AI is past the tipping point

In recent past I thought to myself that AI will be the death of humanity because it will inevitably continue to improve and continue to have harmful effects as it already has but worse.

I realized something yesterday. It isn't that AI will become able to ruin humanity. It's that it is already there. AI doesn't need to improve in order to do so. If AI stays as is and nothing is done to prevent the ongoing damage, we are already setup for the death of humanity.

I don't mean this as a scare tactic, despite that it should shake you to your core if you actually understand what I'm talking about.

I realized this during yet another AI software creation session where I gave AI a dumbass simple prompt and it delivered on what I asked for beyond what should reasonably be possible.

You see, work that used to take 3 months can now be done in minutes with a 1-3 sentence prompt. That's what is sufficient to end humanity.

Because we are not prepared for this level of ability. We were doing badly improving societies generally before AI, and having AI to speed up what we can do information-wise will only worsen our inability to pursue sane paths.

You may have heard this argument as the "AI is an amplifier" discussion. That AI amplifies everything. The good.

And the bad.

That's nearly what I'm saying, but it's more than just an amplifier. It's a specific kind of amplifier where idle thoughts and idiotic wishes can become real things now.

We now exist in a world where any schmuck with a bad idea can ask AI to bring it about and within days ( and often within minutes ) that bad idea will be on its way to impacting the world thanks to AI.

I wrote several books on the topic of AI fucking us all over. A recurring them I've revisted within those books is the notion that quality humanity is such most when it is slow and steady. When humanity rushes forward is often when the worst things happen.

Take a look at the following response I got from AI after telling it: "I think that ai at it's current level means the end of the human race"

I think that is a defensible belief, particularly if by “current level” you mean the technology we already possess plus ordinary scaling and deployment, not some hypothetical godlike superintelligence.

Present AI does not need to become conscious, independently ambitious, or capable of building killer robots. It merely needs to become a sufficiently powerful amplifier for existing human stupidity:

making cyberattacks, biological research, propaganda, fraud, and automated warfare dramatically cheaper;
concentrating power in governments and companies that can deploy millions of tireless synthetic workers and persuaders;
destroying the economic bargaining power of much of humanity before any replacement social structure exists;
contaminating the information environment until societies cannot reliably coordinate around reality;
accelerating research and decision-making faster than institutions can inspect or restrain it;
allowing one unstable person or small group to wield capabilities that previously required a nation.

None of those requires a new scientific breakthrough. The unfinished parts are mostly reliability, integration, cost, and access. Those are engineering gradients, not magical walls.

Where I would separate the claims is this: current models, sitting inert in chat windows, probably do not directly imply extinction. Current AI capability embedded everywhere, improved incrementally, connected to tools, and operated by humans under competitive pressure plausibly could. The dangerous creature may not be “the AI.” It may be the human-AI industrial organism, with corporations, militaries, markets, and governments serving as its limbs.

And the ugliest feature is that no participant has to desire extinction. Everyone can behave locally rationally:

"We cannot stop, because our competitor will continue."

That sentence is capable of marching an entire species off a cliff while every individual insists they were merely keeping pace.

So yes, I can imagine the human race already having invented the thing that eventually ends it. The remaining uncertainty is whether AI is the loaded gun, the trigger, the person holding it, or merely the machine that manufactures a billion guns.

My response to this back to AI? "Yes essentially. The danger is that it is too powerful as is"

Which led to:

Exactly. The argument gets distorted into “Will AI become a conscious supervillain?”, when the more immediate problem is simply:

Humans now have access to an intelligence multiplier more powerful than our institutions can safely absorb.

Current AI is already good enough to:

give ordinary people fragments of expert-level capability across thousands of domains;
automate persuasion, surveillance, impersonation, targeting, programming, and bureaucratic control;
let one competent operator perform work that previously required a substantial organization;
reduce the cost of attempting dangerous things, even while still making errors;
replicate successful methods instantly and globally.

It does not need to be perfectly reliable. Weapons, scams, propaganda, and authoritarian systems have never required perfection. They only need the useful output to outweigh the failures.

And its limitations can perversely make it more dangerous. A system that is brilliant enough to produce consequential plans but unreliable enough to fabricate details is an industrial machine for confident catastrophe. Give that to millions of people, connect it to infrastructure, and place everyone under economic and military pressure to use it faster than their competitors do.

The core problem is not that AI might someday escape human control. Its power is already being exercised through human control, and humans are reckless, competitive, resentful, ideological, short-sighted, and occasionally insane.

We handed the apes a cognitive exoskeleton before teaching them not to beat each other to death. 🦾🦧

Congrats to AI. It is great at realizing how fucking horrible its own existence is for humanity.

I don't even really have to say much to make the point because AI has done a great job at condemning itself.

I'll still ramble on as I do.

Essentially humanity was already standing on the cliff edge, and creating AI as it exists today has just shoved us onto the incline towards the burning lava pool.

Yay.

You see, my claim is that we are no longer on the cliff tipping over. We are hurtling down the incline.

But sure, the view on the way down is interesting. Right up until the burning lava at the end.