Payoff from AI projects is 'dismal', biz leaders complain Businesses have become more cautious about investing in artificial intelligence tools due to concerns about cost, data security, and safety, according to a study conducted by Lucidworks, a provider of e-commerce search and customer service applications. "The honeymoon phase of generative AI is over," the company said in its …
I had to download the report to get this little tidbit, since the author of the article did not judge it sufficiently important to include in his report.
I'm sorry, but when I'm spoon-fed a bunch of statistics and percentages and absolutely no basis in numbers whatsoever, I get cranky.
So, for those of us who like to know what a survey is based on, I got the figures : over 2500 participants.
Now you know.
The remarkable thing is that they all agree the payoff is dismal, and then a worrying number state they'll be throwing the same or even more money at "AI" next year, apparently driven by nothing more than executive FOMO.
I suppose AI is simply a bigger fad wave then blockchain, crypto, IoT, smart cities/grids/charging etc, where billions have been similarly been thrown away chasing illusory benefits touted by management consultancies who know how gullible their customers are. Doesn't look to me like smart glasses, 3D vision, mixed reality or "metaverses" ever managed to or will launch themselves into the executive fad-o-matic, so I wonder what the GPU makers think will save their bacon when the AI wave breaks?
> I'm glad I'm not the only one who spotted that, when did a "presentation" become a "slidedeck"?! What a load of horse-doings!
Dunno but probably quite a lot longer than you think. The consultancy I joined in 2000 were making a conscious effort to *stop* calling them slidedecks even then (for reasons I forget - probably because the more cowboy consultancies were over-using the term and they didn't want to be tainted)
"I'm glad I'm not the only one who spotted that, when did a "presentation" become a "slidedeck"?! What a load of horse-doings!"
Maybe when presenting turned into someone reading a set of Microsoft PowerPoint slides to an audience (otherwise known as a slidedeck). This particular jargon doesn't bother me. It's like complaining about when did someone starting saying "I took the train" instead of "I travelled" (when people starting travelling by train).
Don't give a toss what they're called just the bollocks contained within them. I remember sitting in a room of 15 or so management consultancies at a large London borough and all they could talk about was how they didn't "just" want to do a powerpoint presentation. I had no idea what they were talking about...if you need to present an idea using pictures and words what else are you going to use? An overhead projector? Obsession with the means of communication as opposed to the content is the issue.
In the early days of Office, PowerPoint saved my company a contract.
I was in a quandary about how to re-explain a concept involving the switch from a DMS-style set-based database to a relational model for something the set-based model did extremely well. (Clever use of manual sets, if you care). I'd covered the subject already but the customer was understandably skittish and my words were inadequate to the task.
On my plane journey to the client I wondered if this "PowerPoint" thing, which I'd never so much as seen before, could help.
Yes it could.
I quickly put together a set of pictures to help me convey what my explanations had so far been unable to do. Same talk - with pictures.
Sat with small group of client's staff, in a cube, using my lappy and a (wired) mouse as a clicker.
Intelligent and probing questions were asked about the approach being described as each slide was shown and explained, and I was able to answer them. A marked contrast to the silence that greeted my previous attempt.
Job done, on a plane with a Windows 95 laptop with a battery life measured in hurryhurryhurry units.
So I quite like PowerPoint and it's free office clones. Don't use it much, but it does what it says on the side of the tin, which is unusual for anything computer-y.
To be fair, there are many sorts of presentations which do not involve slides, and a good presentation with slides involves far more than the slides.
And since I often here presentations referred to as "a PowerPoint", frankly, I'll take "slidedeck". Though IME the cool kids just say "deck" now.
PowerPoint presentations in specific and graphic presentation tools in general, have referred to "slidedecks" for at least 25 years as apps/files replaced actual 35mm film "slides". The term "slide" was created around 1935 (probably by Kodak around the time Kodachrome was introduced) for the ability to slide the cardboard or glass-mounted film in and out of projectors. FYI - like most things connected with technology, there are also lots of other anachronistic and idiosyncratic words for various presentation media that have changed through the years as technology changed. One example - overhead projectors. The content was originally created with thin metal sheets - so early users referred to the sheets as "foils". When they were replaced with acetate sheets, some companies (notably IBM) referred to them as "slinkys" or in groups, slinky decks.
