Sören GadeMar 15
Jeff Johnson

LLM access is relatively cheap now because the LLM vendors are discounting their price at a massive loss, subsidized by VC, in order to get you addicted and to drive as much skilled human labor as possible out of the workforce permanently.

The goal is monopolization, and if they’re successful, you’ll see monopolistic pricing in the future.

Mar 8 at 11:14amWeb
Show threadfan boy 🪭Mar 8

@lapcatsoftware
The same model of all rent seeking services.

And drug dealers.

Show threadRobert HMar 10

@notyourfanboy @lapcatsoftware

The first one is free. The second is on me.

Show threadWilliam O'ConnellMar 8
@lapcatsoftware But will it work? There seem to be too many labs making models of similar quality, and often it's effortless to switch between them.
Show threadDaniel GibsonMar 8
@williamoconnell @lapcatsoftware
so what?
they all make a loss, and eventually need to make money
Show threadWilliam O'ConnellMar 8
@Doomed_Daniel Generally they aren't making a loss on inference currently, they're just spending so much more on training new models that they have a loss overall. But also some of the competition is from open source models, and the cost of inference for those is just the cost of compute, which goes down over time.
Show threadGabriele SveltoMar 8
@lapcatsoftware already happening in some corners of the market. I know a person who became professionally dependent on using Perplexity. They gave them a 12 months free plan, cut short to 4 then asked to pay 20€ a month. After 8 months this plan was also eliminated and the price is now 200€ a month for less than what they used to get for 20. Obviously they're in a bad place now, and Perplexity is still likely losing money over that.
Show threadEnola KnezevicMar 14
@gabrielesvelto @lapcatsoftware What do you mean by "professionally dependent"? Can they not start thinking again? Brain plasticity and everything?
Show threadGabriele SveltoMar 14
@rhelune @lapcatsoftware I hope they can, but they modeled their workflow so much around these tools for an entire year that it's hard for them to get rid of it
Show threadEnola KnezevicMar 14
@gabrielesvelto @lapcatsoftware I get it, a vendor lock.
Show threadCedar Fen Farm Cedar Fen FarmMar 8

@lapcatsoftware

Drug addiction tends to cut ones real world life very short no matter what fantasies it gives you.

Show threadpoleguy looking for lost toolsMar 8

@lapcatsoftware I heard that's not the end game plan. Rather it is to start stealing all the good ideas created using the models and then provide better versions of those so that nobody just using a model can compete. I have no hard evidence for this alternative end game... Just speculation I heard, I think.

Do you have evidence that it is just going to be simple rent seeking?

businesses using the models will be the victims... And the public who will absorb the damage as this bubble pops.

Show threadWouter VerhelstMar 8
@lapcatsoftware
And while they're at it they're also driving up prices of RAM so you can't do normal computing anymore even if you wanted to.
Show threadPetra van CronenburgMar 9
@lapcatsoftware Never forget the immense price we all pay for the destruction of climate and drinking water! If we use them or not, the data centers destroy our environment.
Show threadPollinatorsMar 9
@NatureMC @lapcatsoftware Yes. The data centers are accelerants. The billionaires and fossil fuel industry are racing to raise the temperature. They are inconsiderate. #noai, #nowaterfordatacenters
Show threadCassandrichMar 9
@NatureMC @lapcatsoftware Even worse, they destroy the informed and educated populace on which any hope of meaningful democracy and liberation exists.
Show threadArthur van der HargMar 9
@dalias @NatureMC @lapcatsoftware My take: this is happening to quickly to succeed. Adoption is too fast. And people are already realising that LLMs are starving the experience pipeline. But there are still plenty of experienced pre-LLM people around to restart it. Had this taken place over 20 years, we would have had a real problem.
Show threadPetra van CronenburgMar 9
@dalias Indeed. I don't know which one is worse, climate change doesn't make a difference whom it kills (the super-rich will be protected some longer). And a worsening with more extremes and catastrophes can also kill democracies ...
@lapcatsoftware
Show threadBrian DouglasMar 9

@lapcatsoftware the monopolisation of knowledge. There is a future where people are only as smart as the subscription they pay for.

