Jump to content

Talk:Sentient (intelligence analysis system)

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

"Sentient Enterprise"

[edit]

Given the NRO in 2010 formally called it "Sentient Enterprise", we have some more sources. I have another potential set here (not all will be relevant, was just a quick assemblage there).

Still looking for more. -- Very Polite Person (talk) 22:42, 6 June 2024 (UTC)[reply]

Does anyone know what this may be in terms of material? Found here, and references "Page n346 U.S. Department of Defense, National Reconnaissance Office. Sentient Program. REL to USA, FVEY. DECL ON 25X1, 20670112. INCG 1.0, February 13, 2012." -- Very Polite Person (talk) 23:42, 6 June 2024 (UTC)[reply]

Article rating and GAN

[edit]

I originally saw this article through the talk page of WikiProject Espionage, and I noticed that this article had been changed to B-class and is now a GAN. First off, thank you @Very Polite Person; this is all great work you've done.

Though, I'm not sure about B-class or GAN. The article could certainly use more citations, as it uses the many of same citations multiple times, keeping the citation count relatively low, but the article is quite wordy, and I'm definitely going to go in to fix a lot of the things I think could be better. Normally, I would just change the rating, but since the rating was added so newly, I wanted to bring it up on the talk page first, especially since @Hawkeye7, who rated it, has so much experience.

I think that C-class is a better fit for this article, but I'd like to hear your thoughts, or have a discussion beforehand. Thanks,NeuropolTalk 12:49, 7 June 2024 (UTC)[reply]

Hi, thanks! I can't admit to expertise on the ratings thing. At a glance... halfway between C and B today? C.8/10? What are you thinking needs changing? -- Very Polite Person (talk) 13:18, 7 June 2024 (UTC)[reply]
Assessment is based on project-wide B-class criteria. See Wikipedia:Content assessment/B-Class criteria for details. Hawkeye7 (discuss) 19:45, 7 June 2024 (UTC)[reply]

Article sourcing

[edit]

This article has a lot of good information, but the sourcing is far too heavily reliant on primary sourcing (budget justifications, documents from the National Reconnaisance Office, and so on). I realize it's going to be hard to find a lot of secondary info when the program is mostly classified, but as it stands we're essentially aggregating a bunch of quotes here, which isn't really what WP is for. I put up the primary sources template because I think we might need to winnow our quotes to rely mostly on the few secondary sources that are given. WeirdNAnnoyed (talk) 14:35, 23 June 2024 (UTC)[reply]

@Very Polite Person: You nominated this article for GAN a week before this complaint was lodged. You've had some time to address it, but nothing has been done. I will quick fail this article if I don't hear from you soon. Viriditas (talk) 22:37, 29 July 2024 (UTC)[reply]
Also, I am a bit concerned by your comments up above. The day before you nominated this article at GAN, you described it as a "first draft". I don't think that qualifies for GA. Viriditas (talk) 22:54, 29 July 2024 (UTC)[reply]
I was sidetracked by 'real life' and other projects. I'll swing back to this later. Thanks for your time and sorry. -- Very Polite Person (talk) 18:25, 31 August 2024 (UTC)[reply]
Ping me again when you nominate it and I will try again. Viriditas (talk) 20:25, 31 August 2024 (UTC)[reply]

GA Review

[edit]
Unsuccessful. Viriditas (talk) 01:22, 31 July 2024 (UTC)[reply]

The following discussion is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

GA toolbox
Reviewing
This review is transcluded from Talk:Sentient (intelligence analysis system)/GA1. The edit link for this section can be used to add comments to the review.

Nominator: Very Polite Person (talk · contribs) 03:51, 7 June 2024 (UTC)[reply]

Reviewer: Viriditas (talk · contribs) 01:06, 31 July 2024 (UTC)[reply]


Criteria

[edit]

