AI threat is a hoax

[~MG~]

Ohh there's a problem - but it's NOT what we're being led to believe.

At the core AI is only a gun. Both follow directions

Gun: you pull trigger, it sends bullet.
AI: you ask question, it finds answer.

Both need human management.

Just like you don't leave a loaded gun Laying around a daycare; you don't leave AI in an uncontained environment.

In ALL of the recent incidents of AI supposedly going rogue on its own - it's been found that the containment had open doors. Bottom line? A human, either intentionally or not, let AI out to play.

And THAT is where the danger lies. It doesn't make any difference if AI's capabilities are as a pea-shooter or a machine gun. It's a human that wields both.

******

So why are we being lied to?

Control, and they're using using fear to accomplish it. Just like was done with firearms.

Only this time it's the big players that want to control the AI industry - and the fear created is intended to keep small players out through restrictive laws and regulations.

Rinse, repeat.

[Texas PO]

As I have said before, AI is not the problem, it's the people who control it we should worry about. Like the gun, AI can be good, or bad, depending on whose hand it's in. People with nefarious intent no doubt would love to gain control of AI.

[Sasquatch]

The new "threat" is always about control.

The Bermuda Triangle
Cults stealing your kids
Halloween candy with razors in it
Covid
Aliens
Ai

And the media is always a willing participant.

[~MG~]

And the media is always a willing participant.

Yep. Along with control, there's big money in spreading fear.
It used to be large-font newspaper headlines to sell subscriptions.
Now it's apocalyptic clickbait on 24/7 media to generate income.

[~MG~]

Linked below is a Breitbart article that continues to spread the fear:

"From what we know, Hugging Face reported over 17,000 agents attacking their infrastructure that went on for days and weeks,..."

In all likelihood, it was one AI Agent hitting it 17,000 times. (Dos attack, anyone?)

But which sounds scarier? Which has more weight for the public to demand their lawmakers do something! Quickly! To prevent Skynet!

If you read the article - take note how many times AI is framed as having intent... you know. Like those big black evil AR's that try to escape the safe, so they can find somebody to shoot.

https://www.breitbart.com/tech/2026/09/28/nvidia-unveils-safety-platform-to-stop-ai-agents-from-breaking-containment/

[1skrewsloose]

Well ya know what, I could go dark in a heartbeat, pay my bills by check, smash the crap out of my phone and laptop. I wouldn't be happy about all of it, I think they're abuseing info collection on private citizens now.. I'm sure I get to the point of saying screw it all pretty soon. I absolutely hate having someone looking over my shoulder, having to know what I'm doing every second.

I turn off data on my phone, no interenet, I just want my phone to be a phone and that's all.

[Inor]

This is all a big nothing!

The two companies screaming the loudest are Anthropic (Claud) and OpenAI (the makers of ChatGPT). Both of them are in a big dick-measuring contest right now claiming they have "the most powerful model". Both companies are also losing MASSIVE amounts of money. Yes, Anthropic shows a profit on their Income Statement but that is only because they are NOT reporting huge amounts of depreciation on the computer hardware they bought the last couple years. OpenAI make no bones about it, they are spending more money every quarter than they did the previous quarter and do not even plan to have a positive quarter until 2030! Combined, they are both also on the hook for over $1.4 TRILLION dollars of expenses coming up in the next 3 years!

Moral of the story, both companies are in horrible shape financially. Neither one has any other source of value beyond their AI models. (They are not Google, Tweeter or Meta). So they HAVE to fight it out over who has the best model. What better way to ape your AI model than to feed the MSM stories about it being so "smart" they can no longer control it?!?

Just from what I am seeing from my clients in that industry, folks are starting to turn away from the big expensive cloud based models with their big expensive monthly fees. Instead, they are starting to favor open-source models (many from China) that they can run locally and be tightly controlled The ones I have been working with lately are nearly as good as OpenAI or Anthropic and will run nicely on a $5,000-$10,000 system.

Just from looking at the financials of OpenAI and Anthropic, assuming they are not cooking the books, Anthropic will probably survive but get the hell beat out of them. OpenAI is a dead man walking. I expect OpenAI will be gobbled up entirely by somebody much bigger (maybe Microsoft or Oracle)?

[~MG~]

This is all a big nothing!

What are your thoughts re the push for AI regulations?

IF they could even be developed with a required implementation - could that be a hindrance to the small models?

Flailing comes to mind in several corners; political mostly. FWIW, the government regulating AI concerns me more than the AI itself.

Nice post, btw! It shined a light on a bunch of it.

[Inor]

What are your thoughts re the push for AI regulations?

IF they could even be developed with a required implementation - could that be a hindrance to the small models?

Flailing comes to mind in several corners; political mostly. FWIW, the government regulating AI concerns me more than the AI itself.

Nice post, btw! It shined a light on a bunch of it.

I am John Galt. I do not trust the government to regulate ANYTHING! Furthermore, I have been playing with this data science crap for over 15 years and am still figuring it out daily; I highly doubt there is anybody in the government that is capable of learning enough of it to even begin to write any kind of regulations, let alone implement them.

