Close Menu
    Technik

    KI kann jetzt mehr kosten als menschliche Arbeitskräfte

    Share.
    View 38 Comments

    38 Kommentare

    4. Mike-Banachek on

      Yes and by paying more their virtue signaling their distain for human labor. They’re not invincible though and accountability is coming.

    5. strosbro1855 on

      I just don’t think that’s going to stop these companies bc at the end of the day, they will never see humans as humans.

    6. topscreen on

      Outsourcing often costs more than domestic workers in the long run, but that never stopped companies before

    8. K_M_A_2k on

      The general consensus is after the bubble bursts the realistic sustainable version of this will come to market not the general ask something about coding, and a therapist, and recipe, and make a picture, and and and

      Just a bunch of specialized agents to do specific tasks really well for businesses.

    10. gravtix on

      Seeing the article soft paywalled behind this just made me laugh:

      > Axios AI+: Catch up on what’s new and why it matters in just 5 minutes.

      > Sign up for Axios AI+ to continue reading for free.

    11. TheMyzzler on

      Wait until companies have replaced tons of processes with AI and then the real price is charged. Gonna be a bloodbath.

    12. mantrakid on

      But the biggest issue with human workers is just how annoyingly human they are. Can’t put a price on not having to deal with that shit.

      /s?

    13. iamagainstit on

      who would have guessed that the speed of enshitification of AI is even faster than the speed of AI hype.

    14. Eskipony on

      A 1m token Opus 4.6 model carelessly used cost us a few hundred bucks in a single session. Unless companies start teaching people how to choose their models and manage their context windows effectively, its going to be a mega budget explosion, especially since the AI companies are willing to charge B2B much more.

      Those tech companies that are putting performance metrics on tokens used are probably feeling pretty stupid right now after looking at their bills.

    17. Hot_Lava_Dry_Rips on

      If youve looked at enterprise versions of these tools, youd know this is absolutely correct.

    20. thinkB4WeSpeak on

      If we didn’t give deals to data centers for electricity and tax breaks they’d really be paying, which they should

    21. teknoob on

      Not at all surprising. And it will get more expensive when investors stop subsidizing and start looking for returns.

      There is one caveat though, inferencing is relatively quite cheap. Most of the expense is in training the frontier models.

    23. EuropaWeGo on

      It’s only going to get worse. The older agents are being dumbed down while the newer more capable agents are costing 3 or more tokens per request.

      My team was pretty much forced to switch to Opus this month because Sonnet was basically nerfed to being unusable. Which caused us to hit our token limit 10 days into the month. Management approved a larger budget, but it’s not cheap anymore and my manager is getting worried as the execs want us to use agents even more, but are weary on the price. Making this a double edged sword situation.

    24. silatek on

      Anyone have a gift link or smth?

      I’d love to read it but I can’t be arsed to dodge the paywall atm. TIA

    25. LosMorbidus on

      Rember when google realized that making the search engine worse will make them more money?
      If they give you the result immediately you’ll spend 20 seconds on the site. If they don’t you’ll try again and again, see more ads and they make more money.

      Wait until AI companies do the same with your token budget. You spend 1m tokens now to get the result? What if it’ll take 2m next week? 3m next month?

    26. Vespene on

      The only reason you can use frontier models is because the compute/token costs are heavily subsidized by the immense capital investments into AI shops. Your subscription payments alone are not enough to pay for the immense amount of energy and infrastructure needed to generate the outputs you’re getting. This is the same as when Door Dash cost very little to deliver, or how streaming platforms cost less than half of what they charge now. Once the AI platforms settle and establish their markets as essential needs, the prices will triple or quadruple.

    29. Migb1793 on

      Blockchain this, blockchain that – cycle
      Now AI this, AI that – cycle
      What’s next ?

    30. thedoommerchant on

      To be expected. I just do not see how AI can be used at scale at enterprise orgs without huge costs incurred.

    31. jokysatria on

      Tbf, user of AI need to pay electricity bill more rather than local people around data center.

    33. smiling_seal on

      Soon everyone will realize, that AI “workers” is a massive single point of failure because SUDDENLY: AI providers can go down due to a failure, tokens ran out, and your ENTIRE company’s work-force simply doesn’t do a work, because no one can replace it.

      Then “suddenly” companies will learn that AI is opposite to much tolerate humans: AI must be paid in advance otherwise it won’t do a work so delaying payouts is no longer an option, AI doesn’t do anything beyond a paid bill (token limits).

      Eventually it will became obvious that providers need ROI (Return Of Investments) for spent trillions so they will start sucking off everything from their customers and bills will go insane overshadowing spendings on humans. This will be worsened by a massive “vendor lock-in”, because migrating to another AI provider will be a “faith jump” as models aren’t programs that always output X for given Y but random heap of bytes “somehow” trained from “some” sources and somehow “reasoning” and no one can tell how it’s going to work for you.

    34. Salt-Detective1337 on

      Dear medium sized corporations,

      What did you think was going to happen? This has always been the plan. They’ll goad you into firing your whole staff, they’ll wait till there are no longer any humans qualified to do the jobs. Then they’ll jack up the prices and milk you until you **also** have no money.

      Go fuck yourselves.

    36. AlarmingTurnover on

      Judging by the comments, nobody actually read the article. It quotes 1 single company as spending all its IT budget on AI, which is not a measurement of any tech company ever. And it doesn’t give any information on what using all their budget on tokens produced over this amount of time. It doesn’t show the productivity vs man hours saved from this. There is zero research or evidence presented. 

      But good job Reddit, continue to comment on bullshit ragebait without actually reading anything. 

    37. oritfx on

      It’s a repetition of the tactic that was used with Uber or Netflix: enshittification.

      1. New product uses dumping prices, the difference is financed by venture capital. Do you remember how great Netflix was? How cheap was Uber? How good was google?

      2. The competition dies out.

      3. Now that the market is cornered, the customers get the shaft.

      Bonus round 4: now that the advertiser market is cornered, the business clients get the shaft too. Everyone does. That was the idea from the start.

      Grand Finale: ask YAHOO. or AOL.

    38. bearded_mischief on

      It’s a very expensive technology that was extremely subsidized to get a lot of organizations to sign up and once they do they are dependent.

    Leave A Reply