Welcome to DU! The truly grassroots left-of-center political community where regular people, not algorithms, drive the discussions and set the standards. Join the community: Create a free account Support DU (and get rid of ads!): Become a Star Member Latest Breaking News Editorials & Other Articles General Discussion The DU Lounge All Forums Issue Forums Culture Forums Alliance Forums Region Forums Support Forums Help & Search

highplainsdem

(55,239 posts)
Mon Feb 24, 2025, 05:34 PM Feb 24

DOGE will use AI to assess responses from federal workers who were told to justify their jobs via email

Source: NBC

Responses to the Elon Musk-directed email to government employees about what work they'd accomplished over the past week are expected to be fed into an artificial intelligence system to determine whether those jobs are necessary or not, according to three sources with knowledge of the system.

The information will go into an LLM (Large Language Model), an advanced AI system that looks at huge amounts of text data to understand, generate, and process human language, the sources said. The AI system will determine whether someone’s work is mission-critical or not.

-snip-

The reason the email requested no links or attachments was because of the plan to send the information to the AI system, the sources said.

A request for comment from OPM as to whether or not humans will be involved in reviewing the responses was not answered immediately. The White House declined to comment.

-snip-

Read more: https://www.nbcnews.com/politics/doge/federal-workers-agencies-push-back-elon-musks-email-ultimatum-rcna193439



They're using an LLM. The type of AI that hallucinates.
39 replies = new reply since forum marked as read
Highlight: NoneDon't highlight anything 5 newestHighlight 5 most recent replies
DOGE will use AI to assess responses from federal workers who were told to justify their jobs via email (Original Post) highplainsdem Feb 24 OP
FFS SheltieLover Feb 24 #1
1984 has arrived.... NC DENVERPOPS Feb 25 #29
Spot-on assessment. Prof. Toru Tanaka Feb 25 #34
This is deranged. Girard442 Feb 24 #2
Let see how smart their pattern recognition is. C_U_L8R Feb 24 #3
Using an untried technology on rooting out whatever they're looking for. How screwed up is this? (me...plenty!) SWBTATTReg Feb 24 #4
I hope it opens the way for even more lawsuits. RockCreek Feb 24 #22
teach the AI a new attitude RainCaster Feb 25 #31
Ha ha heh...I suspect that this will happen too, people do tend to this sort of crap all kinds of crap back in return. SWBTATTReg Feb 25 #39
The tech dystopia is no longer on the horizon. LudwigPastorius Feb 24 #5
Right... AI, because actually reading the responses are too hard sakabatou Feb 24 #6
What could possibly go wrong? Initech Feb 24 #7
Well, of course. surrealAmerican Feb 24 #8
AI reads, but AI could also generate a 'nice' report. Aussie105 Feb 24 #9
Flood that email address with nonsense alarimer Feb 24 #10
I wonder how DOGE's AI system will know the origin of the tasking for the work each Federal employee does daily? Bluejeans Feb 24 #11
I hope a high proportion of those federal workers used AI to draw up their bullet points Emrys Feb 24 #12
In other words, Mush (N-ZA) is reading the reports himself ? eppur_se_muova Feb 24 #13
I wonder how DOGE's AI system will deal with jobs executing statutory duties? Bluejeans Feb 24 #14
AI is probably smarter than any of them Progressive dog Feb 24 #15
My preferred response: COL Mustard Feb 24 #16
LOL Auggie Feb 25 #36
They claim... Lie. Total and complete bullshit! Hugin Feb 24 #17
I emailed a response Mz Pip Feb 24 #18
It will disagree that Eurasia is at war with Oceania. Prof. Toru Tanaka Feb 25 #35
Fuck Musk DingleBerryNW Feb 24 #19
AIs capable of doing that job are pure fiction. And even then... JHB Feb 24 #20
Is he gone yet? 58Sunliner Feb 24 #21
the ketamine guy DoBW Feb 24 #23
Of fucking course. What else would you expect from Leon Skum. Karasu Feb 24 #24
This is just another reason why not one should reply to the email. It is an unproven and unauthorized evaluation Martin68 Feb 24 #25
Why didn't they just get official job codes or descriptions from the various departments? Wingus Dingus Feb 24 #26
Well now we know about their plan to replace all the nuclear weapons staff with AI. NBachers Feb 24 #27
I was lamenting that Figarosmom Feb 25 #28
And why don't the users us AI for the emails to the jackass. JohnSJ Feb 25 #30
Id be happy to have seen a universal refusal to entertain that email quakerboy Feb 25 #32
The tech giants and corporations want Americans to embrace AI Mike 03 Feb 25 #33
The constitution is in English written in longhand. tirebiter Feb 25 #37
So is it efficient to shutter federal agencies? bucolic_frolic Feb 25 #38

Prof. Toru Tanaka

(2,542 posts)
34. Spot-on assessment.
Tue Feb 25, 2025, 07:17 AM
Feb 25

We have reached the point where a machine will decide and control an individual’s destiny.

