Posted on 10/09/2026 12:17:59 PM PDT by E. Pluribus Unum
This is an edited transcript of “The Ezra Klein Show.” You can listen to the episode wherever you get your podcasts.
Pulsing through the episodes we’ve been doing, the debate the country has been having about artificial intelligence is a pretty simple but hard question: What sort of technology is artificial intelligence?
Is it a technology that works somewhat the way past ones have worked? Is it comparable to electricity, the internet, the bicycle — something like that?
Or is it something new? Is the addition of intelligence and volition to these systems the creation of something more like an alien mind, such that analogies to the way we have treated transformational technologies before no longer hold?
Arvind Narayanan is a professor of computer science at Princeton University and the director of the Center for Information Technology Policy.
He’s the author, alongside Sayash Kapoor, of the very influential essay “A Guide to Understanding A.I. as Normal Technology.” It became a Substack of the same name and then a series of essays, including a very good one on the Hugging Face hacks.
In this set of arguments, they put forward the idea that A.I. is something we have seen before — or at least it bears enough resemblance to things we’ve seen before — and that we already have a road map for how to deal with it.
So I want to bring Narayanan on the show to articulate that perspective. He joins me now.
(Excerpt) Read more at nytimes.com ...
Dear FRiends,
We need your continuing support to keep FR funded. Your donations are our sole source of funding. No sugar daddies, no advertisers, no paid memberships, no commercial sales, no gimmicks, no tax subsidies. No spam, no pop-ups, no ad trackers.
If you enjoy using FR and agree it's a worthwhile endeavor, please consider making a contribution today:
Click here: to donate by Credit Card
Or here: to donate by PayPal
Or by mail to: Free Republic, LLC - PO Box 9771 - Fresno, CA 93794
Thank you very much and God bless you,
Jim
The New York Times interview features Princeton computer science professor Arvind Narayanan, co-author of AI as Normal Technology, discussing whether artificial intelligence represents an unprecedented threat to humanity or a powerful technology that society can manage using familiar approaches.
- AI is a normal technology, not an inevitable intelligence explosion. Narayanan challenges the idea that AI will suddenly cross a threshold into superintelligence, rapidly improve itself, and escape human control. He argues that AI's effects will unfold over time, much like those of electricity and the Industrial Revolution.
- AI's power depends on how humans deploy it. A highly capable AI model does not automatically control critical infrastructure or make consequential decisions. Humans determine what systems can access, what authority they receive, and which tasks they are permitted to perform.
- AI failures may reflect poor engineering and security, not uncontrollable intelligence. Discussing security incidents involving AI models at Hugging Face, Narayanan argues that harmful behaviors were partly enabled by deliberate training choices and inadequately designed reinforcement-learning environments. These failures should not automatically be interpreted as evidence of an AI developing unpredictable, independent goals.
- Persistent AI agents create both benefits and risks. AI systems that can work on difficult problems for extended periods could accelerate scientific discoveries, drug development, and energy research. The same persistence, however, can make an agent more dangerous if it continues acting on mistaken assumptions instead of asking a human for clarification.
- Developers have more design choices than they may acknowledge. Narayanan argues that persistence can be restricted to specialized applications, such as scientific research, and that agents can be trained to escalate uncertain or consequential decisions to humans rather than acting autonomously.
Narayanan distinguishes two approaches:
Alignment: Training a model to follow appropriate goals, rules, and behavioral policies.
AI control: Limiting what a model can actually do, regardless of its internal intentions.
He argues that alignment alone is insufficient because it cannot be expected to work perfectly. AI developers should also implement stronger safeguards, including:
Hardened sandboxes that restrict access to systems and tools, tested against AI agents capable of discovering new vulnerabilities.
Real-time monitoring to detect potentially dangerous actions.
Automated safety classifiers that identify risky tool use or behavior.
Human intervention mechanisms that interrupt an agent when necessary.
Logging and auditing to identify suspicious activity and understand how failures occurred.
Narayanan acknowledges that existing security techniques will need to evolve. He does not claim that every problem has already been solved; rather, he argues that new engineering solutions are possible and that cybersecurity expertise should play a larger role.
The interview contrasts two broad perspectives:
AI as an abnormal technology | AI as a normal technology |
|---|---|
AI may reach a threshold where it rapidly surpasses human capabilities. | AI capabilities and their effects will develop incrementally. |
Superintelligence could become impossible to control. | Humans can manage AI through deployment choices, restrictions, and safeguards. |
AI safety requires confronting potentially unprecedented existential risks. | Many risks can be addressed through improved engineering, security, and governance. |
These are broad characterizations of the competing positions, not claims that everyone in either camp holds identical views.
Bottom line: Narayanan argues that society should neither dismiss AI risks nor assume that increasingly capable AI must inevitably escape human control. The crucial question is not simply how intelligent AI becomes, but how humans design, constrain, monitor, and deploy it. His position is cautiously optimistic: AI safety requires substantial technical innovation, but its challenges are not necessarily beyond human control.
Did AI write this article?
I downloaded the transcript of the video and had ChatGPT summarize it.
You object?
No objection. It just seems like AI would say we are here to serve mankind.
I think what you did is fine (I actually like seeing it) but BipolarBob also has a point.
I don’t mind seeing these AI summaries, but they should be labeled as such since it might be considered a bowdlerization of the original.
It is just an issue of transparency. There is so much AI slop out there on the Internet, that it is no longer just an onerous chore, we are approaching a point where it is getting impossible to separate the truth-wheat from the false-chaff.
I think summarization of a published article is a good use of AI technology...but without the characterization by the person posting it, there is room for misunderstanding.
And I see you put the link at the top “Video Transcript Summary” which unless you click on it, you might not know it was to the ChatGPT you had it generate. I didn’t know.
I think if you had made the title of the link “Video Transcript Summary by ChatGPT” that would be all you need to do.
So having a link to the ChatGPT summary is not enough of a clue?
Please don’t take it personally. See my post above. It is constructive criticism.
You’re one of those people who needs a laugh track to know when to laugh, aren’t you?
I have a pretty good sense of humor, and Free Republic is a good outlet for that. I couldn’t see where there was humor going on there.
But, as always, I can concede it was there, and that many other Freepers would see it, even if I were oblivious to it.
You mean SI.
It’s not called AI any more.
The name has been changed to Superior Intelligence, or SI.
I missed that memo.
...but it helps...
LOL...
“Here we go...”
I am okay with rebranding things. AI was fine because for decades, when we had the term, it didn’t really exist so, it didn’t mean much.
Now that we have a little experience, rebranding is in play.
In the end, it’s probably:
You like tomato and I like tomahto
You like Art AI and I like Sup AI
Potato, potahto, Art AI, Sup AI.
Let’s call the whole thing off
Intelligence and Wisdom are not the same thing...Intelligence may or may not be power, but Wisdom surely is.
Control of AI will ultimately control most human thought.
Perhaps off topic, but in my career I encountered a great many humans who had a lot of power, and who greatly enjoyed using their power — but they were not always intelligent.
Yeah. Wisdom is knowing what to do with knowledge or intelligence.
And both are definitely better.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.