How deliciously binary: AI has yet to pay off – or is transforming business The tech industry's enthusiasm for artificial intelligence software – a conveniently amorphous term – has yet to generate much of an economic windfall. An estimated $1 trillion of capital expenditure commitments in the years ahead to develop and deploy AI "has little to show for it so far," according to a June report [PDF] …
Well, the internet did massively change the world and upend many industries. It just took many years until people found out what worked and what didn't. Most dotcom companies went bust, but a few that didn't are now worth trillions. If you want to win the lottery, you have to participate...
The thing is, I do not want to win the lottery. Winning the lottery is [mostly] a curse. Reasonably high stable income is much better.
And I do not want any tech bro to ‘win the lottery’ either. Why should I want more monopolistic megacorporations controlling everything? We are still trying to deal with the current bunch…
"Well, the internet did massively change the world"
The internet (packet switching and routing and that) has changed the world - that's engineering. LLMs and the like are very fancy and are cool enough to allow the carnies to skim their marks with unprecedented efficiency. I'm actually a Brit and I'm sure we have some home grown terminology but I'm going all in ... left pond!
LLMs are being used to deploy a massive amount of "blogs" which are basically crap to fill search results and sell ads. The input data for LLMs is rapidly becoming its output too and vv.
It won't end well. AI will remain an unattainable dream (let alone I)
I am hard pressed to find solid evidence the Internet actually makes me money or has improved my life. Or for millions of average people for that matter.
I have no doubt the big circle jerk of come-ons, touts and hands out is working for somebody. Seems like mostly the grifters, though.
Personally the internet provides me with a job, but mostly it provides me with endless entertainment. I haven't felt bored or needed to watch daytime TV since 1999.
So, panem et circenses.
But seriously, just the existence of Wikipedia and YouTube are a considerable improvement to my life. Also, accessing government services on the internet is a considerable improvement over queueing.
Amazon and Apple are the only tech companies from that era that are worth more now than before the dot.com crash.
Apple was a recovery stock. It was considered at the time to be in the same category as the likes of Commodore Amiga. It pulled through, all the others from the 1980s home computer boom either didn't, or are companies like Microsoft that are doing fine but not good enough to justify their valuations at that time.
Google IPO'd in 2004. The VCs who invested in it at the time certainly made a lot of money. Google wasn't really on anyone's radar back then.
Ebay, technically the price is up, but it works out at an average increase of about 1% per year, so you would have made more money in a bank deposit account.
Expedia was owned by Microsoft at the time and was spun off later.
Priceline went down a lot between just before the crash and when it was acquired by Booking.com
So a company doesn't count if it wasn't public before the crash? Seems arbitrary, but OK.
Pick any day in 1999 or 2000 and Ebay was somewhere between $5 and $10. Today it's about $55. What new math are you using to come up with 1% per year?
Priceline was not acquired by booking.com, it's the other way around. Also, the company is worth a lot more now than it was during the dotcom boom. Whether it fluctuated in between has no relevance to the discussion.
You seem to be making this shit up as you go along...
“So a company doesn't count if it wasn't public before the crash? Seems arbitrary, but OK”
Yes, because if you were invested in dot.com stocks, you wouldn’t have had Google in your portfolio.
And this is really important. You may have identified that Search was a good way to make money, but none of the big players at the time: Altavista, Lycos, Metacrawler, HotBot, Ask Jeeves, Yahoo!, etc were successful. Some new company came along and fixed all the problems they had and took over the market. I think the same could happen in AI.
In some other sectors, companies were successful but didn’t justify their valuations.
Also, how would you have picked out Amazon or Apple from all the alternatives available at the time? Both made their money by moving into a different sub-sector from where they were at the time.
AI is older than the current hype phase, and has already earned billion$ for plenty of very wealthy companies.
For example, and let’s suppose you were to play any kind of race driving game. Most of those games, by-and-large, are played as single-player experiences, and yet are contests among other “participants” (who are actually computer-controlled rivals following some kind of pre-programmed algorithm).
