I plan to start a new SW engineering SIG. I don't have a planned
scope for this SIG, however, I do have a synopsis of what it will
likely look like:
- A place for SW engineers to share opinions and discuss SW
practices (good or bad)
- Detailed wiki
- Extensive documentation of various programming languages on Forge
- Large amount of focus placed on GPU programming, as well as CPU
programming
- Various software repositories made by the community on Forge
I don't know if this SIG will catch on, but I may as well try. I
originally wanted to start an AI/GPU engineering SIG in order to
address the "Fedora AI Developer Desktop" initiative, however, I
surmised that the AI/GPU Eng. SIG would end up copying Ollama if it
were to go entirely off of community consensus, which likely isn't
a productive use of time.
I'll probably keep AI usage limited, however, it won't be
entirely forbidden. Just note that having an extremely large volume
of code for little reason, will not suffice in this SIG.
I would like to hear your thoughts.
- C. S. Sushi
---- Extra context below, read at your leisure ----
When most people think of AI, they think of large language
models, which is merely a single type of DNN AI. The hype around
LLMs has overshadowed the various other kinds of AI, which may be
better suited towards certain applications.
A good example of a "better suited AI," would be the PID
controller. These can be used to control industrial machines, or
be used within fly-by-wire flight control systems. They work *very*
well for these kinds of applications. LLMs (as far as I know) are
incapable of controlling these systems directly, since there must
be a PID controller in the loop for them to talk to. PID
controllers have been used in control systems for several decades.
Convolutional neural networks, for instance, are good for
recognizing images, and handwritten text, and have been used for
several decades up to this point. LLMs are incapable (or would be
very inefficient) of being used for these purposes.
There are *various* reasons to not copy Ollama. Simply put,
Ollama has already been built. It would be better time spent to
build systems via these other solutions, when/if they are
requested.
Sent with Proton Mail secure email.
--
_______________________________________________
devel mailing list -- [email protected]
To unsubscribe send an email to [email protected]
Fedora Code of Conduct:
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives:
https://lists.fedoraproject.org/archives/list/[email protected]
Do not reply to spam, report it:
https://forge.fedoraproject.org/infra/tickets/issues/new
