A while ago I wrote how I managed to add 13 points to my IQ (as measured by the mean between 4 different tests).

I had 3 “self-experimenters” follow my instructions in San Francisco. One of them dropped off, since, surprise surprise, the intervention is hard.

The other two had an increase of 11 and 10 points in IQ respectively (using the “fluid” components of each test) and an increase of 9 and 7 respectively if we include verbal IQ.

A total of 7 people acted as a control and were given advantages on the test compared to the intervention group to exacerbate the effects of memory and motivation, only 1 scored on par with the intervention group. We get a very good p-value, considering the small n, both when comparing the % change in control vs intervention (0.04) and the before/after intervention values [...]

---

Outline:

(01:03) Working Hypothesis

(01:58) The Rub

(02:45) What's next

(05:56) Confounder elimination

---

First published:

Source:

Linkpost URL:
https://morelucid.substack.com/p/increasing-iq-by-10-points-is-possible

Narrated by TYPE III AUDIO.

A new paper by Finlayson et al. describes how to exploit the softmax bottleneck in large language models to infer the model dimension of closed-source LLMs served to the public via an API. I'll briefly explain the method they use to achieve this and provide a toy model of the phenomenon, though the full paper has many practical details I will elide in the interest of simplicity. I recommend reading the whole paper if this post sounds interesting to you.

Background

First, some background: large language models have a model dimension that corresponds to the size of the vector that each token in the input is represented by. Knowing this dimension _ d_{text{model}} _ and the number of layers _ n_{text{layers}} _ of a dense model allows one to make a fairly rough estimate _ approx 10 n_{text{layers}} d_{text{model}}^2 _ of the [...]

---

Outline:

(00:33) Background

(01:44) The method of attack

(03:39) Results

(06:43) What will labs do about this?

The original text contained 2 footnotes which were omitted from this narration.

---

First published:

Source:

Linkpost URL:
https://arxiv.org/abs/2403.09539

Narrated by TYPE III AUDIO.

This is the seventh post in a sequence that demonstrates a complete naturalist study, specifically a study of query hugging (sort of), as described in The Nuts and Bolts of Naturalism. This one demos phase four: Experimentation. For context on this sequence, see the intro post. Reminder that this is meant as reference material.

Wait, there's more to this study? But we’ve just discussed the main insight that came out of it, and how it illustrates the point of naturalism. Why is there more?

There is more because by this point I was interested not only in insights, but in mastery. There is more to mastery than reconceptualization.

However, I would like to point out that everything I’d done so far preceded experimentation. I had not even begun to try to change anything—yet I had learned quite a lot, through mere observation without interference. This [...]

---

First published:

Source:

Narrated by TYPE III AUDIO.

I've been in many conversations where I've mentioned the idea of using neuroscience for outer alignment, and the people who I'm talking to usually seem pretty confused about why I would want to do that. Well, I'm confused about why one wouldn't want to do that, and in this post I explain why.

As far as I see it, there are three main strategies people have for trying to deal with AI alignment in worlds where AI alignment is hard.

1. Value alignment
2. Corrigibility
3. Control/scalable alignment

In my opinion, these are all great efforts, but I personally like the idea of working on value alignment directly. Why? First some negatives of the others:

1. Corrigibility requires moderate to extreme levels of philosophical deconfusion, an effort worth doing for some, but a very small set not including myself. Another negative of this approach is that by-default the robust solutions to [...]

The original text contained 1 footnote which was omitted from this narration.

---

First published:

Source:

Narrated by TYPE III AUDIO.

Whenever I start telling someone about toki pona, they ask at least some of these questions. So I compile the questions and my answers here.

Toki pona is a constructed language notable for having under 200 words. The strange writing that probably prompted you to ask me about it is sitelen pona.

How do you say anything with so few words?

You refer to most things with multi-word phrases, where some words act as adjectives or adverbs.

Once you know all the words of toki pona, you can combine them to express anything, tho an accurate phrasing can get long.

Did you make it up?

Sonja Lang made it up in 2001.

Is it just a rearrangement of English?

Toki pona has a grammar of its own, which is similar to [...]

---

Outline:

(00:26) How do you say anything with so few words?

(01:11) Did you make it up?

(01:17) Is it just a rearrangement of English?

(01:35) Does anyone actually use it?

(01:50) Thats so stupid.

(01:56) Why do you use it?

(02:14) How does that writing system work?

---

First published:

Source:

Linkpost URL:
https://dkl9.net/essays/toki_pona_faq.html

Narrated by TYPE III AUDIO.