Even though the strategies are rudimentary at best, yeah, the ghosts in pac-man are an early form of artificial intelligence— that paid off hugely for the inventors!
Video gaming is one example of extremely lucrative returns on artificial intelligence (it’s just that nobody calls it that, even though that’s exactly what it is).
Nothing like taking down some Spaceward Ho! algorithms, particularly because they are so diabolically savage, it’s like ohmigod stop wasting my settlements ya cold calculating bastard, here’s my invading fleet on your homeworld you scummy cylon slime-suckin griblets. Eat laser blast!
Well clearly big business is dumping a trillion dollars into it so they can helpfully get rid of those useless employees. Also, hopefully they can replace their customers with AI, since these days customers are more of a liability than a source of revenue. Without the burden of employees or customers their numbers can only go up, so shareholders win for at least one quarter, which is what's really important.
Someone much cleverer than me said:
"The future is already here, it's just not evenly distributed"
My sister in law is a chiropodist. She keeps paper receipts and her accounts are handwritten.
It works for her.
I have used the MS llm at work and it's... ok but just a novelty really.
Other divisions of $job are starting to use llms on live help desk webchats.
The sad thing is that I can't imagine it will be much worse than the present situation. It would be much cheaper and much less frustrating if companies would simply acknowledge that they had absolute contempt for their customers and had no intention of listening to their whining.
Indeed. Taking the crappest of crap companies (say Virgin Media) and a customer service model built on LLM could be light years better than their existing crapshored customer service, which appears to take vagrants off third world streets, crowd them into a skip* and give them a telephone and a script, but no training. Potential advantages of AI for Virgin Media customers: No waiting times; No loss of skilled human agents (as there are currently none); No change to customer expectations (the agents simply lie and make things up, AI can't be any worse); More consistency in the lies and un-help.
* Off topic, but "dumpster" does seem a much better term than "skip". There's very few areas of grammar the USAians have got right, but this is one. It really is about time Blightly started referring to these things as dumpsters, it's just so nicely descriptive.
Yes... calling your ex a cumdumpster sounds much better than a cumskip.
But calling my ex that would be very unfair... even though she cheated on me twice... I would never call her that.
She does actually care for the environment a great deal... so she's more of a cumpost bin.
Ah, Virgin. I recall a while back when I was lucky enough to be one of their customers. When they put the cable in, they bodged it so it popped out of the edge of the pavement, over the edging, before diving down under my front lawn. Then someone/something pulled a loop out making a serious trip hazard. Having elderly neighbours who could be injured, I called Virgin and explained the situation. A week or two later I get a call asking me if I was going to be in on a date for the engineer. I pointed out that the problem was outside and that I didn't need to be in.
Anyway, the whole saga went on for over a year including:
Six visits from people who "didn't have the equipment" to cut the pavement edging.
Cards through my door saying they had called by I was out.
Four or more calls from me pointing out the problem and that they could be in for a nasty lawsuit
“…just 5.4 percent of US businesses reported using AI”
I’m not sure I believe even that low figure. When you say “using”, in what capacity? Enabling that annoying feature in Bing? Messing about making some pretty pictures with one of the picture generators? Hardly a world-conquering business initiative is it?
Yes, it IS all hype. And I’m sorry that El Reg seems to be one of the very worse offenders - every other story these days included those two little letters - it’s getting quite boring.
I’m getting really tired of reading about “AI”. It’s NOT AI. It’s just some (admittedly clever) search algorithm on a boat-load of data that has been stolen from pretty much everyone with any presence in the internet.
I’ll be very happy when the bubble finally bursts and we can all go back to work. Until the next super-hyped-up-bollox comes along in 6 months or a year.
There is undoubtedly a lot of hype surrounding AI. I grow tired of it.
A major part of all this froth is that VCs are desparately seeking the next great thing (and sole control of it). They have their own investors, who also want a piece of the pie. Doing nothing - i.e. applying the common sense filters that most good engineers are born with - is not doing things with their investors' money that the investors are expecting them to do. They'd literally prefer their VC to take big risks than no risks at all. The bigger the potential return, the more risk appetite there is.
