Steinmetz was a genius. And like all good scientists, when he had an idea, he did research to see if someone else had thought of it first, and if not, he did experiments to see if it worked. As it turns out, even if you are a genius, the vast majority of your ideas will fail one of these two tests. You can come up with hundreds of ideas in an hour. But if you come up with only a few successful ideas in your lifetime, you are doing really good.
My interest in data compression was partly in its competitive nature. Benchmarks give you a precise score, a ranking, and bragging rights. (This might be why the vast majority of data compression developers are male). But I was also interested in its ability to precisely measure the quality of language models when compressing text. Developing the PAQ series of compressors took years. I would come up with dozens or hundreds of ideas per day while coding. Some I thought were really big breakthroughs. Most would either fail to improve compression, or would improve it a tiny amount like 0.001%. So I learned to do the smart thing and not announce any of my brilliant ideas until I could back them up with numbers. Compression started as a hobby. It wasn't until years later that it led to a good paying job working from home doing something I enjoy. And now we have an important advance toward AGI. We know that the best language models use neural networks. On Mon, Nov 4, 2019, 12:52 PM John Rose <[email protected]> wrote: > On Monday, November 04, 2019, at 12:36 PM, rouncer81 wrote: > > Lossylossnessness, total goldmine ill say again. Dont doubt it. :) > > > Picture this - when Charles Proteus Steinmetz proposed using imaginary > numbers for alternating current circuit analysis everyone attacked him and > thought he was coo-coo. Now, I'm definitely not a great mind like him but > I have good intuition. I recently made an effort to find the original > location of his cabin to absorb some zen, it's a few miles from here near > Schenectady, NY hidden. The actual cabin was moved to Michigan (I do > amateur archaeology on weekends). We may have a similar situation with > lossy and lossless. Perhaps imaginary/complex numbers can do it. Or a > similar concept. > *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + delivery > options <https://agi.topicbox.com/groups/agi/subscription> Permalink > <https://agi.topicbox.com/groups/agi/T252d8aea50d6d8f9-M04106d4897e03d1e8d2c31d2> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T252d8aea50d6d8f9-M307e237bae1e860d8d14d2c7 Delivery options: https://agi.topicbox.com/groups/agi/subscription
