I think we can all agree that algorithmic information theory applies to every branch of science. The theory that compresses the observed data to the fewest bits is the best predictor of future observations. Occam noticed this in 1485 and it has held ever since. The reason that Solomonoff induction works is that all possible probability distributions over an infinite set of finite length descriptions must favor shorter ones, in the sense that for any description there must be an infinite set of other descriptions that are both less likely and longer, and the set of the other 3 possibilities must be finite.
So my question is what questions do you want to answer and what data would you compress? For example, suppose my question is whether AGI will lead to a Kardashev level IV civilization of 10^80 human mind equivalents spread across the galaxies, or will it lead to self replicating nanotechnology killing all DNA based life (or both)? To answer that, I could compress historical population data to a polynomial fit plus residual errors and get the answer from the sign of the leading coefficient. Hume's guillotine (Laboratory of the Counties) and the Hutter prize (enwik9 from Wikipedia) both have detailed population data. So we could do this. But would you trust the answers? Or did you have a different question and different data in mind? -- Matt Mahoney, [email protected] On Mon, Jul 20, 2026, 9:48 PM James Bowery <[email protected]> wrote: > Like William C. Norris, Jay Forrester was just another farm kid from > Nebraska. He was instrumental in creating the electronic computer (1951 > Whirlwind project). Why then did he almost immediately abandon computer > engineering? > > To pursue what I agree is the most important application of computation: > > Macrosocial Dynamical Systems > <https://github.com/jabowery/HumesGuillotine>. > > This <https://youtu.be/k_GtTqt3QAA>interview shortly before his death > <https://youtu.be/k_GtTqt3QAA> reflects what I, too, have experienced: > > People think "AGI" or "ASI" or whatever, is the most important > application of computers nowadays. I suppose there is a sense of > "important" that may apply. For example, an asteroid "the size of Texas" > hitting Earth at 20km/sec could be considered important. However, it might > be just as important to understand the system dynamics of multibody > gravitational interaction so one could direct a small impulse at apogee to > make its effect less destructive. > > The data selection process must include those with the greatest influence > on macrosocial narratives; challenge them to produce the data upon which > they base their narratives. This is because centalization of social policy > is bringing us to the brink of a rhyme with Gutenberg Press -> > Decentralization of Religious Agency -> 30 Years War in which the Internet > -> Populism ->... is the current state of the rhyme. This will be somewhat > like attempting to bring the Jesuits together with the Calvinists and > Lutherans to debate theology, but the fact that nowadays everyone dons the > robes of the white lab coat gives hope that they may at least be > embarrassed into an honest dialogue. > > Once that dataset is agreed upon -- and we can afford to make it extremely > inclusive so long as we don't let people claim that they jusst HAVE to > include torrents of real time data so they can declare the whole project to > be impractical -- you need devote only about 1% of the computation and > money already going into the Texas-sized asteroid -- to make its > incremental lossless compression a prize competition that is impervious to > conflicts of interest, such as leakage of evaluation data or endless > palavering that will influence no one to do anything about the asteroid > until it is too late. > *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + > delivery options <https://agi.topicbox.com/groups/agi/subscription> > Permalink > <https://agi.topicbox.com/groups/agi/T5b58bcc51c493d41-Mcaab652417203653f5819cea> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T5b58bcc51c493d41-Mbf31dd36dd977c03655b6c5b Delivery options: https://agi.topicbox.com/groups/agi/subscription
