These forums have been archived and are now read-only.

The new forums are live and can be found at https://forums.eveonline.com/

Intergalactic Summit

 
  • Topic is locked indefinitely.
 

Break through in sleeper AI research

First post
Author
Previous Topic Next Topic
Alizabeth Vea
Doomheim
#41 - 2014-01-17 11:15:35 UTC  |  Edited by: Alizabeth Vea
For the Jurou cyborg.

Retainer of Lady Newelle and House Sarum.

"Those who step into the light shall be redeemed, the sins of their past cleansed, so that they may know salvation." -Empress Jamyl Sarum I

Virtue. Valor. Victory.
Jurou Yuan
Wolfraam 74
#42 - 2014-01-17 11:27:25 UTC
I'd put another bullet through your face, but I think the only thing that would accomplish is a whistling sound these days.
Solarienne
Hrimdraugar
#43 - 2014-01-17 11:34:28 UTC  |  Edited by: Solarienne
Now, now, let's get back to the topic at hand.

Have you managed to replicate these efforts with other autonomous entities, such as Rogue Drones? Their AI seems more limited, at least in combat scenarios, so I wonder if this would affect them as significantly, or if it is specific to the potentially more advanced elements of the Sleeper tactical intelligence engine conflicting with itself over something it may never have encountered before. If that were the case, the implications would be pretty astounding - it might suggest that the sleeper drones are not intelligent at all, but instead just have a massive knowledge base with a very fast expert system laid over it.

PY-RE Combat Pilot
Erica Dusette
Division 13
#44 - 2014-01-17 11:34:50 UTC
Karynn Denton wrote:
Erica Dusette, your choice of words to Eran suggests a little unhealthy ignorance of the Matari people.

May I suggest you broaden your cultural horizons - for example, you could spend some time doing your "morale officer" thing with some Valklear recruits. I'm sure you'd be well received. Maybe then you'd think twice about being a ******* racist little ****?

Oh and please could you all stop accusing each other of booster use like it's a bad thing? Some of us have livelihoods, y'know.

You know, I think you got me. I do I tend to lash out a little excessively at your people, especially Matari men, because I in fact have a subconscious attraction to them. Which is why after reading your second comment there I can't help but find myself somewhat aroused. Forbidden fruit, and all that.

It's why I live in deep space. Just can't control myself with all this Empire eye-candy you know.

But, since we're talking about ignorance, I guess it wouldn't have occurred to you that back in my day as a former State combat pilot I'd regularly encounter your kind trying to murder me or those who flew with me. We are at war for a number of reasons, among them a significant racial divide, Miss Denton. I find it rather odd that it's completely socially acceptable for me to sit outside an undock in Hek, murdering your unsuspecting relatives in a racist, yet righteous, hail of scourge fury missiles - but Bob forbid I make any verbal racial slurs while doing so as that'd make me a ****, and as we all know that's so much worse than being a murderer.

Anyway, I'll steer well clear of saying anything bad about your boosters at least. Wouldn't want to threaten your livelihood. Maybe you should try doing the "morale officer" thing sometime? You'll find such financial worries will then disappear almost as quick as Matari claims over contested systems.

Jack Miton > you be nice or you're sleeping on the couch again!

Part-Time Wormhole Pirate ツ Full-Time Supermodel

▷ worмнole dιary + cнaracтer вιoѕ ▷ вσss
Ollie Rundle
#45 - 2014-01-17 13:11:54 UTC  |  Edited by: Ollie Rundle
To try and return this thread to the original post's intention:

Dreygun wrote:
I am certainly aware that sleeper drones will target capsaleer drones and that cycling drones is a useful way to mitigate damage across both, but this is an entirely different matter. Once the tactic is initiated the drone vessel is effectively cloaked towards the sleepers. They can get as close as they want for as long as they want and they will never draw a targeting solution. To my knowledge this is not a commonly known phenomenon, if it were I assume the knowledge would be much more wide spread in guides.

