Decentralized Social Media Is the Only Alternative to the Tech Oligarchy

Running AI Locally Without Spending All Day On Setup – https://hackaday.com/2025/01/08/running-ai-locally-without-spending-all-day-on-setup/

This is the 3rd post about my DIY psi experiment. Here are Part 1 and Part 2.

I was seeing if I could influence a random-number-generator mentally or psychically. The idea was to roll a lot of virtual dice and will the number FOUR to come up.

The results are negative. No evidence of psi.

While willing the number 4 to come up, I got: 1034 ones 982 twos 1043 threes 955 fours 1043 fives 943 sixes

• "It was just a coincidence" – prevented by statistical calculation
• "You analysed the data various ways until you found a way of making it look good" – prevented by this post, which commits me to a particular methodology before I conduct the experiment. I will post the data whether it shows an effect or not.
• "You quit when you were ahead" – prevented by this post; I have decided on the number of trials in advance: 6000
• "You lied" – this I can't prove to you, but I know for my own self. You can try replicate.
• "The dice are biased" – To control this, I rolled 600 virtual dice on the 16th of April and made no attempt to psychically influence them. I got: 984 ones, 973 twos, 1007 threes, 1015 fours, 964 fives, 1057 sixes

1. According to an article in 'Evidence for psi thirteen empirical research reports' edited by Ben Gortzel, psi effects are strongest around 13:47 sidereal time; I will use sidereal.app/calculate to optimise this
2. It's the day of a full moon; Dean Radin has investigated if this affects psi
3. random.org uses a random number generator in Dublin, Ireland, about one mile from where I conducted the experiment

Years ago, I tried to influence the random-number-generator on random.org with my mind. I was surprised to get results with p<0.01 likelihood, but I no longer have the data. Tomorrow, Tuesday the 23rd, I will try a larger repeat performance.

random.org allows you to generate integers, lottery picks, dice, etc. Among these, https://www.random.org/dice/ is the one I've chosen because it allows me to gather a large amount of data quickly: you can roll 60 virtual dice with one click.

I plan to roll 6000 virtual dice, while trying to psychically will the number FOUR to come up. If there is a small but real effect, these large numbers will allow me to detect it.

On the 3rd of April, I did a small experiment: I rolled 600 virtual dice and got 118 fours. This is a p=value of 0.02956, meaning this abundance of fours will occur 3% of the time by chance. Pretty good.

These results are interesting enough that I want to do it once and for all with 6000 rolls instead of 600. This should give me a decisive p-value. If the same proportion of fours is replicated again among 6000 rolls (1180 instead of the chance-expected 1000), the odds of fluke approach zero. Any number of fours above about 1140 is a one-in-a-million fluke.