Summary
AI & wisdom vs intelligence
AI = tool for manipulating information. Big shift. Real.
But “powerful tech → solves big problems” is bad logic. Powerful tech → transforms society. How = unknown.
Shelter bench test: Does AI fix housing crisis? No. Problem isn’t lack of intelligence applied to housing. Problem is landowners have power, block density, protect scarcity. AI can’t fix that. Steam engine can’t fix hurricanes. Different universe.
TB same: Know how to prevent 1M+ deaths/year already. Problem isn’t knowledge. Problem is institutions lack will/capacity. AI diagnostics help at margins. Core blocker unchanged.
Key distinction:
- Intelligence = manipulate information → process, sort, predict, generate
- Wisdom = want right things, solve right problems without making worse problems
AI mass-produces intelligence. Nobody mass-produces wisdom.
World shifting from intelligence-constrained → wisdom-constrained. More intelligence into wisdom-constrained world = unpredictable. Could help. Could concentrate power further. Both simultaneously.
Cancer drug example: easier to make drug than get it to patients. Cancer doesn’t care drug exists if patient never receives it.
Bottom line: AI is powerful. Impact unknown. Will it let wisdom flourish or let powerful route around wisdom? No one knows.
Transcript
[00:00] Good morning, John. You’re right now in [00:01] Sierra Leone. I was just texting with [00:03] you. You’re visiting the maternal center [00:05] of excellence, which I cannot wait to [00:06] see more of and it is making me wonder [00:08] about something that I think is [00:09] important and I apologize, but it is [00:11] about AI. I think that AI is a big deal [00:14] and I do think that it’s going to change [00:15] a lot of things. But whenever somebody [00:17] tells me that AI is going to change [00:19] everything and solve like all problems, [00:21] I check in with a little standard I’ve [00:23] created. It’s maybe you would call it a [00:25] benchmark. Not like how good is it at [00:27] drawing a pelican on a bike or inventing [00:30] a novel protein or beating a grandmaster [00:32] at chess. We’re going to call it shelter [00:33] bench. Uh how will AI help solve the [00:37] housing crisis? Because like to me the [00:38] lack of housing that people want where [00:41] people want it is like one of the big [00:43] problems that underlies many many other [00:46] problems in America. And I know that it [00:48] is worse where I live than other places [00:50] but it is a very big problem and it [00:53] underlies a lot of other problems. And [00:55] on that benchmark, I think AI does very [00:58] poorly. And it’s strange to me that we [01:00] don’t even look at this or think about [01:02] it, but like obviously [01:04] ask this question. If it’s such a big [01:06] deal, how does it solve the biggest [01:09] problems? Maybe it would help like a [01:11] tiny bit of the margins. Maybe you could [01:13] do permitting more quickly, maybe [01:15] cheaper code review or design, but like [01:18] that’s not what’s blocking housing in [01:20] America. I don’t think anybody thinks [01:22] that we haven’t like applied enough [01:23] intelligence to the problem. The lion [01:26] share of the issue and there’s other [01:28] things around this but the lion share is [01:30] that people who own land in an area have [01:33] power over that area and they have all [01:36] kinds of reasons to try and prevent new [01:39] housing from being built there, [01:41] especially more dense new housing. They [01:43] don’t want more traffic. They’re scared [01:44] of change. They know that if 20 people [01:46] move in, maybe two of those people are [01:48] going to be neighbors that they don’t [01:49] particularly like. Their property values [01:52] benefit from scarcity. Nobody thinks AI [01:54] is going to fix that. It’s like saying [01:56] that steam engines are going to fix [01:57] hurricanes. Like there are problems that [01:59] exist in different universes. So how [02:02] does AI do on the shelter bench [02:04] benchmark? Uh badly. like it might [02:06] potentially be actively making it worse [02:08] because it might further concentrate [02:10] wealth and intense inequality creates [02:12] weird economic incentives to only serve [02:14] the wealthiest consumers. Anyway, we [02:16] could also take a look at