I don't see a reason to separate training when we evaluate AI intelligence.
557 karma · joined February 15, 2016
Genetic algorithms are smart enough to make life. It seems like genetic algorithms don't care how complex a task is since it doesn't have to "understand" how its solutions work. But it also can't make predictions. It just has to run experiments and see the results.
“We’re not at the point where we’re outperforming wild-type tobacco, but we’re on the right trajectory,” Gunn said. “We only need fairly modest improvements to Rubisco performance, because even a very small increase over a whole growing season can lead to massive changes in plant growth and yield, and the potential applications span many sectors: higher agricultural production; more efficient and affordable biofuel production; carbon sequestration approaches; and artificial energy possibilities.”
All I see is the normal dampening and dragging effects that I show in my physics classroom.
I should say this is true for the physics of sliding friction forces, which aren't super precise.
I use the site for my own students when I teach physics, and they catch a few typos and bugs each semester.
I shared my attempt at this on Hacker News a few years ago.
I wrote my own free open source browser game. https://landgreen.github.io/sidescroller/
"N-gon is a physics-based single player 2-d platformer shooter with hundreds of different power ups. You move with the keyboard and aim with the mouse. The core game is done, but I'm still adding new content.
You can play here for free. It's open source, no ads, no tracking, no data harvesting, no crypto-mining, no DLCs, no freemium, no microtransactions, no gacha."
Often it's just an equation that fits some data for a range of values.
Almost all laws are found in the physical sciences. Some examples are: Ohm's Law, Universal Gravitation, Coulomb's law, and Kirchhoff's laws. None of these are perfect descriptions of reality.
https://landgreen.github.io/sidescroller/index.html
It's also free, open source, no ads, no micro-transactions, web based, no freemium, no data harvesting, no gacha, no crypto harvesting.
Quantum field theory describes reality with wave functions. In this theory there is a field for each type of fundamental particle spread across all of space and time. Regions of a field can evolve into a localized excited state. Particles are a useful mathematical approximation of these excited states.
For example, an electron is a region where the electron field has more amplitude, or bigger ripples. The electron field exists everywhere, but electrons are more likely to be measured where the field amplitude is larger.
Fields interact with other fields, and with themselves. These interactions can be approximated with a Feynman diagram, which treats everything as particles.
Using Feynman diagrams electron field interactions generally model well as particles. This is because the most complex electron interactions are weak enough to be mathematically canceled out. Quarks, the particles inside protons and neutrons, interact through the "strong force". This force has frequent complex interactions that don't cancel out, making the particle model less useful for calculations.