Finally getting around to reading The Mind in the Wheel from @slimemoldtimemold.bsky.social, and boy howdy does the prologue do double duty as an exploration of both superficial science and overblown claims about LLMs. Statistics ≠ understanding, even when useful for, e.g., prediction.
A presentation slide titled 'Outmaneuvering Peer Review: The field guide for getting past Reviewer 2' by Matt Wedel. It includes a quote from John Gilmore about how 'The Net interprets censorship as damage and routes around it.' The slide introduces two strategies: '1. Pick the right enemy' (revealing that the real adversary is the editor, not Reviewer 2) and '2. Deploy agreement chaff' which explains how to sort reviewer suggestions into four categories ranging from useful changes to completely unwelcome ones. The slide features a dramatic illustration of a hooded figure in a library with candlelight.
A continuation of the peer review strategy guide showing points 3 and 4. '3. Pit reviewers against each other' describes using conflicting reviewer opinions to your advantage. '4. Enthusiastic noncompliance' explains how to strategically comply with most suggestions while quietly ignoring certain ones. The slide features a pop art style illustration of a confident businessman in sunglasses surrounded by papers marked 'CORRECTED' in a comic book style, with the caption suggesting compliance while doing the opposite.
In case you haven't read the latest edition of The Loop by @slimemoldtimemold.bsky.social — you should, it's pure joy.
One highlight is the "Field guide for getting page Reviewer 2"
drive.google.com/file/d/1YR3...
In 2023, @slimemoldtimemold.bsky.social conducted an experiment. They asked strangers on the internet try eating almost nothing but potatoes for a month.
For the ABC, this is that story. Listen here or wherever you get your podcasts.
www.abc.net.au/listen/progr...
Time for a story about how when you heat things up, they almost always get bigger.
slimemoldtimemold.com/2025/02/06/t...
AS USUAL, tons more information in the full blog post.
Thank you for reading! :)
slimemoldtimemold.com/2025/02/02/p...
The fact that we planted our flag and ran this as a riff trial didn’t change the nature of this exploration. But making it one study, clearly marking out its existence, definitely sped things up, and helps make the results easier to compare and interpret.
#3: Riff Trials Harness Cultural Evolution
Even if we hadn’t run the riff trial, people would have experimented with potato diets for the next 10 or 20 years, trying new variations and learning new things about the diet-space. But this process would have been slow and chaotic.
#2: You May Have to Encourage Diversity
That said, there was not much diversity in the riffs.
Most people signed up for some version of Potatoes + Dairy. In the future, organizers should think about what they can do to encourage people to sign up for different riffs.
#1: It Works
We hoped that riff trials would use the power of parallel search to quickly explore the boundary conditions of the base protocol, and discover what might make it work better or worse. Good news! This works.
This is the first-ever riff trial. But it won’t be the last. So for the next time someone does one of these, here’s what we’ve learned about how to do them right.
The strongest evidence is against eggs, because participant 41470698 / 80826704 did a kind of self-experiment.
First he did a Potatoes + Eggs riff and lost only 1.8 lbs. Then he did a standard potato diet and lost 13.2 lbs. That’s a pretty stark comparison.
There’s some evidence that meat, oil, vegetables, and especially eggs make the potato diet less effective. But with such a small sample, it’s hard to know for sure.
The main thing we learned is that Potatoes + Dairy works almost as well as the normal potato diet.
Looking at the 10 cases that did just potatoes and dairy, the average weight lost on these riffs was 9.2 lbs, pretty comparable to the 10.6 lbs lost on the standard potato diet.
The potato diet continues to be really robust. You can eat potatoes and ketchup, protein powder, or even skittles, and still lose more than 10 lbs in four weeks.
Mean weight change was 6.4 lbs lost, with the most gained being 5.2 lbs and the most lost being two people who both lost 19.8 lbs.
One person gained weight, one person saw no change, one person reported no data, one person didn't do potatoes, and the rest lost weight.
With these two last-minute entrants, the final count comes to a total of 40 riffs.
Note that this is not the same as 40 participants, since some people reported multiple riffs, and a few riffs were pairs of participants.
Here they are:
Participant 80826704, who had previously done Potatoes + Eggs as participant 41470698 and lost almost no weight, this time did a standard potato diet and lost more than 13 lbs:
Since the last report, we’ve heard back from an additional two riffs.
Participant 87259648 did a Fried Potatoes riff and lost a little weight:
Between March 18th and October 9th, 2024, we heard back from an additional eleven riffs.
We saw continued support for Potatoes + Dairy, and there were more skittles.
slimemoldtimemold.com/2024/10/09/t...
Between January 5th and March 18th, 2024, we heard back from an additional seventeen riffs.
Potatoes + Dairy continued to work just fine. Also, the Potatoes + Skittles riff was an enormous success.
In the first two months after launching the riff trial, we heard back from ten riffs.
Generally speaking, we learned that Potatoes + Dairy seems to work just fine, at least for some people:
slimemoldtimemold.com/2024/01/05/f...
Why the hell does this happen? Well, there are many theories. The hope was that running a riff trial would help get a sense of which theories are plausible, find some boundary conditions, figure out what slows or stops the effect, or just more randomly explore the diet-space.
As the first test of this new design, we decided to riff on one of our previous studies: the potato diet. For many people, eating a diet of ~nothing but potatoes causes quick, effortless weight loss, 10.6 lbs on average.
slimemoldtimemold.com/2022/07/12/l...
The riff trial is a new type of study design. In most studies, all participants sign up for the same protocol, or for a small number of similar conditions. But in a riff trial, everyone tests a different version of the original protocol, and you see what happens.
Just over a year ago, we launched the Potato Diet Riff Trial, the first study of its kind.
slimemoldtimemold.com/2023/11/03/p...
Let's take a retrospective look at the first ever *riff trial*:
that's part of it
peanut toxic chocolate lactose art gallery 4D lamb chop placebo ant colony
slimemoldtimemold.com/2024/11/29/l...
Participating in obesity research has been a low-key hobby of mine for two years now and the results are interesting. bsky.app/profile/slim... I am not sure how much I've added, but "try something, write down what happens, sustain over time" is something anyone can try and everyone should.