The specifics of the technique limit its utility to only certain sites I have not yet attempted this feat in higher class systems and doing so would be very difficult I lost 1 vessel in the process of figuring out the details of what I know so far. I am not sure it could sustain any higher class sites, but I will update you if I can confirm or deny this later.

Sleeper escalation can be done in a C4. I am unsure as to what effect this would have on this technique I have not tried to initiat the targeting crash with any other vessels present though it could be possible to run this technique in a fleet maybe even with escalation I simply don't have the data to confirm or deny this at this time.

Again, very interesting - if only for the implications and conclusions on Sleeper autonomy and networks that can be drawn from it.

Two points of curiosity:

1. Is the tactic you use effective regardless of the drone class being used? Ie, does it occur only with sentry drones or is it equally likely to occur with the use of light / medium / heavy combat drones, logistics drones or even ECM drones?

2. You note that you have lost at least one ship during your ongoing research. Presuming that this didn't occur as a result of simple human error or hostile capsuleer intervention are you able to provide information on how the circumstance around the vessel's loss differed from your usual operating protocols or why you think it failed on this particular occasion?
Dreygun
Alexylva Paradox
#46 - 2014-01-17 13:31:16 UTC
Solarienne wrote:
Now, now, let's get back to the topic at hand.

Have you managed to replicate these efforts with other autonomous entities, such as Rogue Drones? Their AI seems more limited, at least in combat scenarios, so I wonder if this would affect them as significantly, or if it is specific to the potentially more advanced elements of the Sleeper tactical intelligence engine conflicting with itself over something it may never have encountered before. If that were the case, the implications would be pretty astounding - it might suggest that the sleeper drones are not intelligent at all, but instead just have a massive knowledge base with a very fast expert system laid over it.


that is certainly an interesting question. My research efforts have always focused so much on sleeper drones that I had not really thought to attempt the same process on Rogue drones. In theory their much more infantile AI should be equally vulnerable if not more so to these tactics I will attempt to test this theory next time I encounter them.
Dreygun
Alexylva Paradox
#47 - 2014-01-17 13:36:11 UTC
Ollie Rundle wrote:
To try and return this thread to the original post's intention:

Dreygun wrote:
I am certainly aware that sleeper drones will target capsaleer drones and that cycling drones is a useful way to mitigate damage across both, but this is an entirely different matter. Once the tactic is initiated the drone vessel is effectively cloaked towards the sleepers. They can get as close as they want for as long as they want and they will never draw a targeting solution. To my knowledge this is not a commonly known phenomenon, if it were I assume the knowledge would be much more wide spread in guides.

The specifics of the technique limit its utility to only certain sites I have not yet attempted this feat in higher class systems and doing so would be very difficult I lost 1 vessel in the process of figuring out the details of what I know so far. I am not sure it could sustain any higher class sites, but I will update you if I can confirm or deny this later.

Sleeper escalation can be done in a C4. I am unsure as to what effect this would have on this technique I have not tried to initiat the targeting crash with any other vessels present though it could be possible to run this technique in a fleet maybe even with escalation I simply don't have the data to confirm or deny this at this time.

Again, very interesting - if only for the implications and conclusions on Sleeper autonomy and networks that can be drawn from it.

Two points of curiosity:

1. Is the tactic you use effective regardless of the drone class being used? Ie, does it occur only with sentry drones or is it equally likely to occur with the use of light / medium / heavy combat drones, logistics drones or even ECM drones?

2. You note that you have lost at least one ship during your ongoing research. Presuming that this didn't occur as a result of simple human error or hostile capsuleer intervention are you able to provide information on how the circumstance around the vessel's loss differed from your usual operating protocols or why you think it failed on this particular occasion?



Very good points. The specifics of the tactic make the use of any other type of drone largely inefficient because the sleepers will target the drones and anythign not close enough to be quickly withdrawn from space would most certainly be destroyed, however from a purely scientific basis it could be a useful test to further draw out the exact parameters of the targeting failures.

