Dandelion Break The AI Business Model Is Insane (And Probably Evil)

[AgentPete]

For those people who follow the wacky world of AI business, the perennial conundrum is – how are these hucksters ever going to turn a profit?

Much has been written, by far wiser heads than mine, concerning the dire prospects of any AI industry ever making a positive return on the gargantuan amounts of capital currently being sunk into it. It’s simply not a viable industry, under prevailing circumstances.

So is it a vast bubble?

Until recently I used to think so. I’m not so sure, now.

My initial thinking was as follows. Everything the industry is doing at the moment really amounts to classic loss-leader activity. This is a traditional way that internet startups (and others) have sought to bootstrap themselves: give the product away freely to start with, and eventually you’ll have insinuated yourself into enough punters’ lives that some of them can’t live without it – and those are the ones who will become paying customers. Voilà… Your business succeeds and you’ll become the world’s first trillionaire!

That’s not necessarily going to happen with AI.

Studies show that despite the tsunami of hype, most people rarely use AI. Of those who do, the vast majority don’t pay for it (i.e. it comes bundled into your Outlook or your web browser for free).

Looking at ChatGPT, it has over 700 million users worldwide… yet only 10 million are actually paying customers.

That’s a pretty awful figure when you consider that they’re trying to raise $40 billion alone this year. In fact, the math doesn’t compute. They are never going to make money from captured consumers.

“Godfather of AI” Geoffrey Hinton suggests that this isn’t, and has never been, the real business plan. He believes, and he could be right, that the real payback will come from “labour replacement”… in other words, by wholesale enforced redundancies across wide swathes of the employment spectrum, focusing on those areas that pay the most. Most professional sectors will simply disappear.

If that’s the game plan – and the AI industry is suspiciously silent on this all-important question – then we need to have the biggest and widest possible public discussion. Because we’re looking at nothing short of societal revolution.

Reality check. Is present-day AI capable of massive employment sector replacement? I really don’t think so – yet. Anyone who’s messed around with current-level AI will know that it massively over-promises yet underwhelms on delivery.

Now, I suspect that’s why the industry is so desperate to raise such astronomical sums of capital investment at the moment – they know the tech isn’t at “job replacement” level yet, but are betting that it will be given given the scale of investment they’re currently seeking.

It’s a punt. One that, if they win, means the rest of us will lose catastrophically.

Yet, our politicians seem almost ridiculously naive to this outcome, e.g.:

Artificial intelligence presents a "vast potential" for rejuvenating UK public services, Prime Minister Sir Keir Starmer said on Monday.
In a speech setting out the government's plans to use AI across the UK to boost growth and deliver services more efficiently, Sir Keir said the government had a responsibility to make AI "work for working people".

- PM plans to 'unleash AI' across UK to boost growth - BBC News

The scale of current investments, believes Geoffrey Hinton, suggests the transition is already underway, and without deliberate policy intervention, the trajectory points to widespread [jobs] displacement alongside concentrated profit.

Sounds like dystopia to me...

 

[Aethalope]

The UK might still get away without the PM's plan to unleash AI, but other countries are increasingly going for it. South Korea going to start experiencing acute labour shortages very soon (and is really not keen on immigration). It's an important global supplier of components for things like telecommunications, renewable energy, and high-tech shipping (and K-Pop, of course, that's especially important). The power and control of the chaebols offer a good structure through which to integrate AI into industrial processes, and their government has just made humongous deals with Nvidia.

It may be the first country trying to AI its way out of demographic and economic collapse. It's worth keeping an eye on over the next few years to see how it fares.

 

[Pamela Jo]

The one benefit that I could see, IF accurate, are the reports that AI can diagnose patient symptoms faster and with less failure than human doctors. I mean it doesnt have to check notes it made in med school 25 years ago so... If the unequal medical care could be equalised so that all people get the best instead of the present system- I mean someone has to go to the guys who were at the bottom of the class, right-then it would pay for itself. But as several people have muttered aloud, "How do they expect people to buy stuff if they have no work?" Ireland unfortunately knows the answer to that as famously argued by Johnathon Swift. My parents grew up in the Depression. People were starving to death. And as the current US admin is demonstrating at home and in Gaza -they are unafraid to let families eat grass again.

