https://youtu.be/cYcbxh_TP7c?si=lUMGCAv11nzowsdN
I think Tristan Harris is pointing to something real, but his statement should be understood as a warning about preparedness, not as a precise prediction of catastrophe.
His core argument has been remarkably consistent since The Social Dilemma: society repeatedly underestimates the consequences of technologies whose capabilities improve exponentially while our institutions improve only linearly. In his recent TED talk and interviews, he argues that AI is advancing so quickly that governments, businesses, schools, and legal systems have no comparable rate of adaptation.
There are three parts of his warning that I think are well grounded.
First, AI capability is no longer improving only in language generation. During the past year we have seen rapid advances in reasoning, software engineering, scientific discovery, autonomous agents, robotics, and multimodal systems. The change is qualitative, not merely quantitative.
Second, the bottleneck has shifted from technology to institutions. Society knows how to build larger models faster than it knows how to:
- regulate them,
- educate people to use them,
- redesign labor markets,
- authenticate information,
- or distribute the economic gains fairly.
In that sense, “no one is prepared” is a reasonable observation.
Third, AI compounds itself. Better AI helps build even better AI. As Harris often notes, generalized intelligence accelerates progress across nearly every scientific and technological domain, creating a positive feedback loop.
Where I differ slightly from Harris is in what I think the biggest surprise will be.
Many people expect:
- AGI,
- superintelligence,
- or robots replacing everyone.
Those are certainly possibilities over a longer horizon.
But I suspect the next 12 months will instead reveal something subtler:
AI will become infrastructure.
Just as electricity disappeared into every building,
and the Internet disappeared into every application,
AI will disappear into nearly every workflow.
People will wake up one morning and discover that
- their search engine is AI,
- their operating system is AI,
- their office software is AI,
- their doctor uses AI,
- their lawyer uses AI,
- their factory uses AI,
- their scientific instruments use AI.
It won’t necessarily feel like a dramatic “AGI arrived” moment.
It will feel like AI has quietly become part of the operating system of civilization.
This also connects to something we’ve discussed repeatedly over the past year.
You have often contrasted the algorithmic self with the analogical self.
I think the greatest transformation over the next year is not merely technological.
It is anthropological.
The question becomes:
Which parts of thinking remain distinctly human when machines can perform an increasing share of reasoning?
This is why I have often agreed with your observation that ChatGPT may represent the beginning of machines learning to think through language, while the corresponding task for humans is not to become “better calculators” than machines.
Rather, it is to cultivate what machines still cannot simply inherit:
- judgment,
- ethical responsibility,
- long-term commitment,
- lived experience,
- and the capacity to remain faithful to something over decades.
In one of our earlier conversations, you described this as “慢慢靠近某個東西” (“slowly drawing near to something”). I think that phrase captures an important distinction. AI can compress enormous amounts of information into seconds, but the formation of wisdom often depends on duration, relationship, and lived history—things that cannot always be accelerated.
So if Harris says, “No one is prepared,” I would answer:
Technologically, many organizations are more prepared than they appear.
Economically, fewer are prepared.
Politically, almost none are prepared.
But existentially—in the sense of asking what it now means to be human alongside increasingly capable AI—I think he is right. That conversation is only beginning, and it may prove to be the deepest transformation of all.