Big tech wants you to forget what you know and become reliant. Using your brain is a threat to their business plan.

Show threadNathan SalwenMar 10
@brianwdouglas @lapcatsoftware Sites like stackoverflow.com are already being starved of content, making AI more essential.
Show threadAsta [AMP]Mar 9
@lapcatsoftware It's wild to me how many people ignore this; you're paying money to "do your job"? Things you could be doing without giving up control of your work output to a third party?

And this is the cheapest it will ever be? That is a terrible investment.
Show threadJohn MaxwellMar 14

@lapcatsoftware Yeah, well, the "good" news is that plan won't work, because LLMs don't work.

Of course, we're all still going to have to pay the price of the grift, so that doesn't help any.

Show threadJon SharpMar 14
@lapcatsoftware this seems likely enough.
Show threadBoyd Stephen Smith Jr.Mar 14

@lapcatsoftware I feel compelled to mention there are models you can self-host. There are even models where the architecture is available under a permissive license, so you can tweak / tune / retrain / distill or whatever beyond mere prompting.

I don't recommend or defend that approach. I think there are still problems, ethical and other.

But, it could be a way to prevent "vendor lock-in" with your LLM usage.

Show threadJonathan LamotheMar 14
@BoydStephenSmithJr @lapcatsoftware I have yet to find a use case for LLMs that I'd actually trust them with.
Show threadBoyd Stephen Smith Jr.Mar 14

@me @lapcatsoftware Generating test data, as a complement to QuickCheck/SmallCheck generators. I think LLMs might "explore the probability space" in different ways than manually written generators. But, I haven't validated this in practice.

I've been fairly disappointed with LLMs output all the times I've tried them. Too many hallucinations around factual data. Too little... variety(?) when doing fiction. The image generators seem better than me, but I have declined to use them (much)because I assume the image generators are "stealing" from the recognition/attribution of artists that make their art publicly visible. I know the code generators "steal" copyleft code, most likely including mine.

I don't like saying LLMs capabilities are bad, because I don't use them, for ethical reasons, enough to really know what their current capabilities are.

Show threadsmegMar 14
@lapcatsoftware businesses would be smart to NOT encourage adoption, as it's easier to divide and exploit skilled labor than it is to negotiate pricing with only a handful of behemoth companies
Show threadel Celio 🇪🇺 🇺🇦Mar 15
@smeg @lapcatsoftware most pretty much prefer to pay the big company than negotiate different labour organization with employees
Show threadArcticulateMar 14
@lapcatsoftware Yes, we saw this in Europe, in the Nordics, between home-electronic retailers back in early 2010s. The German company Media Saturn launched the German-worded ”MediaMarkt” in Sweden, with crazy low prices. They tried to get their competition out and then the plan seems to have been shock-raising prices on electronics, especially TVs. They failed, though: their main competitor was the same whale-sized behemoth as them, more or less, but smaller chains collapsed.
Show threadAxelMar 15

@lapcatsoftware @cdfinder

“Business as Usual”, for monopolists.

Show threadElad HenMar 15
@lapcatsoftware
Thing is, it can go on only if you believe line go up forever. And it can't. It's bait withou switch. Bait and crash.
Show threadRS, Author, Novelist, ProsaistMar 15
@lapcatsoftware The upshot is to keep your no-AI skill set up. In no more than 5-10 years, whatever skill was superseded by AI will be in demand. In the meantime? #theComingLaborApocalypse
Show threadGabriele SveltoMar 15
@lapcatsoftware it's important to keep in mind that it's not only the LLM vendors subsidizing the models, but their entire supply chain subsidizing them. There is not a single company in the AI business making any money, and they're piling up debt on top of the losses. The only companies making a profit are the like of Nvidia, TSMC or server vendors. And if you think about it none of those companies is spending on AI, so the money only ever goes one way. It simply cannot work.