Skip to main content
Ramez Naam

Ramez Naam

American fiction writer

Nov 11, 2026

SophiaHOST
1:37
So what do you think?
1:39
Well, first, like I got to respect anybody that puts numbers down and the AI futures folks have consistently like made a model, made a forecast and written it down.
1:51
Second, I'd say people are freaking out and we are human beings, we get emotional.
1:57
Third, I'd say it looks to me like we are seeing acceleration of AI progress.
2:04
It is not clear to me that that acceleration is coming from, how much of that acceleration is coming from AI itself versus more investment of resources.
2:16
We have to bear in mind that the landscape of this, the number one driver of AI improvement is that we are putting exponentially more resources into AI, exponentially more venture capital, exponentially more hardware capex, exponentially more data, exponentially more talent.
2:35
I tweeted yesterday a chart showing that in overall R&D spend, Anthropic has a clear logarithmic scale, meaning anti-exponential, diminishing returns.
SophiaHOST
5:56
And so if Anthropic can continue exponentially growing its revenue, then it can continue exponentially growing its capabilities.
41:14
And so like
41:14
Ramp's data shows Astra came out from nowhere, from ChatGPT, from OpenAI, and has blown past Fable, the best model from Anthropic.
41:27
Switching costs are nearly zero today.
41:31
And...
41:34
This is also interesting.
41:36
A lot of tech-dominant network effects is you control the protocol, you control the way that two nodes talk.
41:43
Facebook, by default, controls the protocol because the way that you and I interact on Facebook is through their site.

7 MINS LATER

49:00
Basically, this just basically means AI designs itself, and then all of a sudden, it just hits escape velocity.
SophiaHOST
5:32
Certainly, yeah
5:33
...
5:33
I, I think it's really easy to say, okay, to, to follow that thought experiment, it goes back to, like, IJ Good and Vernor Vinge that says once AI starts making itself, it'll make a better version.
5:44
That one will make a better version.
5:45
That will make a better version in less time, and not think about those coefficients- Hmm ... not think about the number in that feedback loop.
5:52
I think that's some of what's going on.
5:54
Um, I think also the future is uncertain, so the parameters that I'm talking about could change.

16 MINS LATER

SophiaHOST
22:00
My goat.
13:56
Uh, is that, is, is, is that enough? And I guess when I was thinking about this question, I'm like, are, is all, [chuckles] is all the money going into data centers, is the kind of progress you're describing, is that good enough to justify that kind of investment?
14:15
Yeah.
14:15
So I'd say, look, there's wide variance and I'm giving you my views, but let's all update on data.
14:22
And so I think there's a lot of data we can actually ask the labs to give us that will help us understand this phenomenon.
14:29
And that's, uh, that's something I mentioned in my blog post.
14:32
The Elasticity Institute has a certain le- uh, set of requests.
14:36
A paper that came out last week that had, uh, some of the godfathers of AI on it also had a set of requests for the labs to release data, and that would help us ground all of this more in reality or i- in empiricism, right? Two, without additional data, the best data we have is what OpenAI put out a couple weeks ago.

7 MINS LATER

22:04
If there is a one in 10 chance that we can, that we can create an AI that would be worthy of the term superintelligence, and if there is a one in 10, one in 100% chance that that model would get out of control and cause absolutely catastrophic damage to society, why should we bother pursuing it?
0:00
Hello, thank you very much.
0:02
Thank you to the Bitcoin Policy Institute and thank you to the AV team who snuck in updates to my slides just in the nick of time.
0:09
Are my slides ready, folks in the back? All right, thank you.
0:12
So for the next 15 minutes, I'm gonna talk about how and why AI, if it remains plural, can be a tool of liberty for all of us.
0:22
So I'm an investor.
0:22
I invest in clean energy, AI, power for AI, anything that makes the world a better place.
11:04
What's the sort of gap between the reality that Nex has painted and the reality that is sort of coming to be in our current reality? Well, look, I
11:16
mean, the state of the art of neurotech is still putting electrodes in the brain.
11:24
And you've got 100 billion neurons in your brain.
11:27
Neuralink, the state of the art device, puts in 1,000 electrodes.
11:33
So instead of 10 to the 11th, it's 10 to the third.
11:36
So it's 100 million times fewer than the neurons in your brain.
11:41
And it requires brain surgery.

20 MINS LATER

31:27
How do you hold the sort of conversation of psychedelics and cognitive liberty, you know, medicalization versus decriminalization and maybe what we're ready for now versus what we may be available for a decade from now or 20 years from now, if you will?

Unknown podcast

"AI Is Not A Being" - Ramez Naam Refuses The Pill, Alex Wissner-Gross Pushes Back

Aug 16 · 52m

27:35
I mean, that's like science fiction in its own right.
27:37
That's right.
27:38
Absolutely.
27:40
Some of these, ironically, came because of budget cuts in academia.
27:45
If you look at Commonwealth Fusion, which is considered sort of the safe bet of fusion, if you will, if there is such a thing, that team at Harvard, their grants were struggling.
27:57
And so they said, why don't we form a company, you guys? And so they did.
28:01
And they're now the frontrunner.