It's a complex, self-fulfilling thing. It's created a hive of start-up orientated techies whose only real skill is Powerpoint Engineering with an ability to put some sort of real world hardware together at a crude demo level, but no more. The more dramatic and capable the supposed future capability, the more risk appetite there is. And, even if it doesn't come off (as is almost always the case), there's no particular consequence; the pipe dream buyer has happily spent their money on the pipe dream, and are content that they at least tried to make a $trillion out of their investment.
In short, a large amount of technical endeavour these days is all about selling far fetched pipe dreams to people with a lot of money who want only to buy pipedreams, who would also be astonished if any of them turned into actual marketable things. And whilst there is a flow of pipe dreams from suppliier to consumer, there is room for various complicated submarkets to exploit that flow (e.g. NVidia shipping a shitton of GPUs that "are good for AI").
Objectively speaking, this is an appaling waste of time, money and resources, and a whole lot of real world soluble problems are not being addressed.
Controversial Question
Is humanity as a whole letting far too many people do far to much pointless science / technology related things?
The point behind the question is that, perhaps, educating people to get to the point where they can talk the hind legs off a VC donkey and persuade them to empty their wallet too on topics such as self driving cars, and then letting them do just that is a complete waste of a valuable education. It's happening only because there is a free market in tech investment; people with money are at liberty to invest in crazy ideas.
So, should the market be less free? Should there be more intervention by (I guess) governments to get more real world problems resourced, and stop the money being frittered away on doing nothing substantively useful? Should a government be able to say, "well if you're willing to risk those $billions in a stupid dumb-ass idea, we'll at least tax that so that it can be spent on something that is far less of a dumb-ass idea".
Looked at from a taxation point of view, if VC investors are preferring to waste their money on stupidity RATHER THAN PAYING IT AS TAX TO THEIR GOVERNMENT, it makes VC's look like the worst scum on the planet. They'd rather the money disappeared instead of being given to the government where there's at least a theoretical chance of it being put to use that would actually benefit everyone INCLUDING THE VC INVESTOR THEMSELVES (you know, things like health care, education, infrastructure).
How many VC investors are quite happy to throw billions down the tech plug hole, but still complain about potholes on their local roads?
Discuss!
Whilst I'm firmly of the view that AI is mostly another fad to waste exec's time, there's a slight problem with "Is humanity as a whole letting far too many people do far to much pointless science / technology related things?"
It is wasteful...but if you're going to stop the waste, then you either stop innovation that could lead somewhere, or you're suggesting that somebody somewhere can pick winners. Look at things that when they were being developed had no obvious near term application - the LCD, the personal computer are two that come to mind.
The problem here is venture capitalists. Back when equity markets were first established, that's how somebody raised money - with all the checks and balances of being a listed company, with the investors taking the fall for every investment that was pants. People got their fingers burned, they knew it and either learned from the experience or didn't. Now, anything with a bit of risk is deemed unsuitable for public markets, and along come the VCs, always playing with other people's money, yet no accountability, no transparency, no common sense.
If Jurassic park taught me anything.,... it's that just because you can do something, doesn't mean you should... They can (and are) enshitifying the entire internet, and it will just lead to AI reading AI and regurgitating AI because no one can be bothered to read what no one could be bothered to write.
My mobile phone has gone from a shinning beacon of light, allow me to communicate with the world from almost anywhere... to 'who the fuck is calling me' & 'I'm not answering any calls unless I ahve a scheduled call arranged because it's always some spammer, scammer and never anyone interesting'
>or you're suggesting that somebody somewhere can pick winners.
Less so that, and more somebody somewhere spotting obviously unnecessary and / or obviously not going to be achieved losers. There's innovation, and then there's pointless wastes of time.
Self driving cars are a particularly good example. We don't really need self driving cars. For occassions when we don't want to drive, there's taxis. The "work whilst your travel" argument has been lost, because if you can work in the car on the way to the office one can also never leave home in the first place and acocmplish the same thing. Road safety is more readily transformed by installation of round abouts at road junctions, something the USA is now (finally!) embracing.