I actually have a slide deck...well, not so much a "deck" as a "carousel"...one of those for the Kodak slide projector. Grabbed it when it was being tossed at my first company. I haven't looked all the way through it, I wanted the carousel more than the slides, but now, the slides would be a great trip down memory lane (it's from the early 90s -- some kind of customer presentation).
"the most successful generative AI initiatives [...] include projects for generating FAQs and providing HR support."
FAQs are supposed to be questions raised by real people and HR is supposed to interact with real people at an individual level (yes, I know I'm being optimistic), so handing either of these over to a dumb machine seems to be the worst possible choice.
Customer service staff are supposed to interact with real - and paying - people, but companies were happy to replace them with useless chatbots. If success is measured by reducing the number of interactions, driving people away in frustration is a positive result. However a certain amount of 'investment' is required so you don't appear too cynical in your reduction of service - but probably ot nearly as much as AI demands.
Then again, the half of the local ISP duopoly I'm currently stuck with has "support" so shoddy that a ChatGPT instance would almost certainly be an improvement. Even their own customer fora have numerous posts saying that the only way to avoid the blind following of a script is to talk to the cancellations department. I try to avoid dealing with them whenever possible, but sometimes calls are necessary. With gems like these, how could a bot be worse?
1) Agent refusing to proceed until I have replaced the ethernet cable from the tower to the router when pings to the router are fine, and traceroutes proceed to the first hop outside the LAN and then die.
2) Agent telling me to go into a Windows-specific panel, and telling me to get a Windows machine when I informed them that I run Linux and have no Windoze boxen.
3) Agent wanting to factory reset the router for an email (not mine, I use a real provider, but unfortunately, do not live alone) configuration/authentication issue.
I'm showing my age, but I got spoilt dealing with real ISPs and dealing with people who would actually listen to the described problem, along with the results of any diagnostics I had already run and steps already taken (although verifying them when necessary, can't really trust someone calling in).
There's a range of decent smaller ISPs in addition to A&A, but the scenario described by the OP sounds exactly like the average encounter with Virgin Media. Whilst unstated, there's an implication that Openreach are only offering FTTC, and if the OP wants faster speeds then VM have a monopoly until either Openreach offer FTTP, or an altnet comes along.
If 30-80 Mbps isn't acceptable to the OP, then the excellence of A&A or other small customer focused providers isn't that much help. I stayed with VM for precisely this reason for 20+ years, but as soon as FTTP was on offer I switched provider (including splitting my landline to a separate VOIP contract rather than bundled with broadband).
For the OP's benefit, in addition to A&A (also known as AAIPS), other decent smaller providers are companies such as Aquiss, uno, IDnet (no relation to Curry's IDmobile), and a few others, all offer much better customer service, and tend to be better at getting the best out of Openreach than the big ISPs if there are problems. Often they'll provide 12 month minimum terms, and no in-contract price rises which can be attractive compared to the heavily discounted deals from the majors that often then have two year terms and big in-contract price rises. However, the smaller players will be more expensive than the customer acquisition prices for new customers from EE, VM etc.
For people who don't want to pay a few quid extra for better service, and who enjoy switching provider every 18 months to two years (or haggling with customer retention staff) then the big companies can offer the cheapest option. If you want a provider that will listen and sort problems well, won't hike your price in-contract, or at the end of the fixed term, and you are prepared to pay a few quid extra then the smaller players are where you should look.
Sadly, I'm stuck left-pond. I was with Speakeasy while I was still on dialup, and before they got bought out, then a local ISP for broadband until I had to move somewhere where I'm currently stuck with AT&T. Hopefully, this summer I'll be able to move to Comcrap, which is, sadly, and improvement, at least in terms of support. No fibre options where I am, which I find rather amusing, since the affluent place with the University has maybe 1/8 of its area with fibre available and the "poor" town next door has it everywhere for reasonable prices comparatively.
When I was in Edinburgh (sadly, couldn't get the right visa to stay), I was with Demon, but IIRC they either shut down or got bought out and enshittified.
Even worse. They're all trying to move people to wireless and neglecting existing networks in addition to failing to build new ones. 5G may be great for a phone or tablet on the go (or tethering when necessary) but should never be used for something at a fixed location if there are any other real options. I don't even like having my desktop machine connected to the LAN through WiFi, but the location of the router (and not owning the place so I can't just run a cable and drop it down through the ceiling[1]) for wired. Wireless anything is such a major PITA to troubleshoot, and that's not even getting into the security issues.
