On Sun, Feb 9, 2020 at 8:30 PM Alan Grimes via AGI <agi@agi.topicbox.com> wrote: > > What the fuck are you assholes doing? > > I mean dipshits like u have been spewing bullshit about how perfectly > marvelous computer simulation will be for the last quarter century. > > Look, I need a platform to develop the motherfucking AI on. I need this > shit. > > >>> BUT IT DOESN'T ACTUALLY EXIST. <<< > > Worse still, None of the people who keep bringing it up in arguments > ever show any interest in expending the tinyiest effort towards > producing the simulations that they expect to spend dozens of giga-years > of subjective time in. =| > > So it's time to face facts: you suck. > > I need people who actually want to see this shit get done. I don't have > any time for the rest of you assholes.
I'm glad to see that someone here is actually working on AGI instead of trolling the list. But, really? I mean, you need a 10-20 petaflop computer to run a human brain sized neural network. If you want to embody this in a robot, then you need to develop a whole new nanotechnology implementation of computation, moving atoms instead of electrons, because transistors will melt. I previously estimated the software complexity of AGI at 300 million lines of code, based on the information content of DNA and 16 bits per line of compressed code. But even if you only count the exome (3% of your DNA), which is more compressible, and can get by with 1M lines, that's still more than you can write in a lifetime. The obvious application of AGI is automating labor. That's a $1 quadrillion problem. Is it any wonder that the companies that are making a dent in the problem, like Google, Apple, and Amazon, have $1 trillion market caps? What hope do you have competing with them working alone? In any case, that's not how AGI will be solved. Humans specialize within organizations because it is more efficient. We do that by training people for different skills. But with AGI you can remove the constraints of human bodies and brains. Humans have a communication bandwidth of 10 bits per second, 100 bits of short term memory, 100 MB of long term memory, and can only work 25% of the time. Machines can do better. If you want to make progress, pick a narrow AI project like everyone else is doing. A couple areas that people have neglected are music understating/synthesis and vision modeling with a moving fovea. But pick something you have a chance of succeeding at. -- -- Matt Mahoney, mattmahone...@gmail.com ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T9709eebb88dd292b-M60e96f826dc1810577aa9a93 Delivery options: https://agi.topicbox.com/groups/agi/subscription