If all the money that'd been sunk into self-driving car projects had instead been put into fixing roads and installing roundabouts, there'd be fewer accidents and better quality roads for everyone to benefit from.
This is the problem when you assume that what you don't need is also unnecessary for everyone else.
"Self driving cars are a particularly good example. We don't really need self driving cars."
Let's see your arguments:
"For occassions when we don't want to drive, there's taxis."
Unless you're traveling in a way that taxis don't want to. For instance, a long distance that isn't already covered by public transportation. Taxis work in cities, not so well in suburbs, and may not be available at all in rural places. From your comment, I'm guessing you've got access to taxis at convenient prices and wait times, but not everyone does.
"The "work whilst your travel" argument has been lost, because if you can work in the car on the way to the office one can also never leave home in the first place and acocmplish the same thing."
Unless, that is, you might have to travel somewhere and still want to work. You don't have to work while commuting to an office, but what if you have to drive out to a variety of locations during the day and could work while doing that? Or what if you want to travel to somewhere else and to do something, work or leisure, while you're doing that? I'm guessing from your comment that you mostly drive to commute or used to do so, but that doesn't apply to everyone.
The many other purposes of automatic driving weren't considered. There are many companies that want to have those for automated cargo delivery. The human drivers don't like this idea, but the people paying for deliveries would like it if it can be done more cheaply. People who cannot drive may like the idea because it gives them independence in transportation that they now lack. People who like to go on road trips may appreciate doing something else while still using their preferred mode of transportation. Families with only one car may appreciate the ability to send the vehicle unattended to pick someone up rather than having to do that work manually. From your comment, I'm guessing that none of these situations apply to you.
Be careful when you say that something is a pointless waste of time that nobody needs. Sometimes, people have different circumstances to you which is why someone is spending money trying to solve their problem. If all the things I don't like and would never buy did not exist, your life would likely be less pleasant than it is, because it is quite likely that some of those things are things you use frequently.
"Our customers can say, 'I want to see gloves. Show me only gardening gloves made out of such and such material that are waterproof.' And [the AI software] will immediately filter thousands of records of gloves down to the handful that meet that criteria."
Plus marketing will insist on adding stuff that's "like" gloves but not really,plus stuff that's "trending" plus stuff that other people bought after buying gloves, plus stuff other people bought with gloves.
Maybe an AI driven adblock is in order
Admittedly, those filter boxes rarely work unless this is one manufacturer's set of products, and sometimes not even then. Many sites either lack the filters that we care about or someone fills them in with garbage. I'm thinking of spec lists, lists which go into the filters, which look like this:
Size: 120*58*20mm
Internal memory: included
Memory card: supported
Waterproof: Water resistant design (translation: no)
Material: metal+plastic
However, I don't expect that their bot will be much better at sorting through descriptions or reviews to figure out the real answer. True, if it were a human reading them, it might be more obvious which parts are correct. For instance, for the device above, the disclaimer in the description that indicates that it's not waterproof could be used to cancel out the meaningless stuff in the spec list and they probably have an actual number for internal memory and a specific type of storage card. However, when faced with conflicting information, I am not confident that the bot will figure this out.
The only bonus of adding AI is that it will also hallucinate product data, so the filter boxes won't work at all.
I already see this with Amazon product searches, where half the results explicitly ignore the search terms. I said "15mm", why are half the results 10mm?
"All that investment just to replicate the filter tick boxes on any e-commerce website?"
Well, the filter options on most e-commerce websites are a bit shit once you move away from basic stuff like waist size or colour. Very rare to find a supplier who properly lists every possible aspect you might want to know accurately. Even companies who should know better like Screwfix aren't very good on product dimensions.
My problem with current search algorithms is that they really take "like" to the extreme. Because you like A, you will like B. Nope. That may be true for someone else, not me. It used to be the case that Amazon etc. allowed the use of NOT "B" to remove those results from the list. That feature has been removed. Why? I explicitly don't want to see them, but now must page through endless crap to try and find what I want.