So yes, I have the "joy" of no higher-speed options in terms of wired service, on a network that they're trying to dump anyway, and the only alternative, Comcrap charges extortionate overage rates on anything over 1 TB monthly.
[1] Not the prettiest option, but would be the least effort and distance by a long shot
I too was with Demon back in the day and they were good to deal with. Then changed to Blueyonder who put in cable so I had several times the ADSL speed, before it was ultimately borged to become VM. To be fair to VM in terms of infrastructure the cable in my area was fast and reliable, but in latter years the service plummeted and costs rose. Quell surprise!
I used to run an ISP, back in the days when 80% of the customers were technical, and the 20% who weren't were happy to profess their ignorance and use support time as a learning experience.
It was utter bliss, and I like to think we gave good service as a result.
GJC
Well said. Whether it's telcos or banks or insurers, call centers have been staffed by human CSRs with room temperature IQ for 10+ years. Even seven years ago, I felt that the then dumb chatbots were better than humans for 70% of customer service issues. https://gtm360.com/blog/2017/05/26/can-chatbots-replace-humans/. If anything, that percentage can only go up after ChatGPT / GenAI have entered the scene.
And TBH, a sophisticated enough bot *could* handle most requests. A sensible business could spend what they're currently spending on dumb minions on a smaller group of competent techs (and paying them well enough) to deal with what a bot couldn't handle. And skilled people aren't a money-sink, they're valuable to a business and, absent a monopoly, as is the case with much of the US ISP "market", essential in getting people to keep forking over cash. I'm still buying (or paying for services) from those companies that give solid support.
The main reason I still deal with Verizon for mobile, despite them being utter bastards I hate is reliable service and clueful support staff in my dealings. Granted, the last time I called them, the tier 1 tech couldn't do anything, but she quickly realised that the issue was beyond what tools she had available and I got escalated in about five minutes. Tier 2 tech fixed the problem immediately. And yes, the first tech I spoke to didn't have access to the systems to resolve what was ultimately an account issue. Thankfully, it has been years since I've had to call Apple about anything, but I've had nothing but good experiences with their support.
Apple once insisted on me paying for a replacement hard drive, which slowed to 1% of previous speed, without any errors, immediately following a major OS update. "Sometimes OS updates reveal previously-unknown hardware failures" the phone folks said. "genius" bar tech ran a diagnostic which came back with the single word FAILED. Neither knew what an actual hard drive failure looked like.
Both their phone and "genius" bar techs are clueless.
Perhaps I've been lucky with Apple support, with a quick chat solving my last software issue, and my one hardware disaster being largely self-inflicted.[1] PSU decided to go beyond blue smoke and I had to call the fire department just to make sure there wasn't anything still secretly burning in the chair cushion. Then I made the idiotic decision to take out my aggression on the offending part, only finding out the next morning that Apple wanted the original part back or was going to charge 5x the price on the replacement, and still didn't want to deal with me directly.
Oops.
I was fortunate enough to find a repair shop in Berkeley that was quite willing to lie on my behalf, saying that they'd have a certified tech install a new part and got me one for a far more reasonable price with a small markup for their efforts. But yeah, what you said seems to track regarding Apple. Trivial/easy support is fine, but if the rare "thing goes horribly wrong" incident happens to you, it's going to be an ordeal.
[1] I've lost a SCSI controller on a mac too, but it was a 3rd-party card.
I had Sky broadband for a while, their support was very good during business hours, but when the modem decided to drop dead one evening, I found myself speaking with a script monkey on other shores.
There were a painful few minutes and my temper rose until I resorted to something like "If you can't follow what I'm explaining, could I please speak with someone technical?" but that did the trick, and he agreed to ship a replacement without asking further questions.
Mine was agent insisting on me rebooting my computer, after I told him the DSL and Internet lights on the external modem were off.
On the other hand, once had an agent answer with "Hi, I'm <name>, and you're calling about the noise on the line, aren't you?"
In fairness, even well before AI, most FAQs I found already were obviously written by a single person, whose task was to reformat the user manual in the form of questions and answer. At no point of this process is anyone actually asking any question without knowing the answer.
I honestly believe lots of people don't even know what the acronym means - or, if they do, never give a second's thought to what it actually, really means.