As to my ship loss, a series of specific steps are required to initiate the sleeper targeting crash, which if not done correctly will result in the sleepers proceeding with normal targeting and combat manuvers, Given I was running in a class of wormhole much higher than a BC should be able to run solo in my ship was quickly destroyed when the targeting solution was drawn on me.
Dreygun
Alexylva Paradox
#48 - 2014-01-17 14:23:57 UTC  |  Edited by: Dreygun
As an update, CONCORD officials finally contacted me. They gave no cease and desist order, I was given a short assurance that they would have a research team analyze my ship data, but that the results of their finding would be classified and I may not be contacted again. This cloak and dagger stuff is very unsettling, I must agree with Streya that CONCORD and its associates must know something that we capsuleers have not been deemed worthy worthy to hear.
ISD Tyrozan
ISD Community Communications Liaisons
ISD Alliance
#49 - 2014-01-17 20:22:11 UTC
Post consisting of attack on CCL personnel and discussion of moderation has been removed.

ISD Tyrozan

Captain

Community Communication Liaisons (CCLs)

Interstellar Services Department

@ISDTyrozan | @ISD_CCL
Dreygun
Alexylva Paradox
#50 - 2014-01-30 13:47:45 UTC  |  Edited by: Dreygun
posting error
Dreygun
Alexylva Paradox
#51 - 2014-01-30 13:48:54 UTC
Ollie Rundle wrote:

Again, very interesting - if only for the implications and conclusions on Sleeper autonomy and networks that can be drawn from it.

Two points of curiosity:

1. Is the tactic you use effective regardless of the drone class being used? Ie, does it occur only with sentry drones or is it equally likely to occur with the use of light / medium / heavy combat drones, logistics drones or even ECM drones?

2. You note that you have lost at least one ship during your ongoing research. Presuming that this didn't occur as a result of simple human error or hostile capsuleer intervention are you able to provide information on how the circumstance around the vessel's loss differed from your usual operating protocols or why you think it failed on this particular occasion?


I have been able to complete further testing. To address the first question all drone classes are able to be used once the targeting failure has been initiated. As I said before the use of other drone classes has no apparent benefit because the sleepers will still target the drones.

Further research has shown this tactic can be employed without the use of drones at all. Again specifics are classified, but we are making very interesting strides in understanding the core details of the sleeper targeting AI.

I have also been able to confirm that this tactic is viable against the weaker AI of the rogue drones as well.

Overall, this particular AI failure has been heavily harvested at this point, and they have still been unable to adapt. I will continue to keep you posted if that changes. My research into the various theories surrounding the sleepers suggests that they are not dormant so it is my hypothesis that they will adapt soon. If so it will confirm at least part of the theories.
Eran Mintor
Metropolis Commercial Consortium
#52 - 2014-02-12 20:15:02 UTC
Haven't heard anything about this lately. Have the Sleepers adapted to any of this yet?

-Eran
Fredfredbug4
The Scope
Gallente Federation
#53 - 2014-02-12 22:33:16 UTC
Were there any cosmic anomolies such as high electromagnetic signatures, a nearby supernova, a nearby blackhole or quasar, or other things that could affect drones?

Watch_ Fred Fred Frederation_ and stop [u]cryptozoologist[/u]! Fight against the brutal genocide of fictional creatures across New Eden! Is that a metaphor? Probably not, but the fru-fru- people will sure love it!
Stitcher
School of Applied Knowledge
Caldari State
#54 - 2014-02-13 14:31:25 UTC  |  Edited by: Stitcher
Efforts to duplicate Dreygun's findings have so far met with little to no success. Whether that points to some adaptation on the drones' part or the presence of an unaccounted-for external factor is not yet clear.

