Is it possible to have an ASI that sucks at a wide range of tasks but is VERY good at some?
Is it possible to have an ASI that rocks at a wide range of task but is VERY weak at them all? What about very good at very many? So you can actually have a brute force algorithm solve all but do VERY poorer on Longer problems (harder) due to search depth. While you can solve 1 hard problem really good and suck at the rest. And we can already solve many easy problems both badly and goodly. But the main thing to know is this: In general all things about the universe go like this: In general most problems are easy, a good batch are medium difficulty, and very few are left that are extremely hard. And this is what the mechanisms of a brain do, each one solves the many easy ones, adding their powers together to do good at most. In general, even with such an ASI, the hardest problem still may get solved 1%, and the lower layer of this imagine a pyramid of problems (lowest layer has most bricks but are easiest bricks) may get the 2 bricks underneath the toppest solved by 4%, and so on. So in general, it is a gradual spectrum, by the top you can't do as well. So when you try to build an intelligent machine, you are aiming for getting as many problems as solved as can, i.e. having the bleeding shade go as high as can up there pyramid, like 99% for layer 1, 98% L2, 97...10 later......... 77....44....22..8....3....2....1 %. Of course for the better evaluation called "looking like a human" and the even better evaluation called "how long can you survive in our universe", to score better on those just means the same thing still, but notice you can get far when you start and get less juice out by the end (rocks easily live long). So Dall-e looks pretty human but lacks a few more parts. Game Engines have Graphics Engines that create unseen scenes. Back in IDK 1980 you "could" have a game that looked super real life, but it is only a static photo for example. To be real life looking, it has to have objects move which causes shadows to change location and causes light sources to change and reflect off new objects and diffuse through fuzzy glass etc, and morphable mesh objects. A Graphics Engine is sort of AI therefore, because it can render New Novel scenes! Simple ANI texture generators do a ok job too. A brain can do an even better job at predicting the rest of a scene based on the context of 1) the previous frames and 2) the Player's controller buttons that were pressed - to follow Your command. Now, a Graphics Engine as we know it takes a 3D mesh and can get the render from that, whereas a brain need not that mesh - unless you want to Control that AI that way. So, you can use a player's commands and mesh to control an image-based dream, including rules that control some meshs's own actions ex. enemy AI commands. Bonus information: You can safely battle and kill AGIs in future video games as long as "they" disappear and don't really die/ feel pain. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T086f1a4e494591a7-M999cc3af422c9ab580d401b7 Delivery options: https://agi.topicbox.com/groups/agi/subscription
