949
Microsoft seemingly just revealed that OpenAI lost $11.5B last quarter (www.theregister.com)
submitted 1 year ago by to c/technology@lemmy.world
225 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 39 points 1 year ago (18 children)

I wish. Even knowing it's all a gigantic scam, they'll first protect themselves before letting it burst and screw everybody else. The rich get a buffer period.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 18 child comments
    •
  • [+] -22 points 1 year ago (17 children)

    Is it really a scam when it creates content?

    It may be slop to you, it may not be useful for everything they market it as.

    But plenty of people find it useful, even if they use it for the wrong things.

    It is not like cryptocurrency, which is only used by people who want to get rich from it.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 17 child comments
    •
  • [–] 32 points 1 year ago* (2 children)

    It's a scam because the prices they're charging right now don't reflect the actual costs. AI companies are trying to get people and companies hooked on it so that once they crank the prices up by 10x to start turning a profit, they'll be able to maintain some semblance of a customer base. If they were charging the real prices a year ago, the AI bubble would have never reached the levels it has, and these companies wouldn't be worth what they are now. It's all propped up on a lie.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 40 points 1 year ago* (last edited 1 year ago) (1 child)

    Is it really a scam when it creates content?

    I create content in a ceramic bowl twice a day. Give me a billion.

    The scam is that the business plan is not feasible. Hundreds of techs have died because some cool idea could never make real money.

    And this is the finance model:

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 18 points 1 year ago (4 children)

    Is it really a scam when it creates content?

    No one is claiming that it doesn't output stuff. The scam lies in the air castles that these companies are selling to us. Ideas like how it'll revolutionise the workplace, how it will cure cancer, and bring about some kind of utopia. Like Tesla's full-self-driving, these ideas will never manifest.

    We're still at a stage where companies are throwing the slop at the wall to see what sticks, but for every mediocre success there's a bunch of stories that indicate that it's just costing money and bringing nothing to the table. At some point, the fascination for this novel-seeming technology will wear out, and that's when the castle comes crashing down on us. At that point, the fat cats on top will have cashed out with what they can and us normal people will be forced to carry the consequences.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 10 points 1 year ago (3 children)

    Exactly. Just like the dotcom bubble websites and web services aren't the scam, the promise of it being some magical solution to everything is the scam.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 5 points 1 year ago (2 children)

    Unlike the dotcom bubble, Another big aspect of it is the unit cost to run the models.

    Traditional web applications scale really well. The incremental cost of adding a new user to your app is basically nothing. Fractions of a cent. With LLMs, scaling is linear. Each machine can only handle a few hundred users and they're expensive to run:

    Big beefy GPUs are required for inference as well as training and they require a large amount of VRAM. Your typical home gaming GPU might have 16gb vram, 32 if you go high end and spend $2500 on it (just the GPU, not the whole pc). Frontier models need like 128gb VRAM to run and GPUs manufactured for data centre use cost a lot more. A state of the art Nvidia h200 costs $32k. The servers that can host one of these big frontier models cost, at best, $20 an hour to run and can only handle a handful of user requests so you need to scale linearly as your subscriber count increases. If you're charging $20 a month for access to your model, you are burning a user's monthly subscription every hour for each of these monster servers you have turned on. That's generous and assumes they're not paying the "on-demand" price of $60/hr.

    Sam Altman famously said OpenAI are losing money on their $200/mo subscriptions.

    If/when there is a market correction, a huge factor of the amount of continued interest (like with the internet after dotcom) is whether the quality of output from these models reflects the true, unsubsidized price of running them. I do think local models powered by things like llamacpp and ollama and which can run on high end gaming rigs and macbooks might be a possible direction for these models. Currently though you can't get the same quality as state-of-the-art models from these small, local LLMs.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 1 year ago* (last edited 1 year ago) (1 child)

    Re: your last paragraph:

    I think the future is likely going to be more task-specific, targeted models. I don't have the research handy, but small, targeted LLMs can outperform massive LLMs at a tiny fraction of the compute costs to both train and run the model, and can be run on much more modest hardware to boot.

    Like, an LLM that is targeted only at:

    • teaching writing and reading skills
    • teaching English writing to English Language Learners
    • writing business emails and documents
    • writing/editing only resumes and cover letters
    • summarizing text
    • summarizing fiction texts
    • writing & analyzing poetry
    • analyzing poetry only (not even writing poetry)
    • a counselor
    • an ADHD counselor
    • a depression counselor

    The more specific the model, the smaller the LLM can be that can do the targeted task (s) "well".

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 2 points 1 year ago

    Yeah I agree. Small models is the way. You can also use LoRa/QLoRa adapters to "fine tune" the same big model for specific tasks and swap the use case in realtime. This is what apple do with apple intelligence. You can outperform a big general LLM with an SLM if you have a nice specific use case and some data (which you can synthesise in come cases)

  • permalink
  • fedilink
  • source
  • parent
  • [–] 15 points 1 year ago (2 children)

    creates content? Out of what? I dont deny that there are some use cases for ai that are good, but ultimately its all built on backs of people who have actually contributed to this world. If it was completely non-profit it would be more okay, but as it currently is ai is tool of exploitation and proof that law protects only the rich and binds only us.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 4 points 1 year ago

    It's not that it's not useful for the end customer. It's more that investors are overpromised on the value and return from AI. There has been no returns yet, and consumers are finding less useful than these companies intended. The scam is for the investors, not the end user

  • permalink
  • fedilink
  • source
  • parent
  • [–] 2 points 1 year ago

    enough people use crypto now that it's unlikely to crater entirely...unless western governments finally kick out the neolib types and take back their country's from private equity/big business

  • permalink
  • fedilink
  • source
  • parent