extremely poorly managed with sub·normal managers and a gibbering C-suite idiotcracy all wrapped in appallingly dismal business processes and systems — Miraculously in spite of the organisational shambles these train wrecks are kept on the tracks by the intelligence and work of the lowly minions at the coalface.
Even if, as in Sam Altman's wettest dreams, AI could do all that it is claimed, I very much doubt it could ever make the silk purse from the corporate pig's ear (or is it arse?)
One thing we've started seeing where I work is the Execs being shocked and suprised that the AI vendors they are pushing us to use (who are spending $100B's/year) have started sending us big bills. This is to the point we've been told to limit usage of the best (and most expensive) models and use the cheaper (shitter) ones where possible.
Who'd have though that when a company spends Billions every year on something they want to make that back + profit from the people who are using it! Thankfully it's not like I work for a financial services company where we could expect the Execs to know how capitalism works....
Corporates (and governments) are not technocracies. You implement AI if your boss tells you to, because his boss has read about it being the next big thing.
It doesn't matter whether it works or not.
When it starts to fade from the media or is replaced by something else, you will be able to quietly dump it and start fixing everything.
As it is, AI will destroy customer service for anything complex or unusual. So value resilience and have a 'Plan B', because when something you rely on breaks, as when the Romans left, you may not be able to get it fixed.