Muireall and DaemonicSigil trade physics problems. Answers and discussion of the answers have been spoilered so you can try the problems yourself. Please also use spoiler formatting (type ">!") in the comments.

Smeared Out Sun

Okay, so the first problem is from Thinking Physics. (I promise I'll have some original problems also. And there's a reason I've chosen this particular one.)

It's the problem of the Smeared out Sun (caution! link contains spoilers in the upside-down text).

The problem goes as follows: The sun is far enough away that we could replace it with a disc of equal radius at the same temperature (and with the same frequency-dependent emissivity), and so long as the plane of the disc was facing the Earth, there would be little difference in the way it was heating the Earth. While scientists would certainly be able to tell what had happened, there would be [...]

---

Outline:

(00:27) Smeared Out Sun

(03:48) Measuring Noise and Measurement Noise

(08:09) Relativistic Rain

(14:04) Martian Dome

(20:42) Thermodynamic Computer

---

First published:

Source:

Narrated by TYPE III AUDIO.

This article contains more than 100 uses of logical or mathematical notation, so an audio narration would be too hard to follow. You'll find a link to the original text in the episode description.

---

First published:

Source:

Narrated by TYPE III AUDIO.

I did an exploration into how Community Notes (formerly Birdwatch) from X (formerly Twitter) works, and how its algorithm decides which notes get displayed to the wider community. In this post, I’ll share and explain what I found, as well as offer some comments.

Community Notes is a fact-checking tool available to US-based users of X/Twitter which allows readers to attach notes to posts to give them clarifying context. It uses an open-source bridging-based ranking algorithm intended to promote notes which receive cross-partisan support, and demote notes with a strong partisan lean. The tool seems to be pretty popular overall, and most of the criticism aimed toward it seems to be about how Community Notes fails to be a sufficient replacement for other, more top-down moderation systems.[1]

This seems interesting to me as an experiment in social technology that aims to improve group epistemics [...]

---

Outline:

(01:06) How does the ranking algorithm work?

(06:03) Some further details and comments

(11:12) Academic commentary

(13:06) Conclusion

The original text contained 12 footnotes which were omitted from this narration.

---

First published:

Source:

Narrated by TYPE III AUDIO.

Introducing Devin

Is the era of AI agents writing complex code systems without humans in the loop upon us?

Cognition is calling Devin ‘the first AI software engineer.’

Here is a two minute demo of Devin benchmarking LLM performance.

Devin has its own web browser, which it uses to pull up documentation.

Devin has its own code editor.

Devin has its own command line.

Devin uses debugging print statements and uses the log to fix bugs.

Devin builds and deploys entire stylized websites without even being directly asked.

What could possibly go wrong? Install this on your computer today.

Padme.

The Real Deal

I would by default assume all demos were supremely cherry-picked. My only disagreement with Austen Allred's statement here is that this rule is not new:

Austen Allred: New rule:

If someone only [...]

---

Outline:

(00:02) Introducing Devin

(00:48) The Real Deal

(03:27) The Metric

(04:48) What Could Possibly Go Subtly Wrong?

(07:29) What Could Possibly Go Massively Wrong for You?

(09:36) If This is What Skepticism Looks Like

(12:33) What Happens If You Fully Automate Software Engineering

(16:24) What Could Possibly Go Massively Wrong for Everyone?

(18:24) Conclusion: Whistling Past

---

First published:

Source:

Narrated by TYPE III AUDIO.

Churchill famously called democracy “the worst form of Government except for all those other forms that have been tried from time to time” - referring presumably to the relative success of his native Britain, the US, and more generally Western Europe and today most of the first world.

I claim that Churchill was importantly wrong. Not (necessarily) wrong about the relative success of Britain/US/etc, but about those countries’ governments being well-described as simple democracy. Rather, I claim, the formula which has worked well in e.g. Britain and the US diverges from pure democracy in a crucial load-bearing way; that formula works better than pure democracy both in theory and in practice, and when thinking about good governance structures we should emulate the full formula rather than pure democracy.

Specifically, the actual governance formula which is “worst except for everything else we’ve tried” is:

• Give a de-facto veto to [...]

---

Outline:

(01:08) A Stylized Tale of Democracy

(02:58) The Recipe Which Works In Practice

(05:17) Some Takeaways

The original text contained 1 footnote which was omitted from this narration.

---

First published:

Source:

Narrated by TYPE III AUDIO.