# Signing the Statement on Superintelligence

By [Continuations](https://continuations.com) · 2026-08-18

asi, society, progress, values

---

[Gigi](https://x.com/gigidanziger_) and I have signed the [Statement on Superintelligence](https://superintelligence-statement.org/), which calls for a pause on the development of superintelligence. This blog post lays out our rationale for doing so now after having previously held off on supporting other “pause AI” appeals.

We have a [broad set of initiatives](https://eutopia.global) in the Hudson Valley and beyond aimed at securing human thriving in the AI age. These include a basic income pilot, various steward ownership projects, and a community learning center.

Two preconditions appear, to us, necessary for any positive future. One is slowing and reversing global warming. The other is avoiding existential risk from artificial superintelligence (ASI).

For several years now, Gigi and I have helped fund research on both these fronts. For example, on the climate side, we have contributed to work on solar radiation management and Arctic cloud thinning. These efforts are in keeping with our approach of supporting endeavors that are (so far) underfunded.

On the ASI side, our first step was to activate more philosophers to work on the issue. We supported the launch of the [ValuesLab](https://valueslab.org) at Columbia and were early contributors to [Cosmos Institute](https://www.cosmos-institute.org/). 

We then looked for people taking novel approaches to the tangle of questions around free will, consciousness and causality. One person who stood out is [Erik Hoel](https://x.com/erikphoel), the author of the wonderful book [The World Behind the World](https://www.theintrinsicperspective.com/p/world-behind-the-world). We have helped fund Erik's lab and he's been making strong progress (e.g. [recent paper](https://www.cell.com/patterns/fulltext/S2666-3899\(25\)00320-4)).

Finally we continue to be interested in the idea that models need to be raised rather than trained. Here is a post from last year about [bringing alignment to pretraining](https://continuations.com/more-thoughts-on-values-in-pretraining). We failed last year to find anyone excited about this approach. Recently, however, [a paper pursuing this exact strategy was published](https://arxiv.org/abs/2608.13482); showing promising early results. We will also be supporting [this work](https://modelraising.ai/) to help it scale to bigger models with reasoning that can be subjected to reinforcement learning.

All of this is to say that Gigi and I have been extremely active in working towards a future of high automation and human thriving. We had held off on signing a petition calling for a pause or slowdown because we had hoped that the frontier labs would make more progress on alignment and society would make more progress preparing. Over the last month, however, it has become clear that capabilities have been pushed ahead way beyond what anyone is prepared for.

While signing petitions is a small act without any immediate consequences, not speaking up clearly seems wrong at this point. We are redoubling the work on our initiatives and increasing funding to projects mentioned above. The world, however, badly needs to buy time for these and similar approaches to be sufficiently developed.

P.S. A separate post on how a pause might be achieved is in the works, along with an explanation of why this might even be in the economic interest of the frontier labs.

---

*Originally published on [Continuations](https://continuations.com/signing-the-statement-on-superintelligence)*