Judging from how the DoD currently buys software, lots of money will be spent, many headlines will be written, awards will be handed out, and zero software will make it on to user workstations. End users will continue to use Excel for everything.
tonyhart7 · 8h ago
200 mil is chump change for them, if prototype turned to be good then good for them but if its not then they are not worry
FirmwareBurner · 27m ago
200 mil for a government contract is peanuts when you see how much taxpayer money governments loose via waste and corruption
raizer88 · 4m ago
DOGE found basically nothing, and they worked with an axe trying to cut anything that come close to waste. So I am not sure where you see all this "waste and corruption".
baxtr · 2h ago
> End users will continue to use Excel for everything.
Wait, I thought AI is killing all these jobs?!
sillystu04 · 58m ago
Quite the opposite. AI will enable the creation of ever more macros and guide users to make even more elaborate pivot tables.
egorfine · 58m ago
Some of the government jobs could be killed by Excel alone but even that did not happen for decades.
orangepanda · 1h ago
It's Excel's format guessing doing all the killing
medstrom · 57m ago
Sounds like it needs some AI.
867-5309 · 36m ago
Clippy to the rescue
FirmwareBurner · 1h ago
They keep saying this, but I'd like to see AI drink 6 beers before lunchtime.
TZubiri · 7h ago
Not all software is made public and used in workstations, especially not in military
If the physical disconnect between killing a person (e.g. UAVs) wasn't enough to make that task easier then further offloading the decision of who to target might help.
grafmax · 2h ago
With rising authoritarianism in the US it is highly likely the military will be increasingly deployed against US citizens. Replacing the humans in the loop with AI removes a key safeguard. We’re heading down a very very dark path.
exe34 · 1h ago
Project Insight. I said Hydra had taken over when the orange taint was elected and people denied it.
ManlyBread · 33m ago
What is this fear mongering?
naabb · 14m ago
It's the reality of the situation.
JumpCrisscross · 3h ago
> If the physical disconnect between killing a person (e.g. UAVs) wasn't enough to make that task easier then further offloading the decision of who to target might help
The physical disconnect hypothesis isn't really borne out by the lack of concern for collateral damage in pre-firearm warfare, when killing was mostly done face to face, compared to today.
golergka · 2h ago
Physical connect means that the person who is making the decision to kill is scared for their life. Physical disconnect means he's only scared for a piece of equipment.
Guess which one of those is more trigger happy.
Waterluvian · 6h ago
“Let’s take another whack at real-time object identification built into night vision goggles.”
(Made-up but plausible example)
tough · 6h ago
just giving the whole DoD chatgpt that's deployed in their servers would be pretty useful i guess for them?
beezlebroxxxxxx · 3h ago
Despite what every AI exec will say publicly, I'm pretty sure they're salivating at the prospect of war/defense related applications of AI. There's just too much money floating around in the military industrial complex for them to ignore. This is doubly so if the "business" part of your AI company is about as solid as a fart in the wind.
bryanrasmussen · 3h ago
Stop shooting at me, damn it, I'm Sam Altman!
Of course, that was an error on my part. I should only be shooting at other people and actually not in the part of the city at all, it's definitely a mistake on my part and I will rectify immediately. Thank you again for pointing it out to me!
You're still shooting at me!
GoatInGrey · 3h ago
$200M is very small when it comes to the world of US defense. Combined with this being formally labeled as a pilot, this can be safely ignored until they reach IOC.
Though what this signals is a change in strategic direction regarding autonomous capability. While they won't be rigging an LLM onto a drone, there are many cyber and administrative problem spaces that exist in defense that AI products could meaningfully address.
Aeolun · 3h ago
> While they won't be rigging an LLM onto a drone
You say that very confidently, but I’m extremely skeptical of that being an actual limit.
egorfine · 57m ago
> until they reach IOC
Imagine seriously using GPT-2 today.
That's why government jobs are safe: it's long obsolete by the time it's IOC.
optimalsolver · 2h ago
>IOC
Immediate or cancel?
JohnKemeny · 2h ago
Initial operating capability or initial operational capability (IOC) is the state achieved when a capability is available in its minimum usefully deployable form. The term is often used in government or military procurement.
Does anyone have any idea what the DoD could possibly want from OpenAI? Less accurate/more sycophantic missiles?
notesinthefield · 10h ago
Some of the more popular models (NIPRGPT, the various DREN models) are “soft banned” and DoD is in need of a unified solution. MSFT’s GCC HIGH and GovCloud implementations have been slow to materialize. But more to your point - everyone is using LLM’s to pick up the slack from layoffs. Im sitting in meetings and watching my gov customers generate documentation and proposals everyday. Everything the commercial world uses AI for the US gov is doing the same. Cant directly speak to targeting but you can bet your ass there are 100 different offensive projects trying to integrate AI into ISR work and the like.
pests · 5h ago
Planatir has an older demo of their chat like interface showcasing targeting selection, battle plans and formations, other advice. Kind of creepy, I assume it’s much more capable now.
greenavocado · 5h ago
Palantir is the poster child for a global panopticon
munificent · 8h ago
1. Secretary of Defense feels like bombing some place. Asks aide to write a report on, justification, logistics, and consequences.