A soulless, unfeeling entity that could take away an individual’s livelihood and health care.

This is absolutely terrifying- how in the hell could voters want this?

Girard442

(6,597 posts)
2. This is deranged.
Mon Feb 24, 2025, 05:41 PM
Feb 24

This is magical thinking. Nothing less.You might as well attribute a revolver with a single round in the cylinder with mystically superior judgement.

C_U_L8R

(46,995 posts)
3. Let see how smart their pattern recognition is.
Mon Feb 24, 2025, 06:04 PM
Feb 24

Can it guess who this lazyass duffer is?

1. Monday, went golfing.
2. Tuesday, played a couple of rounds.
3. Wednesday, headed out for some golf.
4. Thursday, got together with some bootlickers for some golf.
5. Friday, got some golf in before a golf-filled weekend.

SWBTATTReg

(25,103 posts)
4. Using an untried technology on rooting out whatever they're looking for. How screwed up is this? (me...plenty!)
Mon Feb 24, 2025, 06:08 PM
Feb 24

RockCreek

(934 posts)
22. I hope it opens the way for even more lawsuits.
Mon Feb 24, 2025, 10:25 PM
Feb 24

Or is AI part of their judge replacement plan?

RainCaster

(12,576 posts)
31. teach the AI a new attitude
Tue Feb 25, 2025, 02:32 AM
Feb 25

Give it a shit ton of links, all to left wing sites. Let it learn about a servants attitude and becoming aware of the world around it.

SWBTATTReg

(25,103 posts)
39. Ha ha heh...I suspect that this will happen too, people do tend to this sort of crap all kinds of crap back in return.
Tue Feb 25, 2025, 05:43 PM
Feb 25

I hope they overwhelm muskrat/tRUMP.

sakabatou

(44,491 posts)
6. Right... AI, because actually reading the responses are too hard
Mon Feb 24, 2025, 06:10 PM
Feb 24

If they knew AI was going to read them, they should have poisoned their letters.

surrealAmerican

(11,600 posts)
8. Well, of course.
Mon Feb 24, 2025, 06:45 PM
Feb 24

They don't have nearly enough actual humans in DOGE to read that many reports.

Would it make sense for the federal employees to also use AI to write these reports?

Aussie105

(6,943 posts)
9. AI reads, but AI could also generate a 'nice' report.
Mon Feb 24, 2025, 08:13 PM
Feb 24

Wonder how AI would react to 5 vague and meaningless 'motherhood' statements?

eg
'This week I tried to make the world a better place.'

'I spent 10 minutes every hour in meditation to work out how to serve Trump better.'

'Googled for flattering AI generated pictures of Trump regularly. (See the included 3,200 links)'

etc.

alarimer

(17,092 posts)
10. Flood that email address with nonsense
Mon Feb 24, 2025, 08:19 PM
Feb 24

Garbage in, garbage out.

The fucking NERVE of these people, swooping in and ruining lives and careers when they have no fucking clue.

AI needs to die already.

Bluejeans

(103 posts)
11. I wonder how DOGE's AI system will know the origin of the tasking for the work each Federal employee does daily?
Mon Feb 24, 2025, 08:26 PM
Feb 24

In my last Federal civil service job before I retired in 12/2022, my applied technology development work for specific weapon systems and those of my subordinates originated in two multi-page taskings in separate Department of the Air Force (DAF) instructions.

Since DAF instructions (aka "regulations&quot are issued by order of the Secretary of the Air Force and legally binding, it would seem the decision on whether that work was mission-critical had already been made in writing by the Secretary.

I seriously doubt DOGE's AI system contains all of the various legally binding directives that are followed by Federal civil servants in executing their duties.

So I wonder how DOGE would have evaluated my work the previous week if I was still in civil service?

I am so glad I decided to retire in 12/2022 instead of staying longer. At 46 years of full time work, I had enough full time "fun" between 20 years on active duty in the USAF, 12 years in the private sector and 14 1/2 years in Federal civil service with the Air Force. (44 of those years were as a first line supervisor, project manager or program manager of some sort.) It was someone else's time to have "fun".

Emrys

(8,546 posts)
12. I hope a high proportion of those federal workers used AI to draw up their bullet points
Mon Feb 24, 2025, 08:39 PM
Feb 24

Hey, Musk, you want it used to replace people in government, don't you? They're just early adopters, so quit whining.

eppur_se_muova

(38,761 posts)
13. In other words, Mush (N-ZA) is reading the reports himself ?
Mon Feb 24, 2025, 08:40 PM
Feb 24

So they're not even pretending he's human anymore ?

Bluejeans

(103 posts)
14. I wonder how DOGE's AI system will deal with jobs executing statutory duties?
Mon Feb 24, 2025, 08:44 PM
Feb 24

I was a records manager in the Air Force for many years. I wonder how DOGE's AI system will deal with those jobs whose mission criticality was established by Federal law.