another [02:17] similar series of problems with TB. We [02:20] here all know that tuberculosis kills [02:22] more than a million people a year and [02:23] we’ve known how to prevent almost all of [02:24] those deaths for decades. So, that [02:26] problem can’t be intelligent because we [02:28] have the knowhow. Now AI might help with [02:30] diagnostic image reading or developing [02:32] new antibiotics or surveillance modeling [02:35] which it would be great like all those [02:37] things are great but those are not the [02:39] thing that would solve the problem [02:41] because we have solutions now and doing [02:43] any of those things or implementing any [02:44] of those things would still require [02:46] institutions that want the outcomes and [02:49] can execute on them. Again it’s the same [02:51] problem. The people who need the help [02:53] who need the resource don’t have power [02:56] over the resources. People who need [02:58] housing don’t have any sway inside of a [03:00] community. They don’t live there. They [03:02] don’t have housing. The story of like a [03:04] transformative technology tends to be [03:07] told as one, this technology is very [03:09] powerful. Two, powerful technologies [03:12] transform society. Three, therefore this [03:15] technology will help solve our biggest [03:16] problems. But I don’t know if you like [03:18] spotted the logical problem there. Three [03:20] does not follow from one and two. What [03:22] follows from one and two is therefore [03:24] our society will be transformed in what [03:27] ways it is transformed is not clear. Now [03:30] John, some people are still out here [03:31] thinking that AI is a scam. To me that’s [03:32] obviously not what’s going on. So I want [03:34] to try and articulate what is going on. [03:37] But in order to do that I have to [03:38] redefine intelligence. I mean redefine. [03:41] Can anyone define intelligence? No. [03:44] Obviously not. But throughout this video [03:45] I have been saying that you like you [03:47] can’t solve the housing crisis with more [03:49] intelligence. But there is like a [03:51] version of intelligence. You could [03:52] define intelligence in such a way that a [03:55] bunch of intelligence would help solve [03:57] the housing crisis, especially if it was [03:59] spread out among a lot of people who [04:00] have power. But there’s also a [04:02] definition of intelligence that [04:04] definitely could not help solve the [04:05] housing crisis. And that’s the one I’m [04:06] using right now. So I want to tease [04:08] these things out. I have lately been [04:10] thinking about intelligence as like [04:12] power over information. So power in like [04:15] the physics sense is the ability to [04:17] basically manipulate atoms. heat them [04:19] up, speed them up, pick them up, drop [04:21] them, slam them into each other, melt [04:23] them, mix them, push them over here. [04:24] It’s doing all of those things. So, [04:26] lately, I’ve been thinking of [04:28] intelligence as the ability to [04:30] manipulate information. Mix them up, [04:32] make new things out of it, push some [04:33] over here, bring a bunch of it to the [04:35] top of a mountain, and throw it off. You [04:37] know, like that. Like, I don’t know what [04:38] I mean exactly, but like the ability to [04:41] manipulate information, which I know is [04:43] not like what we mean precisely by [04:46] intelligence, but like what is it? [04:48] There’s no good definition anyway, but I [04:49] think that there’s some value in looking [04:50] at it through this lens for a moment [04:52] here where intelligence is the ability [04:55] to manipulate information. And yeah, [04:57] crows do that and single- cellled [04:58] organisms do that and and like maybe [05:00] search engines do that. And maybe we [05:02] don’t think of those things as having [05:04] intelligence. You know, certainly [05:05] they’re not intelligent, but they might [05:07] have intelligence. I like maybe they do. [05:09] Regardless, if the shorthand I’m using [05:11] for the ability to manipulate [05:12] information is the word intelligence, [05:14] then intelligence you get like [05:16] processing and sorting and predicting [05:17] and even generating. We just have way [05:19] more of it than we used to have, even [05:21] without AI and potentially very