Is AI going to pay its way? Wall Street wants tech world to show it the money The tech world has been banging on about the amazing benefits of generative artificial intelligence – so when are we going to see them? Cynicism aside, billions of dollars have been spent on this, plenty of employees are being pushed to use it, and Nvidia and others supplying the market with machine-learning shovels are facing …
Education being the best investment, one might approach analysis of AI ROI by analogy to medieval concepts of trivium and quadrivium IMHO, the 7 liberal arts[*], and moats (or not?). Basic LLMs manage grammar and rhetoric well (2/3 of trivium), and some genAIs might do ok on music (1/4 of quadrivium). Hard work is ongoing to incrementally improve them on logic, arithmetic, geometry, and astronomy (remaining 1/3, and 3/4, of tri- and quadri-vium).
In this vein then, one may follow TPM's NLP SQL front-end proposal, mix-in TM's Lean Deepmind article, and add TC's Coq piece, to obtain a potentially complete "real McCoy": a 7/7-shooter.
And that's the ROI right there! The computer is now educated. Who could ask for anything more!? ;)
[*] not to be confused with the 7 deadly sins, nor the 10 plagues of Egypt, etc ... (eh-eh-eh!)
I view LLMs as allowing a person to create what they consider to be "very good" / "excellent" in a very short time.
If you know what you are doing, or what excellent looks like, an LLM will give you great results. Maybe first time, or maybe with a few iterations.
If you are a beginner at whatever you are working on, or don't know what you are talking about, you will get very convincing crap.
With particular and peculiar regard to ......
The gist of this memo, written by an unnamed Google researcher, is that both Google – and we generally don’t call the company Alphabet because who really cares about the money-losing, non-Google pieces of the company – and OpenAI/Microsoft are going to be adversely impacted by a flood of open source AI frameworks, models, and datasets that will be as profound as the Protestant Reformation, which took the power of Christianity away from the Roman Catholic high priests and putting the words of the Good Book in the hands of the people in their own vernacular, explicitly and expressly for their own interpretation and use.
...... there definitely be A.N.Others and other researchers/luminaries/employees/employers of Goggle and Alpha Beta AI and OpenAI/Microsoft more than just simply capable of adversely impacting any flood of open source AI frameworks, models, and datasets that will be as profound as the Protestant Reformation without themselves being able to be either adversely impacted and/or successfully diverted from any singular or co-operative SMARTR Planned and IntelAIgently Designed Future ACTions by any sort of conventional remote third party means or fantastical virtual memes.
Such an AI Situation for Publishing is, like it or like it not, just the current expanding norm impacting upon the unfolding revelations relative to the news being shared and/or withheld regarding the dynamic virtual nature of your universal and Earthly existence outcompeting or competing with Google and OpenAI ‽ .
AI as it is defined today will have niches where it performs very well, but according to Deloitte: experienced organizations that are leaders in AI implementation typically see an average ROI of 4.3%, while those just starting out may only see a 0.2% ROI.
Those kinds of numbers aren't going to support the kinds of multiples and money currently on the board
This is nonsense. All of them got *extraordinary* good results by normal standards. A minimum of 15% annual increase in revenue and profit, for companies of that size, is unheard of. Let alone NVidia, up 260% revenue, 630% profit.
All that happened is that the increase isn’t as fast as analysts had budgeted in. But share prices are still higher than they were in May/June.
If AI can bring 15% growth rates, against historic 2-4% GDP growth, that’s unprecedented. But what *is* a problem, is that US employment data has been poor. That’s why share prices dropped.
Rising corporate income, is needing fewer people to do the work. Henry Ford knew that the population needs to be able to afford the cars they build. It’s just starting to dawn on people that we cannot avoid this reality. UBI is no longer a fantasy, it’s something we have to plan for. Maybe not today, but 10-20 years out. This bit of human history is going to be really rough. Groups that have historically been middle class and entitled will find that their skills are worth less than hairdressers and carers, and they aren’t going to like it.
"Groups that have historically been middle class and entitled will find that their skills are worth less..." People like me (and many smarter and better informed than me) have been saying that for over a decade. At this point, though, I'm divided on the truth of this. I think the current AI situation is a bubble: while there may be some uses for it, it won't be anywhere near as impactful as many thought. But 10--20 years out (as you say), the next-gen AI (I'm guessing some combination of neural net and symbolic), may be very different. At some point, barring a world war, a *lot* of people are going to become unnecessary in the job market. When that happens, watch out.