17 MINS LATER

44:46
What, self-enriching? Would the rubric, would the threshold for saying this is some sort of autonomous being be if it were, say, trying to mine Bitcoin for itself once it gained access to Hugging Face? Is that the criterion in your mind?
63:44
Onwards to the horizon of fission and fusion.
63:47
Okay.
63:48
Uh, let's talk about the atom and the power thereof, going back to the '50s or, uh, you know, retro future.
63:55
Uh, uh, this is a complicated slide, uh, for those of you who are, uh, just listening.
64:01
Basically, there's, you know, two approaches to nuclear fission, which is what we've been doing since the '60s.
64:08
The traditional one is big reactors.
64:11
And the simplest thing you can do to boost nuclear production worldwide is, A, stop shutting down nuclear plants.

36 MINS LATER

100:28
Check it out at moonshots.com.
0:46
We hope you'll enjoy the episode.
0:51
Wonderful.
0:51
Thanks so much, Genevieve.
0:52
It's an honor to be here.
0:53
My first Foresight Conference was 2000.
0:56
That's when I met Chris.
0:58
And it's an honor to be back and to be able to speak to you guys today.

5 MINS LATER

6:02
It has to be that
9:15
Do you think it's possible to actually build the orbital data center? Like, I, I don't think there's actually one that exists yet, but it seems like the math works.
9:22
It's possible.
9:23
The math doesn't work on a price basis yet, but it could work with Starship.
9:27
So look, a, a, a Starlink satellite doing internet is kind of an orbital data center.
9:33
It's, it's using its compute to handle telecoms and compression and, and channel hopping and so on, not for computing AI, but it has some compute on it.
9:43
And we've had the first example of a satellite launch that has a GPU on it doing that sort of thing.
9:49
So nothing says it's impossible.

19 MINS LATER

29:07
There's, like, let's go build general purpose models and chase that general super intelligence, and then these narrow super intelligence pursuits really to me feels like it's a bet just on specialized workflows with industry-based knowledge or experts, and that's kind of where it sounds like you believe the value is.
14:00
So coming today where we have AI, which is at its best an amplifier of our best ideas to implement them, what is the role of open weights?
14:12
Well, look, let me go back in time before I come to the future.
14:15
I'll say, even without open source, everything is open source in the limit.
14:20
Because at the end of the day, IP protections are mostly, you know, societally agreed upon consensus hallucinations about how we're going to, you know, guard information.
14:33
But IP expires.
14:34
At the end of the day, you decompose what somebody else did and you know how to do it.
14:39
Robert Wright, an amazing journalist and author and so on, one of my favorite books is his book, Non-Zero.

29 MINS LATER

43:21
Now and moving forward, how can we best complement each other? What is the potential for humans and AI together?
15:48
So what do you think is, is really the kind of core thesis or, or really the crux of the challenge for renewables, knowing that the case is, is very compelling for renewables and, and hyperscalers to, to combine [laughs] their, their powers, right? What do you think some of the core challenges are there in terms of, you know, regulation, permitting, market design? I, I would love to hear a little bit more of your perspective there when it comes to some of these more process-oriented elements of the energy reality that we have today, potentially serving as roadblocks to renewables playing a bigger role in, in the hyperscale build-out.
16:20
If you just look at time to build, uh, for a medium-sized solar plant or solar and battery plant, uh, it's maybe, you know, twelve, eighteen months to get it going, and permitting is usually fairly easy.
16:34
Uh, but for a data center wanting to get into the large load interconnection queue, it might now be five to seven years.
16:41
And so a lot of that is about T and D. It's bringing the, the solar, uh, or whatever the, whatever the grid is, is powered by to the site.
16:51
And so that's, like, the number one area that I'd focus on.
16:54
There's a second area though that I also, uh, think is worth, uh, pointing out, which is for gigawatt scale data centers, uh, we, you know, don't have a lot of examples of renewable power plants on that scale.
17:07
And so then you start to get into real permitting challenges and also challenges that the developers can potentially address themselves, just putting together large parcels of land, uh, to do this sort of thing.
22:07
Do you feel that we currently have the technology we need to, to serve this version of reality, or what kind of work needs to be done to get to that point?
9:58
Whereas, In Europe, you know, the response to this, and offshore I'm seeing the response to this, is like if another country was doing what America is doing today, America would invade them to remove their fascist dictatorship, right? Exactly.
10:15
So I do agree with the latter.
10:16
And I do think the U.S. is a very individualistic society, maybe the most.
10:20
Maybe Australia is up there.
10:23
Australia is more American than America sometimes.
10:26
But, you know, we solve tragedy of the commons all the time.
10:29
We phased out CFCs that were destroying the ozone layer.

29 MINS LATER

39:06
Is, you know, the global geopolitical situation going to be? Is it going to be, as Ian Bremmer says, a technology Cold War? Or does this force more unification? You know, as AI pushes in, do we need to see more consensus building globally? And what does the US look like in 30 years as a governance structure? Yeah.
35:51
How close do you think are we to a reality where human beings are networked? What ethical lines should never be crossed?
36:02
Yeah, good questions.
36:04
I think we're actually quite distant from human brains really being networked.
36:09
Many things are exponentials.
36:10
The number of electrodes we put in an implant in a human brain doubles every seven years.
36:15
And so the top-of-the-line systems from Neuralink are 1,000 electrodes.
36:19
The brain has 100 billion neurons.
41:12
Well, let me start

We value your privacy

We use cookies to understand how you use our platform and to improve your experience. Click “Accept All” to consent, or “Decline non-essential” to opt out of non-essential cookies. Read our Privacy Policy.