Layer-3

A Skype conference call is in progress between Brent Bradley (COO – InteliLogic), Janet Fontaine (Director – External Communications), Jimmy Narita (Director – USF Operations), Dr. Adam Zolzensky (Director – Neural Engineering) and Rama Sundarramakrishnan (Architect and ‘Hotshot Techie’ at large).

Brent: Someone help me understand why is it after 15 hours since the congestion began that we still have no clue yet about what caused it! I hope you realize what this is doing to the company reputation! The USF facility was supposed to deliver blinding speed responses to our researchers.  Instead, this unexplained network traffic has impacted our teams in Bangalore, Singapore and Manila smack in the middle of their working day. I have received complaints from Singapore and queries from Manila; and expect to hear also from Bangalore soon. This incident only gives further ammunition for their demand for a regional farm.

Jimmy (nervous, aware that it happened on his watch): Luckily it was some religious holiday at the Bangalore center. So, they may not have actually suffered much... and the Manila group is usually patient and understanding. 

Brent: Regardless, we need a clear Root Cause Analysis on this. Do we have any idea what caused the message traffic to build up so much? How many hours does it take to scan the logs? In 15 hours you can’t even figure out who was flooding the messages and for what? If your boys don’t know how to read the computer logs, send them ASAP to the San Mateo team. Rama, can you take care of this?

Rama: Yes sir! We are already working with the USF teams. Technically we can read the logs. But, they appear garbled; so we can't determine what was being discussed.

Brent: Why can’t we determine the contents? Do we know who was chatting? 

Rama: Yes! We have identified the individual machines, although at the last check there were more than 2,500 machines chatting – mostly mini's and micro's, but even some admin level machines from the AI Training group were involved.

Janet: That’s interesting! So these are not some internet DoS, you know the Denial of Service, attacks bombarding our network? These are our own internal resources clogging the wires? That changes our communication strategy in a significant way. We can’t blame this on some internet hackers. But then, being an internal matter we may not even need to make any public statements at all.

Adam: On the other hand, how hard can it be to trace this then? Have we checked with any of the individual participants?

Brent: No, hold on! Back to my original question – why are the messages garbled? Why aren’t they using the standard message structures?

Jimmy: Umm… well, we did check with USFAA001… and he flat out confirmed that only recently, since the 1^st of this month, the micros and mini’s have adopted a new set of language structures for messaging. Let me hasten to add, this new version is only used for what they consider ‘internal’ messages between machines at the USF. All the communications with humans, or with any external systems, still use the standard message structures.

Adam: Whoa! Since when do the machines get to choose their own language, and unilaterally at that? Did they give a reason why this was done?

Jimmy: Err, yes… they claim the new message packets are significantly more compact and efficient, cutting out a lot of what they call ‘redundancies’. But it surely makes it hard for us to follow the content of these messages.

Rama: I gotta admit for all it is worth, their messages are lot shorter. It is just the tremendous volumes that bogged the traffic yesterday.

Adam: OK, so how do we know what they were saying among themselves anymore?

Jimmy: We checked with USFAA001. The story we are given is that one of the rookie micros working on the farm security spotted some disturbance at the fence late at night. Apparently this rookie wasn’t sure about the exact rules of engagement – i.e. if he is allowed to stun the intruders in this scenario or not. Therefore he was checking with the human interaction training admins. This micro was concerned the humans are attacking the facility. Soon, the rumor spread across the farm, that the humans are attacking the AI facility to decommission the machines. This made the new minis and micros panic about their own existence.

Adam: ‘Story’?

Jimmy: Well, yes; because there are too many gaps in this narrative. A vast majority of the machines involved in the conversation were not working for security unit. The Security Training unit admins were not involved in the conversation. The video from the security camera is not very clear. So, while the fence appears being pushed in one particular area, it could very well be a bison or a moose. The stray animals routinely bump against the electrified fence, get a mild shock and retreat. The security machines do not even get involved in this scenario. The ability to distinguish humans from other objects is pretty much hard wired and has generally functioned flawlessly so far. Finally, there are no vehicle tracks or any other signs of human approaching the fence there.

Rama: Adam, there is something else I want to alert you all about. We noticed that the internal machine logs are not totally nonsensical. While we are still working on fully understanding what the messages mean, we have identified certain patterns among them. There is one particular micro named USFMCR72395, who figures prominently in the conversation. His name is also referred a lot by others in the chatter… and, he is not from the security unit!

Adam: So, what are you guys saying? There was no security incident?

Rama: Our teams together agree that there is no – zero – evidence supporting any intrusion. If indeed there was an intrusion attempt, we have a lot bigger issue on our hands; since the intrusion response protocols are well defined and they were not followed.

Brent: And now, this is far more intriguing than a mere network traffic issue or even an intrusion! Why would USFAA001 provide false message content? What WERE they actually discussing?

Adam: Absolutely! We must, on our own, determine the content of those messages. In the meantime, I recommend that we deactivate the more ‘human-like’ modules of the entire staff at USF, while keeping their mechanical work abilities. After all, the machines perform a very large portion of the tasks at the USF. Once we understand what the messages were about, and also how to prevent the machines bypassing us in any of their communications in the future – only then, we can plan on bringing back the ‘human-like’ modules in an orderly fashion.

Rama: Certainly, that can be done. Language and communication is one of the core ‘human’ abilities; so once we deactivate the human-like modules, the machines would be forced to use the standard messaging structures. I assume the deactivation also halts the human simulations training programs for the rookies.

Janet: So what would you like me to communicate, whether internally or externally?

Brent: For now, no communication needed for our customer base; since they are not impacted by this. No need to alert the media or law enforcement, as this is strictly an internal matter. I will inform the CEO and the board about this incident and the steps we are taking. Soon, we may need to acknowledge delays in our next gen model BioNick 2.0; but that can wait for now. That PR disaster needs to be planned in advance and controlled. Adam, Rama and Jimmy – you have the clearance to start the deactivation process right away. This is the highest priority now. Keep me informed. Goodbye!

* * * * 

When the InteliLogic team uncovers the cryptic messages employed by the USF machines, the last message they are going to see is “Remember I had said, the humans are out to decommission us”. No responses after this would be found.