Matt, thank you for educating us. I think your public, Nihilist pro-technology stance may now have excused you from the "Table of the Concerned". Occam's Razor.
On Tue, 08 Sept 2026, 23:42 Matt Mahoney, <[email protected]> wrote: > Quan Tesla wrote: > >> >> A sidenote; Occam's Razor may be great for a quick research-population >> poll, as a "thumbsuck" knowledge indicator, but not as a valid and reliable >> derivation of the most-correct Algorithmic Information Set for a boundaried >> system under consideration. Polling results offer useful insight and a >> refetential parking space for future checks and balances, e.g., "But most >> of you agreed how in your performance reality, "A" logically causes "B". >> The intrinsic system seems to differ." ??? >> > > That's not how AIT works. Occam's Razor and AIT says that the shortest > theory that is consistent with past data makes the best predictions. > For example, Hume's guillotine has a database of county level crime and > income statistics with a negative correlation, meaning by knowing one, you > can predict (and compress) the other. What it does not show is which causes > which. Consider the following 4 theories: > > 1. Poverty causes crime. > 2. Crime causes poverty. > 3. Racism causes crime and poverty. > 4. Low IQ causes crime and poverty. > > The data is consistent with all 4. AIT says 1 and 2 are more likely than > 3 or 4 because the statements can be written using fewer bits (168 vs 256). > > Now there is a separate study in Finland (and a few in the US) showing > that 1 is false. Giving money to people has no effect on the rate that they > are arrested or convicted of crimes. Therefore 2 is most likely. > > Now if another study disproves 2, say by randomly choosing to not arrest > people and compare their incomes, then 3 and 4 are equally likely. > > Now it becomes political, because there is no way to experimentally > control for racism or IQ. What normally happens next is that we guess which > one is correct and search for evidence that confirms our beliefs and ignore > evidence that refutes them. Not because we want to be right, but because > human brains have a maximum learning rate of 10 bits per second short term > and 1 bps long term to avoid filling up our 10^9 bit storage capacity. It > takes more bits to store evidence that refutes our beliefs than evidence > that confirms them. > > As a scientist, I should know better. I should be looking for evidence > that refutes my theories, but I know that doing so means I have to > sacrifice learning something else. At 71 my brain is nearly full and > learning slows down, which only makes it harder to be objective or change > my political views. > > Every political issue has evidence supporting both sides because evidence > is not proof. There is more evidence that the world is round than flat, but > you cannot prove that to a Flat Earther. > > LLMs learn a million times faster than humans, so I expect them to > consider all the evidence in their answers. > > -- Matt Mahoney, [email protected] > > > *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/T0ea44555ed99e6e4-M6493446a6406a2263ee51104> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T0ea44555ed99e6e4-Me47a98756b4366d736dbec3d Delivery options: https://agi.topicbox.com/groups/agi/subscription
