Just referred nfoservers to Tyson internal engineers

This is used for general discussion that is not necessarily server-related.
Post Reply
User avatar
mennes
This is my homepage
This is my homepage
Posts: 85
https://www.youtube.com/channel/UC40BgXanDqOYoVCYFDSTfHA
Joined: Sat Nov 08, 2014 1:08 pm
Location: Michigan
Contact:

Just referred nfoservers to Tyson internal engineers

Post by mennes »

Hey this is mainly for the nfo team; I started working at Tyson about 4 weeks ago, and I lost John's email thread so I figured id just post here

We are in an interesting place with ai and the type of work we are all familiar with. Last year AWS outage created some huge issues. Companies like NFOServers that fly under the radar, and still do standalone machines are going to be a primary interest of high availability OT systems

With the recent ordeal with Openai and hugging face, i think many OT groups are slowly beginning to grasp some of their security risks. I will say I have been at a few high output companies that I believe still are behind in terms of cybersrcurity.

I think a good way to be able to utilize ai for rapid prototyping/system monitoring (like logging motor amp draw over time) with be using external servers, such as NFOServers. Data diodes and external webservers i think is important, but im not expert.

We live in interesting times, thanks for all the help throughout the years NFO, happy friday
Feel free to contact me on discord @ Mennes34
User avatar
Edge100x
Founder
Founder
Posts: 13244
Joined: Thu Apr 18, 2002 11:04 pm
Location: Seattle
Contact:

Re: Just referred nfoservers to Tyson internal engineers

Post by Edge100x »

Thanks for spreading the word!

Our machines aren't designed with AI in mind but the new ones could potentially be modified. Main limitation with it is power usage, though.
User avatar
mennes
This is my homepage
This is my homepage
Posts: 85
Joined: Sat Nov 08, 2014 1:08 pm
Location: Michigan
Contact:

Re: Just referred nfoservers to Tyson internal engineers

Post by mennes »

I dont really think itd be necessary to run a llm on the server; its more sending OT data through a diode to the NFOServer, then you can more safely let a llm evaluate the data/prototype things based on real data vs even thinking about allowing a llm on the local OT network

I think there is some value to a local llm for more sensitive data, which is something worth looking into (im not an expert in this area)


A lot of OT networks basically have some sorta machine that has direct access to the internet, which isnt unsafe by default, but it's a risk.

Basically, trying to solve some of these data security problems is going to be a main talking point. Similar to the discussion of "how do we keep ai in a box, whilst also being able to use it?:

I think data diodes is a very under discussed topic, but will soon be priority #1, once something actually catastrophic happens like hugging face, but with consequences that a non IT person can digest (for lack of better explanation)
Feel free to contact me on discord @ Mennes34
Post Reply