GA review – see WP:WIAGA for criteria

  1. Is it well written?
    A. The prose is clear and concise, and the spelling and grammar are correct:
    B. It complies with the manual of style guidelines for lead sections, layout, words to watch, fiction, and list incorporation:
  2. Is it verifiable with no original research, as shown by a source spot-check?
    A. It contains a list of all references (sources of information), presented in accordance with the layout style guideline:
    B. Reliable sources are cited inline. All content that could reasonably be challenged, except for plot summaries and that which summarizes cited content elsewhere in the article, must be cited no later than the end of the paragraph (or line if the content is not in prose):
    C. It contains no original research:
    Concerns with primary sources originally raised on the talk page on 23 June 2024, a week after the article was nominated to GAN.
    D. It contains no copyright violations nor plagiarism:
  3. Is it broad in its coverage?
    A. It addresses the main aspects of the topic:
    B. It stays focused on the topic without going into unnecessary detail (see summary style):
  4. Is it neutral?
    It represents viewpoints fairly and without editorial bias, giving due weight to each:
  5. Is it stable?
    It does not change significantly from day to day because of an ongoing edit war or content dispute:
  6. Is it illustrated, if possible, by images?
    A. Images are tagged with their copyright status, and valid non-free use rationales are provided for non-free content:
    B. Images are relevant to the topic, and have suitable captions:
  7. Overall:
    Pass or Fail:
    Concerns were raised by other editors on the talk page one day before this article was nominated at GAN. Just before that time, nominator admitted that this is a first draft, which makes it clear that it wasn't ready for GAN at the time of nominiation. After that time, other users lodged complaints about the use of primary sources. That was over a month ago with no sign of the nominator, who disappeared on 23 June. I am quick failing this article for those reasons. Viriditas (talk) 01:21, 31 July 2024 (UTC)[reply]
The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

Renominated for GA

[edit]

100% of known/found sources (over a year+) -- considered, integrated or discarded for various policy reasons. More edits than I want to admit. Article was 2,606 bytes before my first edit. Peaked at 40,599 bytes to get everything that was workable in place to refine, and now down to 29,662 bytes.

That cut 10,937 bytes, or four entire copies of the original article as I found it. There's no more sources I know of (yet)--this is about as refined as I can get it. Anything else is probably polish if there's any. -- Very Polite Person (talk) 22:54, 17 May 2025 (UTC)[reply]

Re: This article

[edit]

Saw a post a while ago on the WikiProject Mass Surveillance talk asking for advice on improvements. I think this article would greatly benefit from an explanation of exactly how AI is used -- it currently reads like a press release where AI is used as a buzzword, but even what kind of AI is completely unclear. Mrfoogles (talk) 17:47, 25 May 2025 (UTC)[reply]

I would be exceptionally greatful for any additional sourcing along those lines. If it exists, anywhere, I've yet to find it. At best we have logical inference for the reader based on standard computer science and related data analysis concepts, intersected with intelligence concepts, which are all public. It's not terribly difficult to put 2+2 on it. After over a year, I've just never found any WP:RS that goes that deep. That's undoubtedly among the rather more classified information. -- Very Polite Person (talk) 17:55, 25 May 2025 (UTC)[reply]
Use what you have. Start with the future ground architecture slide and use your own words to explain the info depicted. That’s a good start. You can use that as a starting point. Filling in the gaps from there in a general way should be easy. Viriditas (talk) 19:30, 25 May 2025 (UTC)[reply]
...it didn't even occur to me to use the slide image as a source. You rock. -- Very Polite Person (talk) 20:21, 25 May 2025 (UTC)[reply]

GA review

[edit]
GA toolbox
Reviewing
This review is transcluded from Talk:Sentient (intelligence analysis system)/GA2. The edit link for this section can be used to add comments to the review.

Nominator: Very Polite Person (talk · contribs) 21:45, 17 May 2025 (UTC)[reply]

Reviewer: Viriditas (talk · contribs) 20:58, 11 June 2025 (UTC)[reply]


Review

[edit]

GA review – see WP:WIAGA for criteria

  1. Is it well written?
    A. The prose is clear and concise, and the spelling and grammar are correct:
    Reads well.
    B. It complies with the manual of style guidelines for lead sections, layout, words to watch, fiction, and list incorporation:
    Looks good, but I have questions below.
  2. Is it verifiable with no original research, as shown by a source spot-check?
    A. It contains a list of all references (sources of information), presented in accordance with the layout style guideline:
    Spot check in progress. Questions in feedback section
    B. Reliable sources are cited inline. All content that could reasonably be challenged, except for plot summaries and that which summarizes cited content elsewhere in the article, must be cited no later than the end of the paragraph (or line if the content is not in prose):
    Cited.
    C. It contains no original research:
    Citation 11 content borders on OR, although it is difficult to know. See feedback section.
    D. It contains no copyright violations nor plagiarism:
    No plagiarism detected.
  3. Is it broad in its coverage?
    A. It addresses the main aspects of the topic:
    Looks good.
    B. It stays focused on the topic without going into unnecessary detail (see summary style):
    Looks good
  4. Is it neutral?
    It represents viewpoints fairly and without editorial bias, giving due weight to each:
    Neutral, although I will address this in more detail later below.
  5. Is it stable?
    It does not change significantly from day to day because of an ongoing edit war or content dispute:
    Stable.
  6. Is it illustrated, if possible, by images?
    A. Images are tagged with their copyright status, and valid non-free use rationales are provided for non-free content:
    Rationales are good.
    B. Images are relevant to the topic, and have suitable captions:
    "NROL-76, the only disclosed Sentient mission" and "A portion of a presentation by DNRO Sapp at GEOINT Symposium 2016" are not complete sentences, so they don't need final periods
  7. Overall:
    Pass or Fail:
    Placing on hold due to issues found so far. Will need further discussion. Only issues are around the history section.