cheaply. [05:24] We have way more ability to manipulate [05:26] information and to have it do stuff for [05:28] us than we previously had. That’s very [05:30] powerful. That’s very interesting. And [05:32] nobody knows what that means. We live in [05:35] an intelligence constrained world. And [05:37] so if you have more of it, like a bunch [05:39] of stuff’s going to get created that [05:40] otherwise wouldn’t happen. But [05:42] intelligence is separate from what I’m [05:44] just going to go ahead and call wisdom. [05:46] And I don’t think that we have a way to [05:48] mass-produce wisdom. So perhaps we have [05:51] been moving throughout my lifetime from [05:53] a world that was intelligence [05:54] constrained to one that is wisdom [05:56] constrained. Perhaps that transition [05:58] started a while back, but we are in the [06:00] midst of it still. So for more examples, [06:03] intelligence would help you get what you [06:05] want, whereas wisdom would help you want [06:08] what you should want or the right [06:09] things. It’s the ability to figure out [06:11] which problems are worth solving and [06:14] then to solve them in ways that don’t [06:15] create worse problems in the process, [06:17] which is not easy. Even the wise fail on [06:20] that sometimes. But like designing a [06:21] more effective slot machine is an [06:23] application of intelligence. I don’t [06:24] think that you would call it an [06:25] application of wisdom. And wisdom also [06:27] has to survive contact with reality and [06:29] also the other people who make up [06:31] reality. I don’t see any reason to think [06:33] that like making intelligence extremely [06:35] efficient would like change the power [06:38] dynamics that create an unjust world. It [06:40] might help. It might hurt. It might do [06:42] both at the same time or in different [06:44] situations or at different scales. It is [06:46] impossible to know though I certainly [06:48] see a concentration of power being [06:51] somewhat inevitable here. But maybe not. [06:53] I don’t know. Oh, a frame that has been [06:55] resonating with me is that AI is to some [06:57] extent a technology. It is a tool that [07:00] already makes like a fairly broad array [07:03] of tasks easier, probably make more [07:06] tasks easier in the future. I think that [07:08] it’s a genuinely a big technological [07:10] shift. That is sort of how I’m imagining [07:12] it. There’s a lot of, you know, leaping [07:14] seven steps down the path that I don’t [07:16] think is valuable because nobody can [07:18] predict any of these things. But like as [07:20] of right now, it is a big technological [07:22] shift and so has been the internet and [07:24] so has been personal computing. These [07:27] things did not solve the housing crisis. [07:28] Like it’s wild to say this, but it is [07:30] obviously true that it will be easier [07:32] for AI to create a cancer drug than it [07:35] will be to get that cancer drug to all [07:38] the people who need it. And I think that [07:39] it is important to recognize that those [07:42] problems are both problems. The cancer [07:46] doesn’t care if the drug exists. that is [07:48] not going to be affected by the [07:49] existence of a drug that is not being [07:51] given to a patient. So the question [07:53] isn’t whether AI is powerful. I think [07:55] that it clearly is. It’s just that no [07:56] one can know what its impact will be. [07:58] Will it allow wisdom to flourish or will [08:02] it allow the powerful to route around [08:05] wisdom as they uh tend to do when given [08:08] the opportunity? And I don’t know how to [08:10] answer that question, but John, I hope [08:12] you’re having a good trip. And please [08:13] tell everyone at the Maternal Center of [08:15] Excellence that they are doing work that [08:17] really matters. [08:22] I’ll see you on Tuesday. I’m about to [08:23] upload this video and I just wanted to [08:25] say that Neil Patel put out a video [08:27] yesterday about some of his thoughts on [08:29] similar topics that in part inspired [08:31] this video and I’ll link that in the [08:33] description. And also John, I’m just [08:34] going to do a screenshot. I’ll show you [08:35] this. Uh you texted me this morning and [08:37] you said the MCOA is incredible. It is [08:40] truly truly incredible. Pretty cool.