On Wednesday, July 08, 2020, at 10:30 AM, James Bowery wrote: > The "surprise" simply means bits were added to the corpus of the intelligence. I disagree. What if it already stored the exact phrase "thank you", and heard someone say it? It'd strengthen the connections / update the frequency, which adds bits.
On Wednesday, July 08, 2020, at 10:30 AM, James Bowery wrote: > The less intelligent will either autistically regurgitate the passage or not > remember it at all. The more intelligent will have compressed those bits > into consilience with the rest of its knowledge structure and _generate_ a > more intelligent interpretation of its meaning. I disagree. A brain models its universe by compressing observations, it doesn't store every observation it sees, some are forgotten, some are already stored, some contain parts of ones already stored, some are similar. As a brain matures/ learns more, it will refuse to store anything new that is not predictable based on its wisdom (if it's not recent, popular, similar, or loved), for example dolphins use computers, we know they have no hands, no education, can't can't sit down on chairs, aren't as intelligent, can't live on land, or for example Stem cells are your favorite domain or the fastest way to immortality, but he loves/knows AGI is! You can definitely screw up this brain if you teach it from birth about some given religion frequently, recently, relate it to all things god made this god made that gods looks after this, and make it loved by saying it gives you infinite food and sex and immortality for future rewards, local optima. My point is that a brain picks a path and gets stuck there, mostly crystallized, and won't store "new things". Hence, if presented with Shakespeare's passage, and repeated it in its own words its understanding of it, it may actually parrot it if it already stores it and believe it like OpenAI believes their mission statement :) Ben, so, 4 questions: Is your OpenCog AI able to predict yet? And is it text or image? Do you understanding your AI (white box, clear mechanisms) much more than people understand Transformers or is it like a magic box may pop out answer no know why it works? Obviously generalization looks at hundreds of similar experiences but I that isn't very blackboxy. And once/if it does predict, [what research goal] i.e. prompt/question do you install in it to [talk about all day]? You want to force it to rarely talk about clothing, unicorns, or politics. For example it says all day "I will cure death by...". And do you let it find its hobby/love? That'd let it switch the question/mission/prompt from curing death to say making stem cells or fixing cryonics or improving AGI, then it can pick AGI or a few and dive deeper into a sub domain of AGI, specialization. Doing so just makes it collect data from those sources more often than other domains, but that's exactly what you want. Exploitation. Diverse knowledge does equal generalness but survival our root mission/prompt causes us to have a specific area of that general space to favor, so only some data is more valuable, yes, clothing is not as cool as computers, sorry clothing nuts. But if it needs to make electricity before computers before AGI, yes, it cool then, but it still is focused on survival. Simply AGI picks next best available it can invent, electricity. This could be sticky web like I'm not sure it may need to go on and off small hobbies like need to refill battery so that's all it talks about, or need a measuring tape, or time to work on cryonics if working too long on one path (half the day devoted, or quits one then takes up other full time). But basically hobbies form slowly and are permanent really, sorta....80s car nut will always be an 80s car nut. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T604ad04bc1ba220c-M9c41e3bbfb0cf72148b65d63 Delivery options: https://agi.topicbox.com/groups/agi/subscription
