# Max Tegmark on Existential Risk and a Moratorium on Artificial Intelligence Training

**URL:** <https://scanalyst.fourmilab.ch/t/max-tegmark-on-existential-risk-and-a-moratorium-on-artificial-intelligence-training/3041>\
**Category:** Tracking with Closeups\
**Tags:** max-tegmark, existential-risk, artificial-intelligence\
**Created:** [14 April 2023 12:38 UTC](https://scanalyst.fourmilab.ch/t/max-tegmark-on-existential-risk-and-a-moratorium-on-artificial-intelligence-training/3041 "2023-04-14T12:38:31Z")\
**Posts on this page:** 1\
**Showing post:** 5

<div class="post-metadata">

**Author:** ![jabowery](https://scanalyst.fourmilab.ch/user_avatar/scanalyst.fourmilab.ch/jabowery/32/5478_2.png) [@jabowery](https://scanalyst.fourmilab.ch/u/jabowery)\
**Post date:** [17 April 2023 14:39 UTC](https://scanalyst.fourmilab.ch/t/max-tegmark-on-existential-risk-and-a-moratorium-on-artificial-intelligence-training/3041/5 "2023-04-17T14:39:01Z")

</div>

> [@johnwalker](#):
>
> wrap applications around transformer models, as I previously discussed in relation to [Auto-GPT](https://scanalyst.fourmilab.ch/t/some-experiments-with-chatgpt/2383/393), [Baby AGI](https://scanalyst.fourmilab.ch/t/some-experiments-with-chatgpt/2383/394), and [Generative Agents](https://scanalyst.fourmilab.ch/t/moore-is-different-generative-agents-interactive-simulacra-of-human-behavior/3039), it may be that recurrence can be achieved by these models by adding external storage and retrieval and feedback mechanisms communicating with it in natural language.

[This paper](https://arxiv.org/pdf/2301.13196.pdf) used a “[one instruction set computer](https://scanalyst.fourmilab.ch/t/a-simple-question-to-quell-agi-hysterics/3044/8)” paradigm to universalize transformers!

 ![image](https://scanalyst.fourmilab.ch/uploads/default/original/2X/b/b703986fa9e5d7c71b86ac1c64617d00d339ed91.png)

---

_[View the full topic](https://scanalyst.fourmilab.ch/t/max-tegmark-on-existential-risk-and-a-moratorium-on-artificial-intelligence-training/3041)._