I imagine the users of the AGI list have rather sore and bloodshot eyes
from reading all of the dren Jim has been spewing out the last few
days.... My answer? MORE TEXT!!! =P
In his spew, he raised exactly one important issue. That is, what would
it take to successfully pull off an AGI sprint. that is, what would it
take to go from a flash of inspiration about how to solve AGI to a
candidate singularity seed AGI.
The problem is that any one of us could come up with the Missing Link of
AGI. The problem is that, in order for it to do anything, it must be
integrated with all of the other links. This creates an insurmountable
obstacle for all but the brightest, most well-funded researchers because
integrating all of those links involves a many different areas of
knowledge including, programming, database theory, algorithmics, obscure
features of various windowing systems to automate a game or advanced
robotics to get a physical platform working. Other important areas of
knowledge include networking, high performance computing, heterogeneous
computing, etc... A successful AGI project will have to implement a
solution integrating all of those areas and, if successful, even more
knowledge such as psychology, education, development, etc...
The ROS system is a help but it's not complete because it still requires
a robot or simulation to interface with.
What we need is a nearly complete AGI system with a nice shiny "insert
magic here" slot.
While I struggle to pay even my HOA dues, I continue to look at ways to
advance my own AGI project with what I do have. One of those efforts is
the Zero State VR project. It's primary intention is to provide a
platform for developing AGI. My writings about it are merely focusing in
on certain applications that I hope to use to sell the idea. The
primary, first use, is AGI.
What is desperately required is a system that takes care of all of the
technical problems that will bog down a programmer attempting to tackle
the AGI problem so the only thing he needs to spend time on is the AGI
problem.
This is the "zeroth" objective of the TEC project. -- Absolute top
priority.
Now one major question is how much of an environment do you really need
to support the development of AGI... I think Minecraft is at that
threshold because of it's open world nature with the combination of
survival, creativity, and redstone logic, even social interactions where
users or other AGI systems are present. -- social interactions appear to
be important for development. The huge advantage is that it already
exists, it runs extremely well, supports multi-users, so it should be
fairly straightforward to run in a datacenter situation where you can
test multiple agents. I found some packages that look promising for
capturing the input and output. The game will have to be moded to lock
down the settings and to force it to connect only to the specified
server, but that's just about it.
The exact amount of hardware we're talking about is also an issue but
let's just assume it's "As much as I can possibly afford." and move
forward. =P I'm really looking forward to the ARM64 based machines AMD
is slated to put out next year. Also there are a number of products that
can boost total compute power for certain types of code... You just have
to know what kind of code yours is... So the first step is to acquire a
machine that can support a number of those accelerator boards and then
step two is to figure out which ones you need.
Now my dream here is, once again, not to copy the human mind, much less
a specific mind. Instead, what I'm trying to do is develop the framework
and technology to realize a truly distributed, asynchronous,
heterogeneous intelligence architecture and converge it with the
theoretical limits as soon as possible. Then I can concentrate on what
kinds of deviant fun to indulge in. =P
--
To anyone it may concern, I just retired a CISCO RV220W router from my
network because it was crashing constantly and returned a much older
Netgear WGR614 V7 to service. I feel obliged to expend effort to
bad-mouth the RV220W and its brand at this point.
Powers are not rights.
-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com
