'Intelligence does have a single goal; Survival. Broken down there is sub goals, and if it isn't General Intelligence then this is where the top goal is not fulfilled as much. And yes goal updating is needed, for stepping among those sub goals.
There is an ASI. Bigger brains (aka big diverse data) and more hands (big team) lead to bigger system; bigger context, bigger intelligence. The future Earth will be incredibly advanced beyond your view. They will repair wounds and missing data or employee 'workers' incredibly fast and share instantly no pondering and will use wireless. It's not just 'more' data, arms, eyes, and processors, it's better/new brain and abilities i.e. think/move fast, clone memories, download and upload memories, morph body, perfect recall, erase memories etc. As for bias sabotage, if you losslessly compress a big diverse chunk of our data on the internet or a forum, you will see major trends and anomalies, so you won't often predict anomalies but those may be the correct best answer just being invented recently! What to do? You need reward on net nodes to ask the desired questions and then pick the desired answer. Also, most the trends are correct, but some aren't and need research. You and the data may have the wrong answer, so to find the right answer you need to start at the last correct question and make sure the next bridge is fully structured to other beliefs; finding the answer in your brain when you don't have it will be based on the frequency in prediction, the reward, and the connection to many other beliefs. For example you may say "We will get past this stone wall by paying to Buddha to make a hole in the wall" and that is a frequent solution and has reward on the node because religious vocab nodes are rewarded in this brain, but its local optima! You need to go back to the roots and change many structures, frequencies, rewards, etc n the net to get Buddha off this example mind so much. So in short, there is ery few correct answers in your brain when you first start out as a kid, and if you learn 500 discoveries so fast, you probably have 498 trash ones. So you gotta weight them all and go to the roots. If you don't, your prediction using frequency and reward DO work, but it will generate more trash if most your knowledge is trash. So while discovery forming works absolutely, you also need to do restructure which is more intense. They are same thing I think, but restructure type is deep/at the root and uses many connections. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tc67faac3048278cf-M84644cb1c43cef9a71e4c84d Delivery options: https://agi.topicbox.com/groups/agi/subscription
