I like this post from you the most. I believe though most time should be spent not training algorithms, but making what we know more simple to grasp and make it into its true form by getting away from ex. Backpropagation. If anyone even cared a tad about AGI, they'd build a complete AGI Guide of what we know that is a short, intuitive read. No where, not even anywhere, will you find half a guide, that easily explains what we know. You'll only find books of 600 pages being sold for cash. And still those books explain only 10% of AGI, in cryptic language, and only an overview of the general idea.
Interesting read, so research takes decades/ years, while making it into a product takes months. Google wants products but research takes longer. OpenAI is too having an identity crisis of turning into a seller than research company. I obviously don't like that. We need to make discoveries faster, then get them working faster, and get them "sold" faster. The 3rd seems easy if you have something better than Google's. The first 2 are the problem, getting something better working. My AGI Guide should help those 2. It allows clear expression of AGI domain and domains in that domain and so on, allowing you to see an idea coming in the dark tunnel and then get close enough to see the details on its face until finally you pop out of the cave on a train and hear "choo choo". A research/ implementation direction gets faster near the end. Same thing. We just need a faster way to understand what we already have. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tee75800a5ab37f5b-M349f10323173a794d751c487 Delivery options: https://agi.topicbox.com/groups/agi/subscription