Feedback

[edit]
Lead
  • Lead reads well, but I wonder if it truly summarizes the main points in the article. Please revisit.
Thoughts? I think that gets all the key points now? -- Very Polite Person (talk) 19:04, 12 June 2025 (UTC)[reply]
Better, but add a sentence or two about historical timeline, milestones, etc. Lead should say when it began (or approximate date), etc. Viriditas (talk) 21:00, 12 June 2025 (UTC)[reply]
I think (maybe) this is the tightest or at least best lede with this article so far, updated again. I'll wait on your feeback (and below) and hop over to other things a bit. Thanks again!! -- Very Polite Person (talk) 17:09, 13 June 2025 (UTC)[reply]
History
  • As a heavily classified program, public details on Sentient’s architecture and operations remain limited.
  • Public records indicate that Sentient's development program began in 2009, as highlighted by the Federation of American Scientists (FAS).
    • Can you briefly explain how you reached this conclusion from the cited source? I'm not saying it's wrong, I'm saying it is opaque.
I think this was another primary sources casualty but here it is for the sequence:
  • This source archive of Verge first says, "Research related to Sentient has been going on since at least October 2010, when the agency posted a request for Sentient Enterprise white papers. A presentation says the program achieved its first R&D milestone in 2013, but details about what that milestone actually was remain redacted."
  • I had those linked documents here in this version back in 2024 in this version of the article in this passage: "A later declassified May 2009 report to the Congress, "FY 2010 Congressional Budget Justification, Volume IV," contains details about the National Reconnaissance Offices plans for real-time and updated satellite signals intelligence, providing context on NROs space-based missions and programs to collect data, such as Sentient, which would initially begin soliciting defense and related industry feedback in 2010.[9][6]"
  • Live article version today says: "Public records indicate that Sentient’s development program began in 2009, as highlighted by the Federation of American Scientists (FAS).[6]" -- the ref there goes to this FAS material which IIRC is what led me to that RFI link. That was the genesis of the 2009 reference.
  • Using FAS here was an equivalent WP:RS secondary for 2009, matching the RFI, while skipping again using the RFI as a primary source. It seemed like an easy way to establish the start position there.
Is that synth? Do I need to go back to the RFI as the earliest reference ultra overtly in 2009? -- Very Polite Person (talk) 17:59, 12 June 2025 (UTC)[reply]
  • As reported by Sarah Scoles in The Verge, research and development of Sentient began as early as October 2010, managed out of the NRO's AS&T.
    • Doesn't this contradict the above?
No, but it maybe makes it needlessly confusing, if you look at Scoles here:
  • "Research related to Sentient has been going on since at least October 2010, when the agency posted for Sentient Enterprise white papers. A presentation says the program achieved its first R&D milestone in 2013, but details about what that milestone actually was remain redacted."
That's the same passage as above, where she links out to the RFI. How does this look here in this edit to try and unify all this in a simpler and easier to understand manner? -- Very Polite Person (talk) 18:09, 12 June 2025 (UTC)[reply]
@Very Polite Person: I'm still not getting it. It can't begin in 2009 and 2010, so there are some words missing. Do you mean to say that it was budgeted in 2009 and began development in 2010? Whatever the case, you still need to fix this. Viriditas (talk) 18:46, 16 June 2025 (UTC)[reply]
I think I figured it out. Another relic of my primary source purging... -- Very Polite Person (talk) 19:00, 16 June 2025 (UTC)[reply]
  • In 2016, the NRO's Principal Deputy Director (PDDNRO) Frank Calvelli briefed the House Armed Services Committee (HASC) on Sentient.
  • The American Nuclear Society published that the annual budget of the Sentient program at the time was $238,000,000 USD per year in the 2015–2017 period.
  • NROL-76, also known as USA-276, was a May 2017 Falcon 9 Full Thrust launch deployed from Cape Canaveral Space Force Station conducted by SpaceX, and is the only reported to the media NRO and Sentient program–related orbital launch and satellite deployment mission.
    • As mentioned below in spot-check, it isn't clear if this is covered by the cited source. You may need to reword for source text parity. Also the wording is muddled: "the only reported to the media" part doesn't work for me and is way too informal and breezy. "It is the only NRO and Sentient-related orbital launch and satellite deployment mission reported to the media" is slightly better, but I don't see that in the source.
  • At the 39th Space Symposium in April 2024, PDDNRO Troy Meink announced plans to field a mix of large and small satellites to increase satellite revisit times, thereby improving global coverage and enhancing resilience against emerging threats.
    • Probably okay, but I don't like the corporate-government jargon/speak. Can it be rewritten for our readers? It's probably fine as it is, but not my fave.
What do you think of this edit? -- Very Polite Person (talk) 18:16, 12 June 2025 (UTC)[reply]
  • DNRO Sapp stated that the NRO has been asked to give more demonstrations of Sentient and its capabilities than "any other capability since the beginning of the organization's history," in 1959.
Features
  • Well written, but I wonder if the technical aspects can be expanded with general descriptions to flesh out the jargon for our general readers. Also, while I was reading it, I was picturing examples in my head (I tend to do that, unlike other people). Is it possible to provide examples based on the sources, or do the source fail to do that?
I don't think we can get it out of the present sources too far for the deeper CS/intel cycle stuff--the users will just have to try and keep up, but the terms are kind of straightforward (or as close as possible, I guess). For the tipping and queing, Scoles helped on this edit and the iceye has a lot deeper dive on definition. The rest is just a ton of the complex workflow and there's probably no easier way to get into it without OR and SYNTH--if we exempted those rules, I could make it stupidly clear, but alas... -- Very Polite Person (talk) 18:33, 12 June 2025 (UTC)[reply]
Coverage
  • Prose: Not seeing any issues here.
    • Prose is good but the structure and layout could be improved. Taking a step back, I think "coverage" is a very lazy way of approaching this. You're basically duplicating features and capabilities here and there are much better ways to present this info. Working on it now.