"Our customers can say, 'I want to see gloves. Show me only gardening gloves made out of such and such material that are waterproof.' And [the AI software] will immediately filter thousands of records of gloves down to the handful that meet that criteria."
The customers here, I think, are the retailers engaging Clearview to produce this sort of market/product research.
This particular query is pretty simple SQL/RDBMS query if the tables covering gloves had columns for material and permeability to fluids eg waterproof.
I would guess these tables are very likely incomplete or don't exist. Clearview probably manually extracted this data from product specifications. Turning a LLM loose on these product pages and consumer reports could extract implicit properties from the data. For example gloves made from Iatex, PVC or nitrile would highly likely be water resistant.
I wasn't clear whether Clearview was selling "AI" ie trained models or normal databases populated by "AI."
>Plus marketing will insist on adding stuff that's "like" gloves but not really
This. Imagine that, after investing enormous effort in getting it to understand product descriptions and customer queries, AI product search actually works.
For some time, we get good search.
At that point, enormous efforts will then be invested into figuring out how to game it so that it returns your product over a competitor's, even if yours is a poor match.
Alternatively, or in addition, the original provider will provide the ability to purchase "sponsored" status for a product, and make the AI return that preferentially, even if it's a poor match.
At which point, we're back to shit search, only with megawatts of power spinning in circles before returning results.
Before AI, Amazon went through this, Google went through this. Adding AI won't fix the underlying problem, because it's not a technical problem; it's an advertisement economy problem.
I don't see any way to prevent this. Personally, I actively avoid "sponsored" products or links whenever I can, but I doubt individual awareness will ever have a meaningful impact. Enjoy the good search period, while it lasts.
To be fair, there are lots of things where calculating the ROI is impractical and sometimes dangerous. For example, calculate the ROI on a new backup system. They can be very expensive, especially if you have a good one, meaning a system that hooks into everything that needs to be backed up, has lots of data capacity, has many tiers of storage accessibility and security, and has been developed thoroughly by the staff so that there's no running around when a backup needs to be restored because they've already developed the restore processes and tested them. Now the finance department wants to see the report on how much we've got for all that time and money.
The answer: either something really bad happened in the meantime and you have a good number to tell them after analyzing all their reports on the event to get the numbers, or it's really simple because the return thus far has been zero, if not negative counting the running costs. However, not having it is still worse than having it, and excessive zeal to calculate a return can cause more problems than it solves.
That doesn't mean that AI is something to spend any money on, but just because it's difficult or impractical to calculate the ROI on something isn't enough to disqualify it. Whether it makes any sense as a business model might be a better reason.
Well, backup ROI is just like insurance ROI. It's about controlling risk.
There are plenty of methods of working it out - median and maximum cost of untoward events vs probability of events.
I insure my house because I cannot afford to rebuild it.
The company backs up its data because losing it would be costly and in some cases an existential risk. The probability of data loss is so close to 100% as makes no difference, and thus a backup ROI is the value of the data so backed up.
'Show me only gardening gloves made out of such and such material that are waterproof.' And [the AI software] will immediately filter thousands of records of gloves down to the handful that meet that criteria. So [not only is it] making [our process] faster, but it's actually making it much more simple for the end user."'
SELECT *
FROM PRODUCT_TABLE
WHERE PRODUCT_CATEGORY = 'Gardening Gloves' AND MATERIAL = 'Nylon' AND WATERPROOF = 'TRUE';
or:
db.product_table.find({"Product_Category":"Gardening Gloves","Material":"Nylon","Waterproof":True})
Or am I missing something?
The thing you're missing is that natural language processing is a hard thing to do and to derive that query from customer input, and some types of AI are good in interpreting human inputs and transforming it to a usable form for querying a back-end (whether a database or some other form of storage). In addition, you're assuming product data is uniform and standardized, which it is usually not (and even if there are standards, suppliers are very good in putting the data in the wrong fields, etc), again, something that AI can help with.
You're missing the part where you have really bad data. Consider this: you're not running a shop, but a marketplace. You have thousands of vendors. Each vendor is supposed to correctly fill each column of your database. But they won't. They'll leave fields blank. They'll make misspelling mistakes. They'll make gross mistakes. They'll outright lie. It's not like you can check every single item to see if it's really nylon.