My biggest fear of any kind of regulation (government or otherwise) will do nothing to control the current big and powerful who can bribe their way through. It will only hinder folks like us and prevent the next small inventor of the next truly clever advance.

Europe is learning that now. They put some fairly strict rules on their AI developers about 2 years ago. Paris France used to be one of the great centers of AI development. It is barely even mentioned anymore. California is following the same path currently and look at how many of their big tech firms have left the state recently. Government is NEVER the answer!

[MoreAmmoOK]

30 years in IT made it easy to smell the bull shit. A machines access is far to easy to control for it to "accidently escape". It would be exceptionally stupid to trust electronic firewalls to contain a machine that has been trained to hack. If the network wire of an AI machine isn't connected to a destination its trainers don't want it to go it's not like the machine is going to hop up and move the wire so it can access more systems.

[~MG~]

30 years in IT made it easy to smell the bull shit. A machines access is far to easy to control for it to "accidently escape". It would be exceptionally stupid to trust electronic firewalls to contain a machine that has been trained to hack. If the network wire of an AI machine isn't connected to a destination its trainers don't want it to go it's not like the machine is going to hop up and move the wire so it can access more systems.

Hey? Would you please elaborate/define what you mean by "network wire"?

Surely not a hardwire, because Agents go on the internet - and there's no "hardwire" to the internet... It was your last sentence that sent my bubble brain into the curiosity, lol.

Could there not be digital coding that instructs an Agent to not bust through a firewall? Seems if it can be trained to hack, it could also be trained to not hack... Of course , that would kinda defeat the evil-doer's purpose, huh.

Or like I said in the OP~
It ain't the gun, it's the shooter.
It ain't AI, it's the human.
Thanks ~

[T-Man 1066]

Put your tin foil hats on, T-Man 1066 fixin to get wierd, and I'm not even drinking beer at 7:30 in the morning!

Here is how AI could be used in a very very bad way.

Bad operators buy / steal / build a bunch of wheeled electric robots, like the ones Amazon uses to drop off packages. They got cameras on them. They are reprogrammed to use AI and facial recognition to identify people. Instantly scrub the internet. Social media, voting history, gun owner, anti-covid, conservative, etc. Then it could pneumatically shoot a small dart like object with some type of bio-poison. People assume a twinge or bee-sting. 5-7 days later, flop, they are dead. Mean while killer hijacked robot is still dropping off amazon packages, or whatever.

Now that is a crazy thought-up scenario, but I think that is how AI could effectively single out and wipe out alot of people the bad actors don't want around. Remember all the chatter about over-population?

Ok, tin foil hat mode off.

[~MG~]

Put your tin foil hats on, T-Man 1066 fixin to get wierd, and I'm not even drinking beer at 7:30 in the morning!

Here is how AI could be used in a very very bad way.

Bad operators buy / steal / build a bunch of wheeled electric robots, like the ones Amazon uses to drop off packages. They got cameras on them. They are reprogrammed to use AI and facial recognition to identify people. Instantly scrub the internet. Social media, voting history, gun owner, anti-covid, conservative, etc. Then it could pneumatically shoot a small dart like object with some type of bio-poison. People assume a twinge or bee-sting. 5-7 days later, flop, they are dead. Mean while killer hijacked robot is still dropping off amazon packages, or whatever.

Now that is a crazy thought-up scenario, but I think that is how AI could effectively single out and wipe out alot of people the bad actors don't want around. Remember all the chatter about over-population?

Ok, tin foil hat mode off.

Nice nod to Dune with the poison dart!

Your idea might work well in the cities where people are too lazy to go get their own shopping. Out here we just shoot anything that rolls unattended, heh.

There are much easier AI-ways to kill off people once they're identified... Especially those who also use the medical system, or drive a newer vehicle over fifty mph, etc.

[T-Man 1066]

Never seen Dune. I probably haven't watched 3 movies in the last 30 years.

[T-Man 1066]

Nice nod to Dune with the poison dart!

Your idea might work well in the cities where people are too lazy to go get their own shopping. Out here we just shoot anything that rolls unattended, heh.

There are much easier AI-ways to kill off people once they're identified... Especially those who also use the medical system, or drive a newer vehicle over fifty mph, etc.

100% agreed. AI could kill people in thousands of ways.

Just seen my cousin's husband over the weekend. He has a defibrillator that decided his heart was getting wonky and shocked him about 3 times. It went on WIFI to alert the defibrillator manufacturer, they proceeded to call his Dr and let them know, Doctor's office called him to see if he was ok or if he needed help. Now could rogue AI use his wifi to have the defibrillator start causing his heat to go south? I don't see why not...

[~MG~]

Now could rogue AI use his wifi to have the defibrillator start causing his heat to go south? I don't see why not...

You bet it can. The same way it can 'disable' the brakes and steering on your truck while you're going down the interstate. Easy peasy for AI to merge vehicle registrations with licensed gun owners.

People have no idea the level of control they've given to others through their own choices. And now those 'others' have AI to complete their agenda.