Thoughts on AI dangers and problems

 Users of Artificial Intelligence might be happy enough to just pay the fee, figure out the basics, and utilize its abilities to make their lives or work easier. But there are things in the background that should be understood by all citizens, not just AI users. 

 

RELIABILITY – The Large Language Model that AI is built on is what gets you the result you want from ChatGTP and others. It is a flawed system. “OpenAI, the creator of ChatGPT, acknowledged in its own research that large language models will always produce hallucinations due to fundamental mathematical constraints that cannot be solved through better engineering, marking a significant admission from one of the AI industry’s leading companies.” Hallucinations are incorrect answers that AI produces in unexpected situations. 

Sam Altman of Open AI said “I no longer think we're going to be able to just patch all the bugs. I think we're going to have to completely switch the defense model to using the same technology for a new kind of defense shield." 

TRAINING BIAS - Who trains the AI? What language is it trained in? What human philosophy is it grounded in? Is a certain religion or culture dominant in the training? Are the current company owners steering AI to serve mankind, or just reel in the greatest profits in the long run? All this is unclear, being part of the Black Box that is AI. The Tech Bros that are pushing and running these companies are each a bit strange and possibly have their own unspoken agendas. Should they be running this?

XAI Musk- “I’ve got a little bit of cyber psychosis,” Musk said. “Maybe a lot actually.” (AI psychosis is where you are influenced by AI into delusions.)

Open AI Altman- “We believe that the world should accept some bad things happening for the benefits of this technology and people having the agency.” 

Peter Thiel- “I no longer believe that democracy and freedom are compatible.” (Nerd Reich, by Gil Duran, p. 8)

Anthropic’s Amodei- "I think we're building a growing and singular capability that has singular national security implications. And democracies need to get there first." 

“Given the accelerating rate of AI capability development, it’s my worry that in 6–12 months such a swarm could be capable of taking over the entire internet with a persistent botnet (potentially causing hundreds of billions of dollars in damage), and that the scale of damage would continue to increase from there if AI becomes more powerful without the necessary guardrails.”

“I am concerned that AI may be uniquely suited to autocracy, and to deepening the repression that we see in autocracy.” [Wall Street Journal video]

LIMITS OF DATA - what data was AI trained on, and what doesn’t it know?
Has AI stolen someone else’s intellectual property? “AI companies create their products by harvesting every bit of accessible content and data in the world, strip mining the creativity, knowledge, and intellectual property of billions of people to feed their networks without permission, payment, or acknowledgement…” (Nerd Reich, p. 174)

The world is accumulating data at an incredible rate, and AI has to somehow keep up. What will the cost of this constant input and training be?

Is all data good data? Can AI just be fed random books and research? How is the good sorted from the bad?

CERTAINTY CLAIMS - AI usually claims certainty rather than state a reliability gauge.

SECURITY – Can AI itself be hacked?

Yes, and here’s an example -

Here’s a paper explaining many security issues within AI.

NOT REGULATED - “No federal legislation establishing broad regulatory authorities for the development or use of AI or prohibitions on AI has been enacted. “ 

POTENTIAL SCAM - The whole thing might be a financial scam. Why the huge rush to grow and build at the same time companies warn that AI could be catastrophic for the world? Could this be a repeat of the dot-com bubble in 2000? There is every reason to believe that the Tech Bros are going way too fast, including overbuilding. “Gartner claims that AI data center companies are unlikely to generate the required $8.5 trillion in revenue by 2030 to justify their aggressive AI investments after accounting for financing costs and a 12% ROIC. “

ENSHITIFICATION – AI slop. Everything comes from one place and becomes gray, lifeless, and most likely with errors. When you see AI slop you know what was behind it was a lazy human trying to cut corners on their job. The slop is getting better, but it is still slop and the product of laziness.

PRIVACY - Is AI accumulating a dossier on you? 

OWNERS VS LABORERS -

The Tech Bros promise that life will be so much better with AI, at least when they’re not warning us we could all die from it. They claim AI will do a lot of work for us so we’ll have more income and more leisure time. But let’s look back through history and see if that has worked out in the past. Let’s think about trains, electricity, the telegraph, cars, etc. Have any of these increased the average wage of laborers? They each helped society, yes. But none of them raised workers’ wages. So who got the benefit from making things more efficient and faster? The owners. When the trains, for example, reached across the country, it was the rail companies, the steel companies, the property owners of the lands along the rail bed that gained. Not the workers. And so it will be with AI. The Tech Bros will get the windfall, not us.

“All I wish to make clear is that, without any increase in population, the progress of invention constantly tends to give a larger proportion of the produce to the owners of land, and a smaller and smaller proportion to labor and capital.”
― Henry George, Progress and Poverty