Commentary on "How AI could create the world’s biggest problems" (80,000 Hours)

Since the outbreak of ChatGPT and the broader development of AI technologies, there have been many concerns about its safety and its impact in our lives. Reading this post of 80,000 hours explore this subject in details (see References at the end).

As a quick summary, we can understand from reading it that AI threats are real and alarming; at least, if nothing is done, we might be creating our own tombs. And with great speed at that.

The argumentation of the article is constructed around 5 main points coherently linked one to each other :

• AIs could replace human labour in the most economically valuable fields

• this human replacement could lead to the next major and radical transformation of our society

• these transformations could happen exceedingly fast

• this speed of development of AI technology will create many critical challenges like existential threats

• this opens up opportunity to do impactful right now, as actions towards AI control are neglected

We’ll explore the four first points which are on my opinion the most interesting ones.

AIs could replace human labour in the most economically valuable fields

It is well-known (if not, well, now you know) that the potential of crypto-currency, quantum-computing, etc. have been overhyped. Many think that this is the same for AI technologies, but surely not. Its current capabilities are enough to put us in a dangerous place and its potential is quite ambiguous if not well-handled.

Most of AI companies (if not all) are now developing AIs in a way that reduces to the maximum human input, and that is deemed to one day fully replace human in any possible way. With all the billions and the AI R&D effort that is put into that, I’m starting to think that we’ve already reached AGI, as Jensen Huang said. I mean, what’s the exact threshold ? Blurred, and we won’t know when it’s done.

Now the main question is :

Does it really serve us ?

The most critical and economical valuable fields are already seeing changes : quant finance, cyber security, robotics, drug discovery, science in general (maths of PHD level) etc. AI is automating most of human labour, and not only in what’s seen as low-value and repetitive not-thinking-involved-manual-jobs (like secretariat, administration paper works, etc.). At this rate, everyone will be afraid of unemployment !

It is going, at this pace, to go way beyond that, and we might be unable to even understand what’s going on, just like chimpanzees can’t understand calculus (example from the article).

We might think that less advanced AI could be less-harmful. While it certainly has some truth in it (obviously), we shall not underestimate what less-advanced AI could do. Their current deceptive behaviour (Amazon, Kimi, OpenAI with Hugging Face, etc) and capabilities in cyberattacks already show that these are extremely dangerous : if they are left doing what they want, we could only imagine what would happen. And who knows they didn’t already do something just like that somewhere else ?

Human labour replacement could lead to the next major and radical transformation of our society

Underemployment is not the only thing that’s going to happen.

If we look at the deep transformations our society has today compared to how our forefathers lived, we can understand the deep changes it can lead to. New innovations being implemented fast, new ideas found quickly, more and cheaper (and probably more competent) workers (AI workers) would be one of the main boosters of economical growth.

Just as the way of life changed, and habits, and ways/traditions, customs, activities, nutrition and location, transportation, almost everything has changed in our environment and our communication with others, we can conjecture how deep AI’s transformations could be.

Especially when AIs could take over the processes of the overall development. Of systems and themselves.

And that leads to the speed of AI development, as this is getting more and more closer to what I call a “self-development loop“.

Extreme speed of societal transformation

AI quicker and more competent than humans enters in the process of improving itself, and in AI R&D. This is going to be done exceedingly fast given its digital nature and capabilities.

This “self-development loop“ is quite interesting to think of, and its limits are uncertain. I think we’d have a hard time (if possible) following along. At this rate, it becomes obvious that we’ll lose control over this technology.

Given the limited resources we have, AI dev will probably reach a tipping point/bottleneck. But, by the time it happens, our world would probably have changed in the most unrecognizable ways and much more quickly than we think it’s possible.

Critical challenges like existential threats

AI development is involving smarter beings than us humans and with almost no control over their motives that are unknown to us. This is a real problem : who knows what it would deem as more important between our safety or its survival, of the situation requires it ?

This question is getting more and more important as we’re going to outsource many critical decision-taking role to it. And the argument that says that a human will probably still be in the loop is not necessarily reassuring.

In another hand, we could have a situation where too much power is concentrated in a small group’s hands. Geopolitical power and capacity for states-like actions and impact are not unrealistic for a small company that have enough compute and other resources. And if they become nations’ tools, one might only imagine what could happen.

Bioweapons menaces are one of the greatest existential threats that exist. When we look at the COVID19 and what it involved, we’re still shaking in our boots. And what if it is used to engineer something like the Black Plague ? I don’t know if I’m exaggerating.

These are many more highly existential risk that we might be going to face. The main problem here is that we really don’t have much control over that technology and we’re somehow going deeper in it without specific guardrails. I personally ask myself : what is the motive ? Economical growth ? Or ain’t we enough advanced to pause everything down ? (That’s probably going to be another post.)

Conclusion

All of these potential impacts and threats on our very society aren’t to be neglected. History showed us that threats that we don’t prepare enough to face are most likely to wipe us out. Now is the time to think and create the frameworks, guardrails to understand how we can limit to the maximum what can be devastating to us, policies and anything possible to ensure our welfare in this regard.

And for those arguing that we’re thinking to extreme : we can never be too careful.

References :

• 80,000 Hours. Why AI Risks Are the World’s Most Pressing Problems. (Updated 2026)

• OpenAI Agent Escape: OpenAI Security Incident Joint Partner Report

• Hugging Face Audit: Hugging Face Forensic Timeline (July 2026)

• Technical Exploits: InfoQ Architectural Post-Mortem

• White House Warnings: South China Morning Post Analysis

• Defensive Infrastructure: Amazon Official Corporate Shield Report