The thing that I like about this article is that you are taking the nebulous “AI is going to transform warfare” and you’re operationalizing it in a practical way that I haven’t seen enough people do.
Having just traveled abroad and used google translate to communicate for just about every interaction, there still is a latency to using these tools that doesn’t feel natural. So much so that the temptation is to just start speaking and seek understanding through body language— even accepting that the communication will be lossy, unverified, untimestamped, etc. I have no experience in the battlefield, but my intuition tells me that a parallel processing route might be easier to swallow — pass the raw signal but digitize it for asynchrony comms and validation
Is there an added EM signature risk to rigging all this tactical equipment with this data collection tech? Or does the collection occur on the receiving side, displaced from the front?
You’d need to be thoughtful about the way you went about doing data collection. In my mind the things that are already transmitted for tactical reasons would be logged instantly/automatically, while the rest (ie the audio files used to generate the transcriptions) would be bulk uploaded at opportune times (off the front, connected to starlink/wifi/ethernet).
Oh and in terms of rigging tactical equipment like the physical hardware, this all runs on a smartphone, like the Galaxy S20s currently fielded or (better) the Galaxy S23s being fielded now. My prototype ran on a S21 that was my personal cellphone.
Now of course these need to push information somehow- one solution is commercial internet via cell towers or low earth orbit satellites (starlink). You could also push over tactical radios, but I would argue would need a better solution than the current monstrosity of cables that I address in my first post. But again, while the data still needs to be transmitted, the proposed solution offers significant compression over what is currently being transmitted and therefore a lower corresponding EM signature.
As I say above, higher volume stuff used to improve models etc but not of immediate tactical relevance could be stored on device and uploaded post-mission.
These both make sense. Thank you again for the good explanations. Pushing over tactical radios seems like it’s not worth the cost in soldier training or bandwidth or time.
The thing that I like about this article is that you are taking the nebulous “AI is going to transform warfare” and you’re operationalizing it in a practical way that I haven’t seen enough people do.
More of this, please.
I think the concept of starting with the soldiers on the ground is extremely useful in thinking about how to integrate these systems.
Having just traveled abroad and used google translate to communicate for just about every interaction, there still is a latency to using these tools that doesn’t feel natural. So much so that the temptation is to just start speaking and seek understanding through body language— even accepting that the communication will be lossy, unverified, untimestamped, etc. I have no experience in the battlefield, but my intuition tells me that a parallel processing route might be easier to swallow — pass the raw signal but digitize it for asynchrony comms and validation
Is there an added EM signature risk to rigging all this tactical equipment with this data collection tech? Or does the collection occur on the receiving side, displaced from the front?
You’d need to be thoughtful about the way you went about doing data collection. In my mind the things that are already transmitted for tactical reasons would be logged instantly/automatically, while the rest (ie the audio files used to generate the transcriptions) would be bulk uploaded at opportune times (off the front, connected to starlink/wifi/ethernet).
Oh and in terms of rigging tactical equipment like the physical hardware, this all runs on a smartphone, like the Galaxy S20s currently fielded or (better) the Galaxy S23s being fielded now. My prototype ran on a S21 that was my personal cellphone.
Now of course these need to push information somehow- one solution is commercial internet via cell towers or low earth orbit satellites (starlink). You could also push over tactical radios, but I would argue would need a better solution than the current monstrosity of cables that I address in my first post. But again, while the data still needs to be transmitted, the proposed solution offers significant compression over what is currently being transmitted and therefore a lower corresponding EM signature.
As I say above, higher volume stuff used to improve models etc but not of immediate tactical relevance could be stored on device and uploaded post-mission.
These both make sense. Thank you again for the good explanations. Pushing over tactical radios seems like it’s not worth the cost in soldier training or bandwidth or time.