Pick hard problems, because problems that last your whole life, you are not continuously starting and stopping
You only need to look at the things you have not already mastered. I took a speed reading test one time in high school, and read at something like 50,000 words a minute. It was a book on robotics and I only had to glance at pages to see they did nothing new. I answered the questions perfectly to prove I had read it that fast with comprehension. Now, almost 60 years later, I have to read hundreds of papers each week in every conceivable topic (for the Internet Foundation) and most things I know all the units and dimensions, basic models, all of mathematics, all the phenomena – by heart. I only have to identify and bring to mind the pieces already compiled and standardized. A paper of 50 pages, dense mathematics, tables, data, software, groups, history, future plans, economics, many groups, many devices and systems, many languages — a few minutes.
AI groups waste a lot of effort because they use arbitrary tokens and try to process that into workable knowledge. But it is not verified knowledge that each piece is known to work and where it fits in. There are problems that method can NEVER solve. Because the condensed it to a game of moving symbols. What I do it work with digital (mental) twins of all things, and those do fit together. When I got to about 65 my mental view of the world (working long hours and constantly challenging myself to find and process “everything”) clicked into place. All the many things seen frame a personal viewpoint moved into 3D volumetric animations of real things. Much of memory is really compression algorithms. And the symbols used must be easy to remember and bring the correct information to mind. I call it lossless storage and retrieval. Those GPT methods are lossy statistical. They need not be but the groups all cut corners to sell stuff. Their aim is not to share and work with the world.
There are many problems that cannot be represented completely with sequences of words. Because we live in a 3D universe and every voxel has something going on. There are methods for imaging that way, and the limitations of humans can be filled by memory and memory processing systems that use lossless and (lossy + lossless) methods. There are many things I do not know, and I avoid things I simply do not have time for in one life. But I try to find and remember who is working on what, where they store it, and how it is shared. The Internet has about 5.4 Billion human users, but many independent and some continuous machines working. There are 8.2 Billion humans. What I keep in mind constantly now it that at one name per second, it would take about 8.2E9/(365.25*86400) = 259.84 mean solar years to say the name of every living person now. And in that time there will be many more billions. So it is not possible for an unaided human to master much of anything. We live, have some sort of life, and die. It is impossible for unaided or aided humans to record the whole, and compression is difficult aided, and impossible unaided.
The usual habit of most humans is when asked a difficult question, they consider for a few seconds, figure it will take them a few minutes or hours or days, then they give up. “Not my problem, no one pays me to think, I don’t feel like it, I don’t want to”. Many things are considered “impossible” and tagged by media that way. Many things are labeled that way because the tag is applied when it was impossible, and no one removed all the old warnings left in books, paper, peoples minds, the Internet. “Gravity is impossible”. Well actually if you focus on what you want to do, keep track of things, much of that bias comes from when it was “impossible” to calculate and form 3D fields with the right energy density and properties at the right places and times. But magnetic levitation, acoustic levitation, and new true AI based methods allow “anything you can think of, and are willing to work out the details”.
Stop limiting your futures. Anything you have not tried yourself, check widely, work hard, find what is going on, and see what you might do if you devote 50 or 60 years, 7 days a week, 18 hours a day, never forgetting, practicing hard the essentials so they are instantaneous.
It is not hard, just tedious. And eventually true AIs that use lossless and complete memories of all things, verified and supported things — can perhaps give all humans and related species — lives with dignity and purpose.
Filed as (Pick hard problems, because problems will last your whole life, you are not continuously starting and stopping)
Richard Collins, The Internet Foundation