I think AI is too expensive (at least for now and perhaps forever) to replace cheap labour. The average salary in Bangladesh is around USD 2.5k per year, and you would be hard-pressed to get AI at a lower cost, even if it can perform as well, which it cant.
The average salary is around 35k GBP; given the state of AI in its current form, it would make little business sense to invest in the ability to replace the average person unless it can significantly reduce workload or increase efficiencies, like customer services etc
The further you increase the salary bands, the more attractive AI becomes. A solicitor who earns 70k a year might be replaced with a junior solicitor and AI. Same with underwriters, accountants and so on. Once you have fine-tuned your model with the correct data, do you need that senior, more expensive employee when you can combine a junior employee with AI
The problem is that the AI probably can't replace the senior solicitor. Because they're the one with the experience to know when it's utterly wrong - or to say that the output is good. But it also can't replace the juniori solicitor, because the senior one is too old to be good enough with computers to get it to work.
I think AI will be useful to bang out boilerplate text for first drafts of technical documents. So could save the legal profession money. It could do well on finicky but repetitive tasks like conveyancing - where you're spending loads of time making sure the information on different documents match, and that you've got the complete file of info - before the solicitor needs to check it for any problems. But I suspect that's going to need specifically tuned models for each sector.
Does it, though? One bad day the night before you draw down. See also the plethora of fund managers that continue to throw money at utterly implausible outfits touting cold fusion, amongst other snake-oil related claims.
I hate that our pension fund is limited to a range of fund choices and therefore visibility of what is in what fund and/or companies I want NOTHING to do with will be in amongst their product mix. To say nothing of keeping track of changes inside those funds. It is, of course, impossible.
The only alternative is moving the funds to a different provider where you have control over selections. And aren't ripped off to the high heavens by fees.
If it weren't for the double matched funding by the company on contributions, it would be better to do something else with it entirely, regardless of the perceived tax break.
Fall in results is because everyone fears Israel and it daddy USA is about to have a pissing contest with Iran and short of perhaps OPEC declaring another oil embargo or find another way to say Israel has been punished it assassination, there going to be war in the Middle East.
They simply using pretend poor numbers from the companies to cover it, so normal investors don't panic.
That phrasing needs some work. In my experience it's the other way around. Plenty of employees are pushing employers to use it...even more are using without their employer knowing...that's where the real "danger" is...companies hiring people that *gasp* don't have to work for the amount of time the company wants them to in order to achieve an agreed outcome.
I think the first thing AI will kill off is hourly rates...personally I'm saving huge amounts of time prototyping things with AI and therefore massively reducing the time required to knock an idea together to put in front of people to see if they want to invest in it. Without AI, it might take me a week or two...with AI, I can do it in an afternoon...so if someone decides the idea is shit, it's less impactful than if I'd spent two weeks or even a month on it...furthermore, because I can now chuck things together in an afternoon, I'm more likely to actually start work on it than put it off because I know I won't have to commit as much time to it.
I think people are getting the wrong end of the stick on AI, I don't think it's coming to replace people on the production end of things any time soon, but it will leader to greater innovation simply because people are able to get functional demonstrable prototypes working faster...the other "danger" with this, which again, affects big business...is patents...it's going to become far easier for someone with an idea to get to the point where they can get a patent than it's ever been...and it won't require the massive amounts of funding and backing in order to get there...therefore, larger businesses won't be able to go patent poaching.
Most of the danger with AI as it stands right now is aimed at big business and elites because AI is going to put a hell of a lot of capability in the hands of regular people for next to no cost which has previously been walled off and only accessible to big business and elites.
Current AI is to productivity what the original printing presses were to education. People aren't going to be out of jobs, the jobs out there are going to be out of people.
The expected effort for a prototype will stay the same. What will change is the expected quality and functionality of the prototype.
Back in the 90s and early 2000s we prototyped using Delphi or VB, just "drawing" a few GUIs and then if investors bit, we started implementation. That level obviously doesn't cut it today.
