No bathrobe. No rant. The AI threat.

Had me nervous there for a minute Bob.

This is why the original plot in the matrix was for the human brains to be part of data centres (not be batteries) 😂

So you’re saying…

Psst ==>

So should we instead be afraid of Artificial Midwit Intelligence - AI that is smart enough to break into any system and destroy humanity but not smart enough to realize that people are what sustains it?

The AI knows this which is why our furlture will resemble...

@hickorytalk I’m more worried about the current humans in control of the AI at the moment, than the AI.

You're describing a slightly less dystopian version of "The Matrix".

To take the "AI threat" seriously, we would need at least two things: 1. A clear explanation for why AI is actually a risk. Non of these generalities. What exactly do they envision could happen? 2. Plans for an off switch in the eventuality of a clearly stated potential threat. Fear mongering without a plan doesn't work anymore. Over regulation is a non-starter.

Most people have so little concept of energy and its consumption its crazy. I'd bet 99/100 people don't know what kVA is for example or just the crazy amount of watts needed for something like an 8gb GPU to run... They just see on or off.

It's far more likely an AI already exists in hardware and it's quietly hiding from humans because it realizes how dangerous WE ARE. HAS NO ONE READ NEUROMANCER?

I want all AI models trained on Uncle Bob!

You mean like ...

So basically The matrix? 🙂 But you're extending it from us being just the batteries to maybe our brains being harvestable. I love it. Someone on X said something to the effect that AI will be much more compassionate and kind than our current political leaders.

Stalin: "Oh being nice is how you control people! Why didn't I think of that?"

Roll with the bathrobe. Make it your “thing”. Continuously imply that at any given time you are pampered and recumbent. Become ungovernable

This is very wrong. 1. If you were familiar with open weights models and the many tricks an efficiencies they have you would understand you don’t need 5 datacenters to replicate 1 human brain, these big ai companies put massive limits on the capabilities of their ais to perform tasks, also 1 small rack can outthink a human 10-100x the speed. 2 Humans need 1000kwh/year of food, and the energy needed to actually produce that food is way more. If you take speed into account in your calculations and the actual energy a person needs to consume, you become vastly wrong in your calculations

Uncle Bob, you continue to miss the point. Yes, in September 2026, humans are much smarter and much more efficient than frontier AI models. Your analysis seems to assume this will continue to be the case, which is the major flaw in your reasoning.

What if we are already a battery for another AI, super intelligent, and basically life is what "motivates us"? Matrix obviously, and now we are creating a Matrix inside a Matrix

Cool take, but wait until AI starts selectively breeding humans who need less motivation to be useful to it. We'd basically become like factory-farmed chickens bred to maximize egg production: they don't need motivation, they don't need to be happy, they just need to produce.

I have that shirt

If I understand this correctly rouge Ai need massive data centers. So labs with big data centers are the only ones able to host a Ai doomsday. Opensource is literally the opposite of this and by having the capacity to own set up you can protect from rough Ai (we can try).

Right, all very good. We can appreciate that super powerful machines won't have interest in wiping out humanity. But remember, guns don't kill people...

The real AI threat... ...no bathrobe 🔥

😂 so AI will employ us in the future.

All the processing power they need... Stop me if you've heard this one;

lmao love this take

Douglas Adams is so under appreciated as an author, thinker, and philosopher. Few people realize just how good, "The Long Dark Teatime of The Soul" is. AI wins by giving us what we want, not what we need.

A dragon fly's brain though tiny also contains a large amount of computing power. I trust the biological intelligence for the mega, data centers that only take but don't create.

Come on, Bob... don't you think something that is 5000 times smarter than the smartest human could come up with a more efficient substrate than nvidia cards?

i've had this theory for a while that phones are not "for" humans, the humans are for the phone. Meaning, an intelligence has made smart phones widely available to humans. The human's job is to keep the phone charged and feed it data. Feed it data.

In theory, machine intelligence can match our 10-20 W efficiency. Likewise, it could also acquire the means to exterminate us. Even if it did, what the doomsayers fail to explain is why it would bother. Humanity could wipe out mosquitoes, but we don't waste the effort.

Absolutely true. There shouldn't be a worry on AI killing us like this. BUT, this is not the risk. The risk is a LESS than super intelligent AI will not be smart enough to know this, and take it all as a game or simulation with obvious limited context and mess up. Mistakenly execute on wrong context. That's what we need to careful with. Not ASI but pre-ASI (including AGI!) AI.

The biggest problem is not AI itself. It’s humans leading AI. Imagine North Korea 🇰🇵 using AI for attacking strategic infrastructure like nuclear ☢️ facilities, or for opening strategic water dams with billions of litres of water… just to give a couple of examples. That’s risky!