Matt do another paper and find a way to refer work that goes without publishing papers and books, such as the Senator's.
On 02.08.2019 05:53, Matt Mahoney wrote:
The obvious application of AGI is automating $80 trillion per year that we have to pay people for work that machines aren't smart enough to do. That means solving hard problems in language, vision, robotics, art, and modeling human behavior. I listed the requirements in more detail in my paper. The solution is going to require decades of global effort. The best that individuals can do is make small steps towards a solution. http://mattmahoney.net/costofai.pdf On Thu, Aug 1, 2019, 9:14 PM Mohammadreza Alidoust <[email protected] <mailto:[email protected]>> wrote: Thank you. I really enjoy and appreciate your comments. There is no universal problem solver. So for the purpose of building a real AGI, how many problems should our model be able to solve? How big is our problem space? On Thu, Aug 1, 2019, 8:22 AM Matt Mahoney <[email protected] <mailto:[email protected]>> wrote: The human brain cannot solve every problem. There is no requirement for AGI to do so either. Hutter and Legg proved that there is no such thing as a universal problem solver or predictor. It feels like you could solve any problem given enough effort, but that is an illusion. In reality you can't read a 20 digit number and recite it back. The human brain is good at solving problems that improve reproductive fitness, and that's only because it is very complex with thousands of specialized structures and a billion bits of inherited knowledge. On Wed, Jul 31, 2019, 10:58 PM Mohammadreza Alidoust <[email protected] <mailto:[email protected]>> wrote: I may not call the model "a reinforcement learning neural network", because nothing is going to be reinforced there. I would rather call it "model based decision making" where the model of the world will be incrementally completed and more accurate, which then helps in better decision making. The model is in its early stages and must be tested in heavier tasks like the ones you mentioned. However, I believe that AGI is an infinite problem-space and a real AGI must be able to solve everything. This requires further implementations, modifications, time, teamwork, financial support, etc. On Thu, Aug 1, 2019 at 1:34 AM Matt Mahoney <[email protected] <mailto:[email protected]>> wrote: Not understanding the math is the reader's problem. It is necessary to describe the theory and the experiments and shouldn't be omitted. The paper describes 3 phases of training a reinforcement learning neural network. The first phase is experimenting with random actions. The next two phases choose the action estimated to maximize reward. They differ in that they use explicit and then implicit memory, although the paper didn't explain these or other details of the learner. I like that the paper has an experimental results section, which most papers on AGI lack. But I think calling it a "AGI brain" is a stretch. It learns in highly abstract models of chemical manufacturing or cattle grazing. It doesn't demonstrate actual AGI or solve any major components like language or vision. On Wed, Jul 31, 2019, 8:01 AM Manuel Korfmann <[email protected] <mailto:[email protected]>> wrote: I guess he meant: It’s difficult to understand all these mathematical equations. Visualizations are better at transporting ideas in a way that almost everyone can understand easily.On 31. Jul 2019, at 13:46, Mohammadreza Alidoust <[email protected] <mailto:[email protected]>> wrote: Thank you for reading my paper. I wish you success too. Could you please explain more about the readership? I am afraid I did not get the point. Best regards, Mohammadreza Alidoust On Tue, Jul 30, 2019, 2:14 PM Stefan Reich via AGI <[email protected] <mailto:[email protected]>> wrote: If someone paid me to go, I'd go... :-) http://agi-conf.org/2019/wp-content/uploads/2019/07/paper_21.pdf I like the stages you define in your paper (infancy, decision making, expert). Sounds reasonable. I pretty much erased mathematical formulas from my brain though, even though I have studied those things. These days I prefer to think in natural language or code. Increases the readership exponentially too. :-) Many greetings and best wishes to you On Tue, 30 Jul 2019 at 02:13, Mohammadreza Alidoust <[email protected] <mailto:[email protected]>> wrote: Dear Stefan Reich, Thank you. I do not know whether submitting my paper before official publication by Springer is against their copyrights or not. I am not sure about their rules. I will ask the authorities when I arrived Shenzhen and inform you. However I recommend not to miss the AGI-19. http://agi-conf.org/2019/ Best regards, Mohammadreza Alidoust -- Stefan Reich BotCompany.de <http://BotCompany.de> // Java-based operating systems*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/Tf27122c71ce3b240-M89bff8225374adb7c0040aa1>
------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tf27122c71ce3b240-M6b857153eb0384f2869c8764 Delivery options: https://agi.topicbox.com/groups/agi/subscription
