So there’s something I need to get off my chest, and if I post it on my LinkedIn it would be career suicide at my level.

In a company, the largest line item by far is usually payroll. I have been at a number of companies that are trying to cut costs and don’t care if you come up with the correct amount via other OpEx categories, they want headcount reduced because “it’s so much”.

So along comes the promise of a computer bot that understands the normal person and also:

  • Does not require sick/vacation time
  • Does not take FMLA
  • Does not want a bonus/profit sharing/equity
  • Don’t have to pay unemployment taxes, Medicare or SSI
  • Does not require them to spend money on health insurance
  • Will not form a union
  • Wont ever file a lawsuit for any number of reasons
  • Will work 24/7

And this right there is the exact reason so many CEOs are salivating at the idea of AI. Not for worker efficiency, not for any number of “positive” benefits they may taut, but they finally have a glimpse of the chance to rid themselves of one of the largest headaches that they perceive in a company.

  • Aceticon@lemmy.dbzer0.comEnglish
    10·
    4 days ago

    Well, current High-level Management Culture is very short-termist and mainly anchored on Priviledged Upbringing, Salesmanship and People Skills (networking, spinning good stories, growing up in the right families and so on) rather than Competence in Strategy or Analytical ability.

    As it so happens, the upsides of AI are of the immediate and obvious variety - basically as you listed - and pretty similar to those of outsourcing (but on the shorter term AI is actually cheaper). Meanwhile the downsides are mainly longer term costs and risks, and/or derived from complex or systemic characteristics of AI, for example:

    • LLMs have a flat profile in the gravity of the consequences of its errors: basically they are just as likely to make errors with serious consequences (say, advise somebody who expressed suicidal thought to “kill yourself”) as they are to make errors with minor consequences, whilst even untrained humans have some awareness of consequences and thus have a lower probability of doing higher consequence errors. This means that over the medium and long term AI has a worse risk profile than even untrained people and is thus more likely to do things which bankrupt companies and kill people.
    • LLMs don’t learn, at all. You can explain it all you want but even if that helps, it only helps as far as their input window contains that explanation since that stuff doesn’t actually feed back into the model so doesn’t get persisted. Most people do learn, and once they learn they seldom go back. So if one position is occupied by a Junior Something and another is “occupied” by an LLM, in 2 years time the former will have become a lot more productive and capable whilst the latter will be no better than 2 years before (possibly worse: see below).
    • If you’re running AI hosted by a 3rd party, you’ve now become dependent on that 3rd party, with all the associated future risks that it entails, especially for companies which have far less money and thus legal power than said 3rd party. The most obvious is that what’s “cheaper than hiring” today will not be so tommorrow once that 3rd party has locked your company in.
    • More generally, most AI gets trained from materials which were gathered from public contributions and those are getting worse, both because people are now less likely to end up in those places and ending up giving their own contributions and because “contributions” done by AI or with AI generated text claiming to be from people are increasing in those places and, as it has been already been provide in actual scientific studies, the quality of AI goes does the more AI-produced content is used in the training of a new AI, so mid and long term as the fraction of training materials which are pure human-created falls, AIs will get worse, not better (at best, stop altogether improving as no new models are trained), which for companies replacing human labour with AI means that the quality of the work done for them will never improve.
    • Then there’s the even more broad and systemic problem which also happens with outsourcing: if nobody employs people today for their Junior positions, tomorrow there will be no Seniors.

    All this to say, that since the “qualities” of present day upper management are seldom in areas like Strategy and Risk Analysis whilst the problems of AI are mainly complex, long-term and/or systemic whilst its upsides are mainly of the highly visible and immediate kind (thus easy to stop and appealing for Salesmen and Tactician types), it totally makes sense that so many CEOs are going into AI.