Amazon's checkboxes seem to be in exactly this situation. Some time ago, I wanted to buy a warm LED E2 light above 15 watts. There are checkboxes for all of that; no natural language anywhere, extremely easy to turn into a query. What do I get in the results? Dozens and dozens of 10 watts lights (including sponsored!), some that are cold, some that are not E2, and some that are not LEDs. Possibly worse, you are likely to miss the product that would actually match your requirements, because it too is mislabeled. I had to remove the "warm" tick to actually find a warm light. Most products categories on Amazon are like that.
On the other hand, vendors generally give correct information on the brochure, because bad information there is what causes returns (if not lawsuits). But that's natural language or, worse, pictures.
AI is this decades pyramid scheme... those who got in early will get rich, everyone else will lose out.
Every decade there's one or two new ways to enshitify and scam people out of money... in the 200's when 'everything' went 'digital'... which meant reproducing the same cheap shit, but with a tiny LCD on it... and charging 3x the price for a few pennies upgrade.
In the 2010's... crypto and the cloud... a pyramid scheme and a the beginnings of the enshitification of everything.
2020's... More crypto bollocks, NFTs and a now AI.
Unless you got in early, you'll lose out.. I got lucky with some bitcoin I had from around 2013 that I sold as prices rose... a few at 5k a few more around 10 and the last at 15k. It cleared most of my mortgage and all my debt. Allowed me to finally work fewer hours and now that I've been mortgage free a couple of years... live a more peaceful, less stressful life.
But a part of me knows... if I'd held my nerve 2 more yrs until 2019 when they hit 50k... I could have retired at 40
Block-chain and proof of work algorithms have many applications, but crypto was/is mostly a pyramid scheme run by hucksters. I suspect that LLMs will end up being about the same. There will be many applications where it is applicable. Right now, this seems to be different variations on the "spicey search" application. If the tech bros want to convince me that AI is really the future they will need to show me more applications.
Had to search "what does Apple AI actually DO" only to be reassured that after years of emails my brain does a more useful summary at a single glance than my Mac or Iphone will.
But then, being human, I just can't wait to ***k with the system : less common words, unusual phraseology, deliberately acting tangentially etc
I mean, one of the first things we learn is how to ...urrrmm... mislead the boss. Why should an AI algorithm be any different?
>Is anyone getting a return on their AI investment?
Well, Microsoft is selling a lot of Copilot access, that's probably the winner right now.
And NVidia selling blue jeans to gold miners.
And the electric company.
OpenAI is losing money but Microsoft is backing them to the tune of $100b, and maybe it's mostly funny money, but it should last another few years anyway.
They're hoping for a 10x payoff or better, stealing business from Google. Google of course is trying to prevent it. We shall see.
Otherwise, not so much.
Remember 1985? The big AI breakthrough was American Express watching your charges to see if they broke pattern indicating possible fraud. Everyone does that now, but do we even consider that AI any more?
And there was something about fuzzy logic and washing machines, I forget the details.
And Roomba?
Also 1985 people were asking, "hey, all these PCs, are they costing us or saving us?" Whatever the answer we're still using them, LOL.
The distinguishing mark of Google was that it delivered in a simple way, what the customers wanted.
The example of the AI scraping customer comments to deliver something they want is going to be skewed given that these are self selecting and also by fake responses. It seems like another investment in time and learning by employees in a system that will not deliver.
Given that many current IT systems, particularly the large one don't deliver even the kown goals eg Birminham Council IT extravaganza, ans the many many expensive failures to build integrated health care informtation systems, this is looking like "good" money (ie from the poorer masses) going after bad and ending up in off shore accounts. Money for nothing and ....
The companies reporting that they are using AI are just estimates - unreliable at best.
To say that you ouoght not to measure ROI with the massive investment figures being quoted is reprehensible. Spend money but we're not going to be able to know if it will be worth it. - That is no way to run a business - or a government!
What a world!
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