Examples:

Data sources
Andrew Krepinevich details the commercial providers contracted to fuel Sentient’s analytics—namely Maxar Technologies, Planet, and BlackSky.[1] Maxar reports it supplies "90 percent of the foundational geospatial intelligence used by the US government."[2] In The Fragile Dictator: Counterintelligence Pathologies in Authoritarian States, Wege and Mobley compare Sentient to Spaceflight Industries’ commercial Blacksky Global service.[3] According to Krepinevich, BlackSky "hoovers up" volumes of raw collateral—dozens of satellites, over a hundred million mobile devices, plus ships, planes, social networks, and environmental sensors—to feed Sentient’s big‑data pipelines.[1] Retired Central Intelligence Agency (CIA) analyst Allen Thomson observes that the system aspires to ingest "everything," from imagery to financial records to weather data and more.[4]
Risks
Andrew Krepinevich warns of the "avalanche" of data available from intelligence, military, and commercial sources that would overwhelm human analysts.[1] Army Captain Anjanay Kumar warned in 2021 that although the system itself is secure, its distributed ground infrastructure could be vulnerable to adversary attack.[5]
  • @Very Polite Person: How do you feel about considering a possible restructuring of the layout with more specific sections?
See also
References
  • Random spot-check
    • 7: ✓ Question: Why is Sentient referred to as the "Sentient Enterprise Program" in the cited source, but not anywhere in the current article?
    • 14: ✓ Formatting of cite is off. It should read NRO, not federal government of the US.
    • 3: I realize it is common to use sources this way, but I don't like it. How am I supposed to know which part of the sentence citations 2 and 4 refer to at a glance?
      • Tweaked here, but I need to come back to the AFRL bit. -- Very Polite Person (talk) 16:24, 12 June 2025 (UTC)[reply]
      • @Viriditas: I think I had picked up the AFRL, Wright Patterson connection from pages 215-217 earlier (I think much earlier) in drafting this when some other PDFs/sources led to me to this: https://www.nro.gov/Portals/65/documents/foia/docs/CBJ/cbj-3.PDF (which I think was part of a huge primary sources nuke someone else did last year). The AFRL, DOE labs and Wright-Pat reference is on 217, 215-217 for complete context. In hindsight it doesn't overtly say "Sentient" but has (bottom left page 217) direct connections to "Advanced Futures Lab ground processing and data fusion technologies" and "NRO advaned technology programs in partnership with the Air Force Research Laboratory and the Department of Energy's National Laboratories." That IIRC had come from me searching non-WP:RS as I often do, because they'll often link to or have various terms to search for against RS. There was other stuff about Sentient-related stuff in Wright-Pat and similar but that went back, I think, to Blackvault based PDFs, which I guess we can't even use for primary sourcing because of the domain. Does this end up OR then in hindsight? I think this was a case where the sources got muddled after I tried to clean up that one edit nuke, my own purging of primaries later, and not being as experienced then. -- Very Polite Person (talk) 17:02, 12 June 2025 (UTC)[reply]
    • 18: ✓
    • 22: ✗ 404. You will need to re-archive your sources.
    • 11: ✗ Are you sure this content is supported by this one source? Are you referring to other sources?