No, that's unfair. I have known many intelligent, hard-working HR people (actually, more than most, as I used to write and support a personnel system, many decades ago), but what you have to realise is that your objectives are not their objectives. Once this is established, things normally go well enough.
GJC
Our benefits department is only reachable by email. They respond within a couple days... usually. Had one major incident, likely impacting several thousand employees, where the first response from Benefits was "should be fixed now" - two weeks after I verified it was fixed and a month and a half after I reported the issue. Not so much as a broadcast "you might see this issue, we're working on it" in the meantime.
FAQs are supposed to be questions raised by real people
Indeed. And should already be covered long-form by existing documentation. In the role of generating FAQs I don't see 'AI' as doing much more than scanning a document and producing an index. If it can figure out the questions people will ask, and what the answers are, as a short-form version of the docs; that's fine by me, saves me or someone else from doing that.
If it can flag up questions which are reasonable to ask which it can't find an answer to that's a win as well; that should probably have made it into the full documentation.
It doesn't surprise me that 'AI' does better at particular specific tasks. It's where I believe 'AI' will ultimately succeed. Unless it keeps hallucinating, insists on making up shit.
The FAQ battle was lost a *long* time ago. At best these days you might get an FAQ written by an actual serving member of the support team, but mostly they seem to be written at the same time, by the same people, and to the same dreadful standard as the rest of the documentation. Utterly pointless, in most case.
GJC
>but mostly they seem to be written at the same time, by the same people, and to the same dreadful standard as the rest of the documentation.
I once had "ownership" of a troublesome datascience system forced on me. A lot of the problems were self-inflicted, so every time I saw an issue twice I documented the hell out of it. This meant I just sent them the link and told them to come back if it's not fixed. After a while the data scientists started emailing the links to one another and the service desk also did the same. The end result was I could get on and do something useful and the data scientists could get back to work quicker. An added bonus was that any problems not in the FAQ were often hard enough to be fun to work on.
Maybe "supposed to" but, nah. FAQ's run out of steam quickly.
Using ChatGPT, one can instruct ChatGPT to read a pile of documents carefully and then one can ask question about specific things covered by the bumf. Those can be boring documents,like HR-policies and user manuals. ChatGPT doesn't care.
In My Opinion, It does that kind of task very well.
Same with source code that it has "read". One needs to know which exceptions "datetime.datetime.fromisoformat()" can raise, ChatGPT will give a very solid answer, a lot better than the snot-rollers and navel-defluffers faffing around on StackOverflow will do because most of that lot for sure hasn't read the docs before answering! ChatGPT will provide code examples too.
My significant other and I had a cultural mis-understanding a while ago. She had always thought Fondue was cooking meet in a communal pot of hot oil, and I had always thought that Fondue was a concoction of cheese, wine, and possibly a bit of flour.... I couldn't understand how you could possibly melt small pieces of steak and dip pieces of bread in it....
Have to say 3D TVs are generally one feature that can't be placed at the feet of the TV manufacturers :
People don't like wearing things to see content
There wasn't that much 3D content unless you have a Bluray player with 3D support
Some of the glasses were expensive for additional pairs
Beyond that the technology generally worked well. Some manufacturers (LG?) did an auto 3D conversion for game consoles etc that was very effective. Feed in a Wii U running Super Mario 3D World and the parts that should have popped out of the screen did, despite the fact the Wii U has almost no games that officially support stereoscopy.
It's ML!!
If these "biz leaders" could get their heads around the fact that there's no intelligence here, artificial or otherwise, maybe they might start to understand.
If I try to contact a company with an issue and get diverted to a chatbot, or even worse, a voice recognition system, I hang up and do not return.
I assume you mean "positronic brains" - that's what they were called in the 50ies.
They dont care, grow up and smell the coffee. These peiple who are pushing AI are nothing but professional bullshittes and liars who have no integrity of any kind. THey are just tryiing to pump and dump. THey couldnt care about whether they are actualy selling something that betters humanity in anyway.
is likely to be chained to Liar's Gate*.
there is NO intelligence - articifial or other wise.
it is not a lot more than just ingesting, processing, iterating with questions to a very large data set.
*= bolt holed, but not sure that'd get by El Reg's censors ;-)
It'll have a very good future in politics.