We're out here for the long haul and are conducting lots of different kinds of research, some of which will occasionally turn out to be red herrings, false positives or just plain human error. I'm taking this Sleeper AI thing as a reminder about the importance of peer-review and due caution before announcing your findings or even the nature of your research, and I've already suggested to Dreygun that future "findings" should be discussed at length internally before release to the capsuleer community.

Not least because this research is potentially lucrative. Roll

AKA Hambone

Author of The Deathworlders
Dreygun
Alexylva Paradox
#55 - 2014-02-13 14:35:10 UTC
Eran Mintor wrote:
Haven't heard anything about this lately. Have the Sleepers adapted to any of this yet?

-Eran


My most recent research has shown no change at this point. I will keep you posted if this changes. I still believe that my hypothesis will be confirmed, and the sleepers will prove adaptation ability, however I am usually forced to destroy all the sleeper vessels when using this tactic to gather research material. This could very well be making it hard for the details to be transmitted out to other sleepers. Who knows what amount of programing is required to initiate a change to all sleeper AI or for the issue to even be processed by any sentient agents still in existence.
Dreygun
Alexylva Paradox
#56 - 2014-02-13 14:40:01 UTC
Fredfredbug4 wrote:
Were there any cosmic anomolies such as high electromagnetic signatures, a nearby supernova, a nearby blackhole or quasar, or other things that could affect drones?


cosmic anomalies seem to play no role in the AI failure. I will be sure to make a full report on the specifics as soon as the matter is properly declassified.
Dreygun
Alexylva Paradox
#57 - 2014-02-13 17:11:43 UTC
I must redact my prior statement... its incredible, I really didn't it would happen so soon. They did it they adapted to the glitch. I will work on creating a new thread to fully debrief everyone on the specifics. Until then feel free to post any questions here.
Stitcher
School of Applied Knowledge
Caldari State
#58 - 2014-02-13 17:43:20 UTC  |  Edited by: Stitcher
*sigh*

Dreygun, kirjuun, exactly which bit of "It's important to be cautious and avoid leaping to conclusions until your data and findings have been properly peer-reviewed, please keep things internal to the corporation until they've been appropriately vetted for public release." are you having trouble with?

AKA Hambone

Author of The Deathworlders
Nicolas Merovech
Doomheim
#59 - 2014-02-14 20:48:25 UTC
If the Sleeper's internal processes can be tricked by a simple "technique" then perhaps they are far more rudimentary than we thought. A true AI or Infomorphic intelligence is able to adapt to unknown circumstances and function normally. If this report is accurate, then it suggests Sleepers are more in line with combat utility drones, or a virtual intelligence. A fascinating implication, but inconclusive at best. This technique must be tested thoroughly and it's results meticulously recorded.

Dr. Nicolas A. Merovech, Ph. D, M.D.
Dreygun
Alexylva Paradox
#60 - 2014-02-14 21:17:34 UTC
Nicolas Merovech wrote:
If the Sleeper's internal processes can be tricked by a simple "technique" then perhaps they are far more rudimentary than we thought. A true AI or Infomorphic intelligence is able to adapt to unknown circumstances and function normally. If this report is accurate, then it suggests Sleepers are more in line with combat utility drones, or a virtual intelligence. A fascinating implication, but inconclusive at best. This technique must be tested thoroughly and it's results meticulously recorded.



I agree the sleeper drones are most certainly non sentient programed drones, the simplicity of their patterned attacks and reinforcement schedules are fair proof of this. This more impressive implication of this finding is that after a period of time they adapted to small scale glitch in their targeting AI. A glitch that if applied by the large scale capsuleer community would have been certainly devastating to them. This suggests planning, forethought, strategy analysis, and large scale adaptability. They don't adapt to our normal tactics for destroying them but they choose to adapt to a tactic that has only been used by a few isolated instances but if applied wide scale would devastate them. I believe this is very strong evidence of a sentient controller who if not actively controlling the drones is at least monitoring them at intermittent intervals.