 

[Pamela Jo]

The UK might still get away without the PM's plan to unleash AI, but other countries are increasingly going for it. South Korea going to start experiencing acute labour shortages very soon (and is really not keen on immigration). It's an important global supplier of components for things like telecommunications, renewable energy, and high-tech shipping (and K-Pop, of course, that's especially important). The power and control of the chaebols offer a good structure through which to integrate AI into industrial processes, and their government has just made humongous deals with Nvidia.

It may be the first country trying to AI its way out of demographic and economic collapse. It's worth keeping an eye on over the next few years to see how it fares.

Japan has also been panicking about the baby boom of old people vs young since I was there in the 80's. It makes sense they have been roboting for about 40 years now

 

[Sarit]

Like you said Pete, the goal is not to make money now, but to revolutionize society. It is a race of who will control the next 100 years and attain AGI first. It's not a bubble. Blue collar will become the new white collar. Premium on 'human' content/services etc. I for one think it's great to do away with laborious and draining administrative/computing tasks and subjective human gatekeeping.

 

[Pamela Jo]

 

[Pamela Jo]

Seems like the SOA would be onto this. It;s one reason I think audiobooks should be on the same platforms as independent music. Maybe, Your constellation should include gaming and music, because readers will have similar tastes

 

[AgentPete]

. I for one think it's great to do away with laborious and draining administrative/computing tasks and subjective human gatekeeping.

I agree! But… this is a strategy by stealth, which has to raise serious suspicions.

And I have zero – absolutely zilch – idea of how society would look post-AGI. Does anyone, in fact?

The most obvious issue being, what do we do with the millions of recently de-jobbed?

We’re getting into some deeply weird realities here. Need to start talking about them.

 

[Jake E]

I don't get it.
If you replace people in their jobs, they don't earn any money.
If they don't earn any money, they can't buy things.
So... Who are they going to sell stuff to?

 

[Pamela Jo]

I don't get it.
If you replace people in their jobs, they don't earn any money.
If they don't earn any money, they can't buy things.
So... Who are they going to sell stuff to?

They aren't going to bother. Once the capitalist model is gone there is feudalism. Ask Ireland and the American Indigenous what happens when the land and all the food it produces is taken.

 

[Pamela Jo]

 

[Bloo]

The cake is a lie.

 

[Jake E]

The cake is a lie.

 

[Katie-Ellen]

Techno-feudalism. The original serfs only got a break thanks to plague. But that was when the feudal barons were still in situ and needed people to herd the livestock and till the land. Plumbers, joiner, fitters, multi disciplinary skills, short in supply now compared with 20 years ago. They can't keep up with demand round here. Jobsworths are always a problem, but gatekeepers absolutely need a quick and finetuned capacity of subjective judgment, warts and all. I was feeling almost ready to riot a few weeks ago but settled for buying a camping stove and butane. Because zombies.

 

[RK Wallis]

If they don't earn any money, they can't buy things.

or pay taxes...

 

[AgentPete]

I don't get it.
If you replace people in their jobs, they don't earn any money.
If they don't earn any money, they can't buy things.
So... Who are they going to sell stuff to?

Well exactly.

But a peek into some of the disturbing notions now being openly discussed by the technocracy does suggest that PJ is right: their “democracy is dead” meme will easily morph into the new feudalism.

Young people are increasingly dissatisfied with democracy and probably won’t fight to keep it alive.

Now is the time for writers, in particular, to use their skills in defence of the freedoms we have today (albeit imperfect). The values we fought two world wars over. The future can no longer be taken for granted.

 

[Pamela Jo]

l. I was feeling almost ready to riot a few weeks ago but settled for buying a camping stove and butane. Because zombies.

Ah but the zombies today dont have to be all the way dead. Just dead inside. They can still do AI prompts.

If you want to be talked into rioting...

The Rise of Neo-Feudalism - The American Prospect

The private capture of entire legal systems by corporate America goes far beyond neoliberalism. It evokes the private fiefdoms of the Middle Ages.

prospect.org

 

[Pamela Jo]

TH White did this beautifully with wit, romance and the ability to make you think. The audible version is good. Imagine addressing future generations about the need for democracy while your country is under attack from fascists and doing it in such a way that young people want to read it.