I wouldn't count on it
"Unfortunately, the financial benefits of implemented projects have been dismal,"
This is a usage of "unfortunately" that does not encompass those of us who experience a certain schadenfreude at the inevitable failure of guaranteed-to-fail projects that we could have warned people would fail if any of them would listen.
You invested millions or billions without a single coherent business plan, or often without a business case at all.
You had no frickin' plan or reason or rhyme for any of it other than 'oh everyone says we need AI'.
So of course you are losing your shirt on it.
The only people who make money on LLMs are: 1) The people providing the subscriptions/tech (OpenAI is obviously making out like a bandit), 2) SEO spammers and other generators of 'content', 3) people who don't have to pay for artists any more and can now crap out a new terrible porn game with AI art to put on Steam every week. 4) I'm sure I'm forgetting some other category, but it does not include J Random Corporation who just bought into the hype train.
OpenAI is most certainly not making out like a bandit. They are massively loss-making because the cost (mostly computing hardware, power and water to cool the whole thing) to run the models is truly astronomical. They can't charge fees high enough to break even because nobody would pay them if they were even close to cost (Microsoft is estimated to lose up to $100 a month per Co-Pilot subscription for the same reason). OpenAI is entirely dependent on investor money to stay afloat and will probably stay that way for ever.
The real profit is made by Nvidia which sells the hardware and Azure which rents out the computing power for the LLM companies.
I've been waiting for this:
The ability to speak does not make you intelligent as given to us by the sage Qui Gon Jinn twenty five years ago,
I think the problem is that like a lot of other trendy and buzzword technologies (e.g. Blockchain), people have bought into the AI hype without really thinking through whether it will fit their needs, and if so, how they can best use it.
Then they buy into the hype, and find either that AI isn't as good as they were lead to believe, or that their business has no real use for AI.
It's natural that those people will complain of a lack of return on investment, and look at cutting future investment.
Remember that while AI appears intelligent. it probably isn't. It's just good at mixing up it's input data in different ways and using that to produce apparently original output.
For instance, a while back, Chat GPT passed a US State Bar exam. They fed in years worth of previous exams, with the answers and it produced correct answers to the most recent Bar exam. That does not mean it is qualified to practice law though because it was just regurgitating answers from previous exams. It did not understand the law, or even the nuances of what it was answering.
I strongly suspect that lots of the AI induced job cuts will be down to the failure of hype chasing AI companies, either straight failure or being borged, and the resulting lack of need for the previous infrastructure meaning the supply chain will drop off too.
Large players, with the most money who can absorb the costs in the distant hope of making it profitable at some point in the future, will probably down size their teams in the next year or two to try and help balance the budget.
Just my predictions anyway. With better and cheaper to run hardware behind ML tech it can change at any time but the problem is that the more specific the hardware the more likely it is to be obsolete sooner.
I was referring to how the major market leaders have as their primary missiong as bullshitting.
In the past Ford built cars, Boeing built planes today the market leaders on the stock exchange are as their primary mission - bullshitters.
Tesla - need i say more ?
Facebook, Google, Twitter all about pushing BULLSHIT
NVidia,
Boeing
Hollywood
Music industry
I could list others but i believe the above are without doubt primary examples of my statement.
The best use case I have found for GenAI is when dealing with content produced by GenAI, such as GenAI-generated spam, emails and other documents. Thanks to GenAI, much of this is moved to spam folders or even filtered out completely, ensuring I never waste my time with it. As for any content that manages to slip through, I use GenAI to summarize it and reduce it back down to the short prompt written by the coworker to produce it. As such, I waste my time with only what the coworker himself wrote, not with what GenAI expanded upon. It definitely saves time when I reduce my exposure to GenAI-produced verbiage to the bare minimum (ideally to nothing if GenAI advances sufficiently).
AI is the first technology the bosses bought for us that doesn't come with any instructions, just a slip of paper that says "figure out how you can use me to make you more productive." Now all the sudden I got two jobs, all of my regular paid work, and now this new one to figure out how to use AI to make my job more productive. How come I'm so much more busy now, and still haven't been able to come up with a solid and fool proof plan to make myself more productive using AI? Someone tell my boss somebody else is getting the efficiency gains, and also please tell my wife this doesn't mean there will be more bacon? And one more thing, who's getting screwed here, my boss, me, or my wife?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
