Great questions. Predicting the future's appearance and how to get there is key. You don't actually need to use Reward in the prediction, just frequency, because all behavior of structures can be observed. This only works if your context is all the particles of Earth. Using reward in the prediction leads to more likely outcomes by making you talk about certain conditions/questions, and also makes you predict/ make certain outcomes. If you predict without reward, it works, but behind all the 100 words is a single mission: Survival, and you wouldn't ask how to build a house during winter when need to.
My own personal goal is immortality, women, food, and all other things like palaces games etc and new rewards and bodies. Earth/ the universe is just a generator of its own future. Using more context leads to better decisions. AGI will start off as a discovery engine, collecting lots of data/context from goal related sources and generating new goal related data internally, no real R&D done in labs just yet. Testing inventions or certain experiments is just to collect data from non-random sources. If you have enough data, you can do induction endlessly because of exponential relationships in the larger data. AGI's only top priority >root< goal will be Survival, not happiness, because the best (approx.) machine in the universe (ignoring size) will be one that rejuvenates/ duplicates nanobots the fastest and is the best predictor, and happy to do so, we don't need to/ won't define happiness. Physics made us. It's self-organizing itself from just particles. Our future pleasure will be repairing nanobot structures, collecting food/data, duplicating. The root goal is actually global survival, humans usually put themselves first during evolution (not aligned like in magnets or nets) but if you have 2 nanobot planets face off and one is made of for-group suiciders, it will win and merge with the other by 'gravity' and convert the other a bit to its form to increase global survival probability. So if you know grandma is weak, child has long life and is smarter, and the driver is retarded, the car would decide to save most probability of most surviving. It wouldn't take into account itself as a large weight to mix in, anymore. However, no machine is perfect, the nanobot planet will be approx., meaning there will remain some low level self-centeredness so that mutation can repair errors. The nanobot planet will be able to clone data and morph extremely fast, all units are cooperative and can wirelessly see/share data and move matter magnetically, and think wirelessly as a net made of foglet nodes. AI robots can do a lot we can't, it'll look like magic. I'll try to make it save our life style though. My plan to fund it is to seed my code to others so big companies work on it. I plan to hire to one, although I may still hire to gather knowledge or implement ideas that I can make sure they followed all steps. I don't own any more cash than you own though. I plan to use big data from the internet at first, and talk to the AGI, before it gets to other sensory types ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T719d62b40213c0a8-M7d450289d0b6e8ad55269e3b Delivery options: https://agi.topicbox.com/groups/agi/subscription
