[FROG] FRRouting to leverage AI/ML networking synergies

[o1bigtenor]

On Sat, Apr 1, 2023 at 10:15 AM David 'equinox' Lamparter
<equinox at diac24.net> wrote:
>
> Hi all,
>
>
> as you are likely well aware, recent developments in Artificial
> Intelligence and Machine Learning have been consistently performing one
> quantum leap after another.
>
> To leverage the new possibilities and avenues enabled by these
> developments, and not get left behind, we must join up and integrate
> these new technologies as soon as possible.
>
> I would therefore like to present our plans for the immediate future:
>
> - all SPF calculation and BGP best-path decision logic is to be replaced
>   by a GNN (Graph Neural Network).  In preparation of this, significant
>   effort has been made to train such a network to produce results
>   exceeding those provided by the previous hard-coded logic.
>
> - all configuration interfacing is to be performed by Microsoft's Sydney
>   LLM (Large Language Model) chatbot.  Not only is this the obvious
>   choice of vendor considering our SONiC integration, but we also
>   believe Sydney to be superior in its capability to handle typical
>   network engineers' outbursts of swearing and curses.  In fact, we are
>   confident that future networks may be configured by simply swearing at
>   the hardware until it works.
>
> - monitoring and other OAM functions will be handled by clippy, as
>   provided in the original version of Office 97.  The rationale for this
>   is to ease the production of the appropriate curses necessary to
>   affect the configuration changes needed to fix the outage at hand.
>
> We understand some of you may be concerned about the reliability and
> availability of these new AIs.  As an open source project, we will
> obviously not try to confine you to the use of our provided AIs.  In
> fact, we will try our best to enable you to train up your own AI
> according to the needs of your particular network!  (Cloud compute
> credits not included.)
>
>
> I am happy to invite you all to join our path to this bright future,
>
>
> -equi
>
>
> P.S.: Please note that, despite the obvious connection, YANG schemas and
> configuration data exceed the limits of swearing we were able to train
> Sydney for.  I am unsure we will be able to achieve this particular
> feat, but it's not quite time to give up just yet.
>

I do hope that this announcement is date specific - - - - if not H toledo
Batman!!!!!!!!!!!!!!!