On the AI revolution's uncertainty, "Imago Dei" and more...

On the heels of the OpenAI/Hugging Face incident, Dwarkesh Patel of Dwarkesh Podcast published “The Rise and Fall of Agent Civilizations” on 8-30-26, a “plain English” account of research findings – one published by OpenAI and the other by METR – that outlined how autonomous AI agents coordinated a cyberattack on the Hugging Face infrastructure.

It took the internet by storm, with wide-ranging reactions from doomsday-esque remarks to critiques on the diction Patel adopted, such as “civilization” and “conspiracy.” 

The latter’s point of view is simple: the anthropomorphism of AI (i.e., the attribution of human traits to non-human entities) is stretched and, to Ben Thompson’s point in his 9-8-26 Stratechery note, “there is no evidence, theoretical or anecdotal, of agents having any sort of internal volition or intrinsic motivation or sense of morality.” 

And just when the frenzy quiets down slightly, OpenAI releases GPT-6 Astra, to which Nvidia CEO Jensen Huang reacts by declaring that “AGI has arrived” – all within a week from Patel’s summary. 

But there’s more. A 9-8-26 WSJ exclusive reported that Anthropic researcher Jacob Coxon decided to depart from Anthropic due to fears of rogue AI models. He believes “no company can responsibly develop AI that can outperform humans in a range of tasks,” i.e., no company can responsibly develop AGI, “absent government intervention or a coordinated industry slowdown,” according to the WSJ. 

Coxon further shared on X that “the people building AI earnestly believe that it could kill us all by the end of the decade” and urged lab researchers to rethink their influence on superintelligence in the coming years. 

Where are we now? 

I am by no means an AI lab researcher and am not equipped with any technical analysis nor intel on what’s going on behind the scenes among frontier AI labs.

But I have many questions from an outsider’s perspective.

I wonder where to draw the line. When can we officially declare that AGI is achieved? When is AI officially “human” enough so the anthropomorphism of AI is no longer exaggerated? At what point do governments, frontier labs, and institutions say “enough is enough” and agree to terminate all AI developments overnight, and hence prevent humanity from fully realizing the full potential of AI – the promises of breakthroughs in healthcare, innovation, and the creation of jobs that are beyond our imagination. Or is that trade off – risking interfering with the full potential of AI – too great and too insurmountable, so we should just YOLO it? 

How much of limiting government intervention is actually beneficial in helping us reach the full potential of AI? How much is it motivated by winning the U.S.-China AI race? Or are we driven by capitalistic gains because we have a 10x levered investment in frontier AI labs or capital deployed in AI beneficiaries such as power generation? 

What if limited government intervention is too limited? What if we wait till the last-second for intervention and end up in a scenario that Coxon outlined, and humanity is ultimately eliminated by AI by the end of the decade?

I probably exaggerated, but those are some of the questions on the top of my head. 

And perhaps my questions, which are nowhere near exhaustive, are a clear indication that the only thing certain about the AI revolution is the uncertainty that comes with it. 

“One of the most important things to know about this AI cycle is that no one knows what they’re talking [about],” Jack Altman wrote on X. 

My thoughts on AI 

If Jack Altman is right, that no AI expert truly knows what they’re doing, then I think we should lean in to the uncertainty, acknowledge them, and showcase the humility that I think is necessary for people to collaborate on frameworks to think through how we can coexist in what the Shanghai Macro Strategist defined on X as a “Three-Body world,” divided by the Adventists, the Redemptionists, and the Survivors.

From the Shanghai Macro Strategist, on X: 

“The Adventists want the Trisolarans to come. They see the arrival of a superior intelligence not as an existential threat, but as a potential replacement for a civilization they believe has reached its limits.

The Redemptionists also believe the Trisolarans can save us—but they hope that the two civilizations can coexist, with superior intelligence serving as a means of transcending the flaws of human civilization rather than replacing it.

The Survivors, by contrast, ask a much more primitive question: How do we survive the Trisolarans? For them, the preservation of human agency and human civilization takes precedence over the promise of whatever the future intelligence might bring…

Perhaps the future of AI will be determined not simply by how intelligent machines become, but by the relative dynamics and power of the Adventists, the Redemptionists, and the Survivors among us.”

There won’t be one single truth of what constitutes AGI, what the full potential of AI looks like, and what safety measures exactly entails. 

And I think that’s fine. There’s no right or wrong, like many things in life. 

For now, the best thing we perhaps can do is to pick one of Shanghai Macro Strategist’s bodies defined above and follow it with a balanced, questioning attitude. 

Note: Obviously there can be a whole other debate on why the three bodies outlined by the Shanghai Macro Strategist shouldn’t be the source of truth. I’m aware of it. 

Why I’m referencing it, in short: I think it does effectively capture one framework to think through the series of events I outlined in the intro. At least it makes logical sense to me, for now. It shouldn’t be treated as the truth, though.

I for example fall under the Redemptionists. I believe humanity and AI can co-exist, with AI having the potential to advance research and innovation – healthcare being one area I’m particularly excited about. 

But that doesn’t necessarily mean replacement. There's a distinction I keep trying to reach for here – between intelligence and personhood and between simulation and soul – and I'll be honest that I can't yet articulate.

I tried it through the lens of Imago Dei, the Catholic idea that humans are unique because they bear the image of God – inherent value, rational powers, moral responsibility. But it didn’t seem fitting, and I felt like I was trying to force an argument that directly opposes Dwarkesh’s claims. 

Plus, I recognize this level of thinking isn’t comprehensive. It is just one (attempted) theological framework in a pool of billions (theological or not) I’m using to formulate one possible line of thinking. 

We also haven’t touched on the economics of AI at all, for example, as one could argue that job disruption caused by AI is already physically replacing humans (i.e., layoffs), just not yet all. But some forms of it are enough to justify replacement. Perhaps more on this in a future piece (I need to do research! I don’t even know if the hypothetical argument I outlined aligns with what economists believe.).

But the point is I picked a side, the Redemptionists, while being open to actively engaging with different views and not immediately rejecting them just because they oppose my own (e.g., actively reading Coxon’s comments on the existential threat of AI, while engaging with rebuttals/ on Coxon’s remarks). 

All that to say, we live in a very exciting time. This is a great opportunity to engage in dialogue, to challenge our individual views, be comfortable co-existing with different views, and acknowledge and lean into the uncertainty created by AI. 

I’m optimistic about the future of AI and believe you should be too. But of course, I say this with a healthy dose of skepticism…