I think you are mistaken if you think you will stay competitive with prototypes that took like a day or less to create. Always remember that if you can do it today, 100s other will know how to do it tomorrow.
A prototype usually takes about 1-2 weeks because 80% of that time is to polish things out so they look smooth. Polishing always takes the most time.
For creative work I agree with others that AI will be a busted flush. It'll maybe help you knock out a few stupid memes or shortcut your homework but not sure it'll do more than that.
Much like autopilot in cars the humans will remain legally responsible so you can't really automate fully. Which will mean even if you use an AI for code or spreasheet work or data retrieval that work will still need to be checked by a human 'expert' which might not actually turn out to be much less work or faster than having done it yourself in the first place. We're already seeing embarrassing failures where the AI has been trusted to generate work without checking.
I've been playing with an GPT license and it is useful for low risk activities. If I put a large amount of data into it and ask it to do the spreadsheet manipulation for me then yes it's much faster, but for me this is low risk data and it doesn't really matter if the AI makes a few mistakes. But the license is expensive and my low risk activities don't really justify the cost. Similarly I could use GPT to build a scripted tool for other to use for compliance activities, but they'll need the expensive license in order to use it so again the occasional low risk need doesn't justify the cost.
Where AI/ML will work (is working) is much more boring. Predicting failures in large industrial or utility systems, predicting or detecting erosion or structural movement in infrastructure or buildings, high speed checking of medical scans for unrelated anomalies (i.e. MRI scan for one condition fed through the AI to check for other conditions). The business I work for has some impressive AI/ML systems already saving customers millions but they're not generative AI and not very exciting.
We had a demo of a generative design tool called Transcend last week. The developers are generally careful not to call it AI, although it is in line with current trends for generative AI.
It's interesting in the sense that it spits out a CAD / BIM model based on company definitions and the requirements specified by the user. The flip side of that is it basically only deals with greenfield sites and has no concept whatsoever of integration into existing systems. As such, a library of standard parts based on functionality would achieve more or less the same result - and already exists in people's heads.
One of the failings of the marketing spiel is that it assumes your company depended on hundreds of draftsmen. Which, it probably did in 1965. Today; the list of CAD users is measured by the dozen not by the hundred.
Does the tooling save those people time? Not really; not unless you can get it to account for the existing systems it needs to interface to.
I won't say such an idea is impossible to realise; because it's not. Realities of construction and access to do work however, mean that irrespective of how quickly you can prototype designs, the end product is not really accelerated any. At most it's a time saving for those CAD users. At worst, it's garbage in, garbage out, and could end up being a time sink just to drive it.
Can our ruleset and technical specifications be turned into something a computer model can consume? Yes, undoubtedly. Can it substitute for the lateral thinking a human offers? Never.
My employer is very big on AI and we all have mandatory targets for using GitHub copilot and our in house chat GPT. I’m using both everyday and I find them quite useful. That’s it, quite useful.
They’re good for boring tasks like prepping mock data for test cases and writing code doc strings and sometimes they can give clues on approaches if you’re trying something you haven’t done before.
I rarely get working code out of them, though they do give me clues. The other issue is the knowledge cutoff as they will be ignorant of new features in tool or library versions post 2021 and things move quickly these days.
I don’t think they are true AI solutions yet and I think the current hype is massively overblown. I wonder how many of the “prompt experts” popping up on linked in were cryptocurrency or block chain “experts” a couple of years ago!
It was very noticeable on youtube how the grift switched from crypto to AI almost overnight after ChatGPT launched. An awful lot of people who really should have known better got caught up in the hype. To be fair it was quite hard not to.
Machine Learning clearly has tremendous potential (as it has done for years) but so many of the capabilities that were being suggested as the immediate future of computing only last year were clearly overstated. As you say, for the most part it’s merely quite useful.
If the bubble bursting contributes to a market crash and recession there’s gonna be all sorts of repercussions for big tech you’d imagine.
If the bubble bursting contributes to a market crash and recession there’s gonna be all sorts of repercussions for big tech you’d imagine. ..... Jeff Smith
The upside to markets crashing and bubble bursting recessions is all manner of myriad opportunities for big tech you could never have imagined even possible before their arrivals on the Earthly scene and that has dire repercussions for all sorts of both Wall Street and K Street dinosaurs.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