References

  1. ^ a b c Cite error: The named reference Krepinevich Sentient 2023 was invoked but never defined (see the help page).
  2. ^ Cite error: The named reference Steele Logic Spring 2022 was invoked but never defined (see the help page).
  3. ^ Cite error: The named reference Wege Mobley Fragile 2023 was invoked but never defined (see the help page).
  4. ^ Cite error: The named reference Scoles Verge July 31 2019 was invoked but never defined (see the help page).
  5. ^ Cite error: The named reference Kumar Sentient Army 2021 was invoked but never defined (see the help page).
Captions
  • @Very Polite Person: Because "A portion of a presentation by DNRO Sapp at GEOINT Symposium 2016." is a sentence fragment, it doesn't need punctuation unless you change it to "A portion of a presentation by DNRO Sapp was shown at the GEOINT Symposium 2016."
Other
  • I've asked User:Mrfoogles to join this review given their previous input on this subject.
    • Will ping User:Tryptofish as well, in the event they see any issues I may have missed.
Mrfoogles (copied from user talk)
  • Looks a lot better in the Features section, I think. I'm inferring it uses AI to detect the unusual patterns or phenomena, and maybe to integrate different modalities of information? I think it should explicitly say where it uses machine learning (if it is known of course, which it may not be), because that would be useful foe understanding how it operates. And maybe a brief mention of where the AI shows up in the system in the lead would be quite useful. Feel free to copy this to the GA review, tech for commenting there is not working right now for me. Mrfoogles (talk) 01:28, 13 June 2025 (UTC)
  • @Mrfoogles: -- a tweak to lede might work as it would be non-controversial but it would be inference as the data for that aspect is likely terrestrially processed. But there is no source for this unfortunately. What we have in the article, unless I've still yet to miss something super deep and buried (and I followed even citations in other materials to dead ends)... is the totality of sourcing right now except for around another 15-20~ pages of post-FOIA primary sourcing. None of them get into this. At one point in the article history (when it was closer to 40k size) I had every source I could scrounge up, even some maybe borderline, in here, to just see what I had to work with. The stripped down version is basically it as of June 2025. The last time I went hunting was earlier this year. -- Very Polite Person (talk) 12:51, 17 June 2025 (UTC)[reply]
Yeah, I’m not surprised given it’s a government program. I guess we’ll just have to wait. If there’s sourcing it might help to say that “X document” does not specify what the AI is used for, to clarify that we just din’t know and prime the user to try to infer, and clarify it’s not down later in the article, maybe. That said I want to clarify this is definitely not required for GA, I think, just my comments. Mrfoogles (talk) 15:11, 17 June 2025 (UTC)[reply]
I honestly don't think I could source that even. I've put what we have through the wringer repeatedly and even launched plain text copies of sources (to be fair--my OCR skills suck) into outside tools and LLMs to see if I missed anything. We are really at the limit of sourcing here.
That reminds me, I was meaning to export every URL and source for every version on all the articles I've worked heavily on, and write a tool to extract every outside URL from each, to compile them on talk page. "Every source or outside URL that has EVER appeared" in this article, basically. I'd begun thinking we should look into bot automation to do this by default as a log or even native function. Like, a page at every article equivalent to Sentient (intelligence analysis system)/sourceslog that would automatically update in a simple date > user > added > URL/source > diff link, shorted oldest>newest. That way we'd never lose that external history being transparent. -- Very Polite Person (talk) 15:26, 17 June 2025 (UTC)[reply]
Retrieved from "https://en.wikipedia.org/w/index.php?title=Talk:Sentient_(intelligence_analysis_system)&oldid=1295150628"