The Once and Future King Books In Publication Order​

1. The Sword in the Stone (1938)
2. The Witch in the Wood (1939)
3. The Ill-Made Knight (1940)
4. The Candle in the Wind (1940)
5. The Book of Merlyn (1977)

Now is the time for writers, in particular, to use their skills in defence of the freedoms we have today (albeit imperfect). The values we fought two world wars over. The future can no longer be taken for granted.

 

[anaximander]

The one benefit that I could see, IF accurate, are the reports that AI can diagnose patient symptoms faster and with less failure than human doctors.

This is a valid, but very narrow business case. Would you trust AI to make healthcare decisions for you or your loved ones? What if that AI was made by a health insurer? What if you had to sign a EULA agreeing with an AI making mistakes before getting medical advice, signing away legal recourse?

 

[AgentPete]

This is a valid, but very narrow business case. Would you trust AI to make healthcare decisions for you or your loved ones? What if that AI was made by a health insurer? What if you had to sign a EULA agreeing with an AI making mistakes before getting medical advice, signing away legal recourse?

And, I think, there’s a bit of blurring the boundaries between Machine Learning (ML) and Large language Models (LLMs).

Machine Learning has been with us for a while now and can be useful in specific circumstances (I’m thinking about diagnoses etc).

Machine Learning focuses on algorithms that learn from data, e.g. “this is what a benign mole looks like”.

Large Language models, by contrast, are a specific type of ML model designed to understand and generate human language. Which is why some folk simply call LLMs “bullshit machines”.

 

[Pamela Jo]

This is a valid, but very narrow business case. Would you trust AI to make healthcare decisions for you or your loved ones? What if that AI was made by a health insurer? What if you had to sign a EULA agreeing with an AI making mistakes before getting medical advice, signing away legal recourse?

Medical researchers are testing programs and if the test results can be trusted... AI takes the symptoms and test results and matches it with a diagnosis far more accurately than local GPs. And especially interesting for women it doesnt follow a "whiny woman" trajectory of dismissing symptoms becasue the patient is overweight, and over 50. So yeah, I'd be VERY interested in hearing a diagnosis from a program geared to actually take note of my symptoms, then search all the medical data bases without deciding to just hand over antibiotics and NSAIDS. In other words what rich people get when they go to a doctor.
These are of course a medical programs-nothing to do with Chat GPT. The mystery remains. If there are actual scientific and medical uses that could monetise AI why are the "great" minds trying to addict us to Chapt GPT sort of models?
The ridiculous amounts invested suggest it is simply because of the control, power and imaginary money that social media has returned. That worked, so AI has to be the next step from that . DUH It's evolution. But they dont understand that is not how evolution works. Real evolutionary models dont duplicate results. If we were to reboot earth we wouldnt necessarily get the same species evolving to look the same way they do now.

And frankly Anaximander your 'what ifs' are already normal for millions of people.
So what if a government that wants to provide universal healthcare equally to all people but cant because they dont want to pay the doctors and nurses they've trained with taxpayer money enough salary to live on so that they emigrate... can actually fulfil the social contract. Even venal, trickledown governments (Not including the current US admin) will probably opt for giving poor people the same quality of care - if it's even cheaper than the shitty unequal system.

A few years back Ireland offered free cancer screening to hundreds of thousands of women then outsourced the testing to US labs who sent back unreliable results. Upshot women dying of cancer becasue it wasnt detected-and not only can't they sue the government for malpractice-the gov told the women they are on their own suing the US firms-because the gov has to stay neutral.

 

[Sedayne]

All GP surgeries in England are already using AI. This is government mandated.
There have been plenty of fuck ups already which the humans have to apologise for.
I work with this every day in general practice. It's not all bad. The staff are more keen on it than the patients are. But it's not so simple as replacing doctors with AI.
I've just finished late off a nine hour shift with no break and am too tired to go into it all now, but there are good reasons to have a human diagnose anything more than the simplest pattern recognition cases, particularly for elderly medicine (over 65, sorry, that's the official age - isn't it dreadful?) and palliative care where things get extremely complicated. The problem comes when the NHS scrimps and uses inexperienced or unqualified humans to do the diagnosing. (Or the just plain jaded and uncaring docs who shouldn't be in the profession - there are still some of those around.) But AI can't do the complicated stuff - yet. And dying patients still want humans around oddly enough.