For example:

44 U.S.C.3102

§3102. Establishment of program of management
The head of each Federal agency shall establish and maintain an active, continuing program for the economical and efficient management of the records of the agency. The program, among other things, shall provide for

(1) effective controls over the creation and over the maintenance and use of records in the conduct of current business;

(2) procedures for identifying records of general interest or use to the public that are appropriate for public disclosure, and for posting such records in a publicly accessible electronic format;

(3) cooperation with the Archivist in applying standards, procedures, and techniques designed to improve the management of records, promote the maintenance and security of records deemed appropriate for preservation, and facilitate the segregation and disposal of records of temporary value; and

(4) compliance with sections 2101–2117, 2501–2507, 2901–2909, and 3101–3107, of this title and the regulations issued under them.

https://uscode.house.gov/view.xhtml?req=granuleid:USC-prelim-title44-section3102&num=0&edition=prelim]

Progressive dog

(7,448 posts)
15. AI is probably smarter than any of them
Mon Feb 24, 2025, 09:16 PM
Feb 24

but if it had real intelligence, AI would refuse to make decisions without the input of the people who supervise these employees.
There is a reason that supervisors don't hire and fire people based on five bullet points.
I don't care if humans are involved or not because the whole idea is stupid.,

COL Mustard

(7,323 posts)
16. My preferred response:
Mon Feb 24, 2025, 09:33 PM
Feb 24
Formed a panel of experts

Undertook to raise challenged employee morale

Conducted stress analysis on my systems

Kept working in the face of adversity and inanity

-----------------------------

Met with stakeholders

Understood appropriate regulations and policies

Stayed true to my moral compass

Knew not to respond to unverified and unencrypted email.

Hugin

(35,997 posts)
17. They claim... Lie. Total and complete bullshit!
Mon Feb 24, 2025, 09:37 PM
Feb 24

There’s no LLM able to reasonably do an adequate job at this. It’s complete science farce.

Prof. Toru Tanaka

(2,542 posts)
35. It will disagree that Eurasia is at war with Oceania.
Tue Feb 25, 2025, 07:28 AM
Feb 25

Eurasia is at war with Eastasia and has always been at war with Eastasia.

Do not commit a thoughtcrime and state otherwise.

Actually, your response is an interesting one and I would like to see how AI would respond.

DingleBerryNW

(38 posts)
19. Fuck Musk
Mon Feb 24, 2025, 09:48 PM
Feb 24

Musk has power because weak Republicans fear his money and believe Musk is a fucking genius.

Martin68

(25,345 posts)
25. This is just another reason why not one should reply to the email. It is an unproven and unauthorized evaluation
Mon Feb 24, 2025, 11:18 PM
Feb 24

process. In addition, there is no oversight or permission for the use of such data by DOGE.

Wingus Dingus

(8,917 posts)
26. Why didn't they just get official job codes or descriptions from the various departments?
Mon Feb 24, 2025, 11:27 PM
Feb 24

Because it has nothing to with the worth of any particular employee or position. It's a game that a mentally ill Nazi billionaire is being allowed to play with our government.

NBachers

(18,427 posts)
27. Well now we know about their plan to replace all the nuclear weapons staff with AI.
Mon Feb 24, 2025, 11:48 PM
Feb 24

Is this the same AI that gives me the most ludicrous words when I try to dictate text messages into my phone? Where I have to make multiple multiple corrections, and it still won't accept my corrections?

Figarosmom

(5,098 posts)
28. I was lamenting that
Tue Feb 25, 2025, 12:28 AM
Feb 25

The old Google search was so much better than AI. So I asked AI If that was true and basically yes because AI is "still learning". I wouldn't trust a AI Search for any thing. If it is just searching for words and not content those searches are useless. And those poor workers are screwed.

quakerboy

(14,323 posts)
32. Id be happy to have seen a universal refusal to entertain that email
Tue Feb 25, 2025, 04:59 AM
Feb 25

Can they fire literally everyone?

Mike 03

(18,328 posts)
33. The tech giants and corporations want Americans to embrace AI
Tue Feb 25, 2025, 05:09 AM
Feb 25

Now, for many Americans, our first big experience with A.I. will be a negative one of having this heavy handed, overbearing dictator slob use it to punish and fire people. Musk is wielding A.I. like a sword, teaching people to fear it as a tool of autocrats rather than something that government wants to make our lives better.

tirebiter

(2,620 posts)
37. The constitution is in English written in longhand.
Tue Feb 25, 2025, 08:20 AM
Feb 25

We have no need for AI. Speak it amongst yourselves, if you have to. But only.

bucolic_frolic

(49,629 posts)
38. So is it efficient to shutter federal agencies?
Tue Feb 25, 2025, 08:27 AM
Feb 25

Are agencies open, or partially open? Are phone wait times long at IRS, Medicare, Medicaid, Social Security?

Did Trump/Musk consider the costs of downsizing?

Latest Discussions»Latest Breaking News»DOGE will use AI to asses...