Safety
I read some strange criticisms of this post. So here is my take. I’m not fully endorsing the piece by Newport, but there’s quite some factua…
I read some strange criticisms of this post. So here is my take. I’m not fully endorsing the piece by Newport, but there’s quite some factually true information in it. Responsible and careful people (
I read some strange criticisms of this post. So here is my take. I’m not fully endorsing the piece by Newport, but there’s quite some factually true information in it. Responsible and careful people (in or out of the mentioned subcultures) should acknowledge the disservice (alongside the service) that EA and Yudkowskian rationalist communities have done to AI governance AND AI governance discourse. I think a charitable reading of Pinker's post here is this: the classic doomer AI-extinction narrative (a singleton AI or a family of super powerful AIs will likely kill all of humanity so literal “extinction” via literal AI “takeover”) have hurtfully monopolized quite a lot of attention for quite some time. Attention is a scarce resource. Much more attention could have otherwise gone toward mitigating concrete, more immediate risks like cybersecurity or biosecurity disasters. The catalogue of risks in relation to existential risks was "popularized" by detailed argumentation only quite recently. Here are two sources: in January 2024 via the concept of "gradual and cumulative x-risk" (https://link.springer.com/article/10.1007/s11098-025-02301-3), and later in January 2025 via the "gradual disempowerment thesis": https://arxiv.org/pdf/2501.16946 These recent papers bridged "immediate, concrete risks" with "existential risks." @SigalSamuel from @voxdotcom has given a quite extensive coverage of the disourse here: https://www.vox.com/future-perfect/461680/if-anyone-builds-it-yudkowsky-soares-ai-risk I think it's time to stop strawmaning those who challenge the classic "ASI takeover-and-literal extinction" scenario. Those who strawman, intentionally or unintentionally, frequently suffer from a "dataset bias" fallacy by falsely claiming that those challengers of the classic scenario are "skeptics" or "deniers" (without saying skeptics or deniers of what exactly") and then saying that the challengers only care about immediate, localized harms of dataset bias. This binary is so deeply outdated. For Many people outside the EA and Yudkowskian-Rationalist subcultures, the catalog of serious, dangerous, systemic, and catastrophic risks has always been far broader. It included complex, compounding deeply dangerous vulnerabilities by the development and/or deployment of advanced AI that could threaten society’s long-term resilience without requiring a singular, speculative "apocalyptic machine" moment. As Newport points out, harping on doom for the species changes the subject from immediate and obvious threats from AI, such as enabling bioterrorism, undermining truth-seeking institutions, and breaching cybersecurity.
Related
- Why oh why do so many people take this man’s predictions seriously?
- true. a lot of people here just don’t have the historical context.
- Seriously. If you don’t understand the below, you really shouldn’t speculate about AI. It’s absolutely fundamental.
Source: Gary Marcus (X) | 2026-09-06