How you can fix AI problems

Sat, 02 Aug 2025 01:21:33 -0400 last edited: Mon, 04 Aug 2025 18:50:35 -0400 from ArrCay

Seems to be alot of news reports lately about data centers' increasing cooling and power demands. This is a problem of bigtechmegacorp AI. It's not so much of an issue with AI in general.

The first obvious answer is really quite simple. Stop feeding bigtechmegacorps.

Secondly, I've promoted decentralized services for many years. I think it may be even more important for AI systems.

As to my first point - "Yeah, easier said than done", right?
The sheeple using bigtech products generally see no reason to stop or select a better alternative.

Most people don't really care [enough] about things like privacy, their personal intellectual property (aka content they post/create) or copyright, whether their data/content/habits are being harvested, analyzed and sold or used to train generative AI, etc. But they will complain about fake/scam/spam garbage, and all the advertising, censorship and various features or usability problems.
Still for some strange reason people on Fecesbook (or fill in the blank with practically any bigtechmegaco) seem to have no interest in using a platform which has better technical features and doesn't have all the garbage. The whole dopamine/addiction triggers and the huge sums spent on applying such psychological research data to their interface/platforms must be a big part of that "some strange reason."

It's time to bring "Buy Local" [back?] to 'cloud' and other IT services

After all, service xyz or platform abc is "free". Most people will acknowledge that it's not really free. However, their concession, and understanding, typically goes no further than the basic 'yeah they have ad's; I'm ok with the ad's plastered everywhere, I can just ignore the ads.' Even the advertising portion goes deeper than most realize.

For example, take the one I call Fecesbook (because it is now quite a cesspool). I won't even go into the privacy issues, selling user data, violation of FTC orders that has them in the news again. For now, let's just consider the basic ad stuff. In addition to the typical advertising plastered everywhere and injected into one's feed/stream, a few years ago they started a different kind of ad posting that appears like any other normal post. I don't recall what it's technically called but I believe the option to buy such paid posts are available to anyone. While there may on rare occasion be a legit individual or small business attempting some real advertising/marketing, these posts are mostly bought by spam/scam/fake accounts. What matters to bigtechmegacorp (aka Meta, in this case) most is revenue.

Why the legacy social media model of bigtech is destined for the dumpster

However, it's not just the ad's. If anyone has read a "Terms of Service" agreement lately, one would notice that most of the bigtechs are using their data (posted content aka their intellectual property or copyrighted works) for AI training. Here's the relevant section excerpts from the X TOS (emphasis added).

By submitting, posting or displaying Content on or through the Services, you grant us a worldwide, non-exclusive, royalty-free license (with the right to sublicense) to use, copy, reproduce, process, adapt, modify, publish, transmit, display, upload, download, and distribute such Content in any and all media or distribution methods now known or later developed, for any purpose.
...
You agree that this license includes the right for us to (i) analyze text and other information you provide and to otherwise provide, promote, and improve the Services, including, for example, for use with and training of our machine learning and artificial intelligence models, whether generative or another type;
...
Such additional uses by us, or other companies, organizations or individuals, is made with no compensation paid to you with respect to the Content that you submit, post, transmit or otherwise make available through the Services as the use of the Services by you is hereby agreed as being sufficient compensation for the Content and grant of rights herein.

Since the average person just clicks "yes" to accept the TOS without reading the 'legal mumbo-jumbo' they are blissfully ignorant of such things. Add in the fact that most people greatly undervalue their time, creativity, intellectual contributions, etc., etc. so they don't really care what bigtechmegacorp does with their data. An attitude somewhat akin to the "nothing to hide" argument commonly used as justification to violate basic privacy rights/protections.

The cost for AI training data is in the multimillion$ range. And most bigtechs are getting that training data for free (literally free) from the users of their "free" (not really free at all) service.

So the question remains; Why are so many people still feeding the bigtechs that are pushing for these huge AI's?

Practically all the applications for AI used by the average person could be self-hosted and not really use much more than the typical homelab network that most IT people run at home. Or onboard in my case where 'home' is a sailboat - not connected to the grid. Even if I increased the hardware on my network say 2-4x to run some demanding AI stuff, it would still likely use less power than my refrigerator and water pump. For LLM's (Large Language Models), STT/TTS (Speech to Text, Text to Speech), and practically every other AI application there are numerous FOSS (free open source software) projects available. The ability to entirely self-host your personal assistant (ala Alexa, Siri) 'offline' is possible now, and has been for quite a few years actually.

The huge AI's that are placing such a demand on electrical grids and cooling systems are really only good for the types of applications which benefit bigtechmegacorps. This is massive data collection/analysis, surveillance state stuff mostly. Basically a furtherance of all the bad behaviors that Meta has become notorious for. The specialized scientific/medical/etc. applications that use very large data models are another matter, but these don't really fall into the bigtech general purpose AI. Or at least shouldn't - such systems should be running at their respective company, university, .gov, etc. in-house computers / data centers. The number of practical (altruistic or benignly beneficial) applications for such huge AI modeling is very small. Of course, its the marketing that such possible applications exist, and just maybe they might be able to [insert hypothetical altruistic cause here] someday, which helps sell the whole idea.

As I pointed out in AI seems to have a personality disorder, the bigtech AI's have some problems. Not the least of which is GIGO; the old saying that should be familiar to anyone with IT 101 knowledge: Garbage In, Garbage Out. Without even considering garbage training model data and going into examples of typical AI outputs (which can be found on most social media sites [see also: Why the legacy social media model of bigtech is destined for the dumpster ] or elsewhere online);
It appears that bigtech AI may epitomize this maxim with inherent flaws at the core.
Many of the novelty applications that the average person uses AI for can be done with any old "dumb" program developed by anyone with intelligence on the upper third of the bell curve, maybe just on the upper half. It's even likely that several so-called AI apps really are just that - just very good apps programmed by a really smart dev.

Even those apps which do use AI/ML concepts, as I said previously, could be run locally.

In any case, it looks like the AI genie is out of the bottle. Whether it leads to 'non-AI' or 'human-intelligence only' 'spaces' being something that people place a value on remains to be seen.

I believe there should be more demand for decentralization of AI rather than the current opposite trend. Besides the benefits I've already mentioned, or at least eluded to, it just might be the answer to preventing SkyNet. 😎

arrcay.com
https://arrcay.7t4.us/.well-known/apgateway/did:key:z6MkvFaQGVpMgvFftLcnymFtRHneLHLgoQcoBE38bWWYqCM5/actor
arrcay.7t4.us
https://arrcay.7t4.us/.well-known/apgateway/did:key:z6MkvFaQGVpMgvFftLcnymFtRHneLHLgoQcoBE38bWWYqCM5/actor
arrcay.7t4.us
https://arrcay.7t4.us/channel/thomas_johnson_iii
7t4
https://www.7t4.us/cover.php
Codeberg
https://codeberg.org/7t4
Github
https://github.com/7t4