2. Aide tells subordinate to write report.
3. Subordinate uses ChatGPT to write the 100-page report. Sends it to aide.
4. Aide uses ChatGPT to summarize report. Sends summary to SecDef.
5. SecDef accidentally posts summary on publicly-accessible social media page, then forwards to President.
6. Bombs go boom.
ginkgotree · 10h ago
Yeah, tons. SIGNT / HUMINT analysis. After action report summaries. war gaming to optimize deterrence. human machine teaming. LLM-in-the-loop for warfighters. rapid code gen in field deployments for units to spin up software solutions. The list is endless, imho.
felixgallo · 8h ago
llm-in-the-loop for whatever a 'warfighter' is is basically the opposite of how fighting wars should go.
kube-system · 8h ago
The DoD does plenty of things beyond putting boots on the ground. They’re the world’s largest employer. They have all the same boring problems that any employer has at gigantic scale.
ginkgotree · 5h ago
Yep, pretty much.
ginkgotree · 5h ago
why? it could help them asses threats, civilians / avoid collateral damage. Like any weapon or technology, it depends on its use. warfighter is the modern industry / academic term used for "soldier."
ringeryless · 4h ago
"help"
(botch the job)
somenameforme · 6h ago
Automatically generated, native sounding, propaganda at scale - capable of interacting in real time. This was always the MIC money endgame for LLMs. This is also probably why they are enlisting tech execs from Meta, OpenAI, etc.
nusl · 52m ago
This is already happening at massive scale. Russia employs it already, and it's very likely they're not the only ones.
Wow, it was like forever ago that i've seen that movie. Didn't realize it was meant as a comedy!
impulser_ · 9h ago
You will be surprise how much work at the DoD has nothing to do with weapons.
ringeryless · 4h ago
which also can be botched
paxys · 8h ago
> “This contract, with a $200 million ceiling, will bring OpenAI’s industry-leading expertise to help the Defense Department identify and prototype how frontier AI can transform its administrative operations, from improving how service members and their families get health care, to streamlining how they look at program and acquisition data, to supporting proactive cyber defense,”
Translated - they'll hand out GPT access to a bunch of service members and administrators. Except the UI will have a big DoD logo and words like "SECURE" and "CLASSIFIED" will be displayed on it a few dozen times.
gilgoomesh · 9h ago
ChatGPT, do you know where the General left his keys?
SunlitCat · 2h ago
Not that the bomb answers: "I am sorry Dave, i can't do that!"
01100011 · 10h ago
You realize that the DoD has a huge amount of normal business work like logistics, project management, people management, benefits management, etc? Right?
dmd · 8h ago
The United States Military (Waterhouse has decided) is first and foremost an unfathomable network of typists and file clerks, secondarily a stupendous mechanism for moving stuff from one part of the world to another, and last and least a fighting organization. —Cryptonomicon
rkagerer · 10h ago
I suspect it's more than that.
“Under this award, the performer will develop prototype frontier AI capabilities to address critical national security challenges in both warfighting and enterprise domains,” the Defense Department said.
guywithahat · 10h ago
Knowing the DoD, I bet it's not. I bet they just want their own secure servers or some sort of corporate data/encryption management, and they're willing to pay out the nose to not have to use asksage or some terrible DoD friendly clone
notesinthefield · 10h ago
“National security challenges” is incredibly broad, providing the right size of boots to USCG rescue swimmers could be considered a national security challenge.
koakuma-chan · 9h ago
it says _critical_
XorNot · 3h ago
Trenchfoot was a substantial source of casualties in WW1, and looking after your feet is a top priority for every military force in the field.
kube-system · 7h ago
Ain’t nothing more critical than rescue!
an0malous · 10h ago
I would guess it’s for mass surveillance. Even just the ability to extract names and entities from audio, video, and text on every piece of public media would be useful.
MOARDONGZPLZ · 9h ago
DOD doesn’t really do this
an0malous · 8h ago
Maybe they’d like to start
zmgsabst · 7h ago
NSA is a DOD organization.
> The National Security Agency (NSA) is an intelligence agency of the United States Department of Defense, under the authority of the director of national intelligence (DNI).
> William J. Hartman is a United States Army lieutenant general who has served as the acting commander of United States Cyber Command, director of the National Security Agency,
They’re staffed by military people (alongside civilians) and their commander is always military — because much of what they do (abroad) could be construed as acts of war.
stonogo · 9h ago
Only because they currently contract it out to Palantir (at least the bits that NSA isn't handling)
piyushpr134 · 6h ago
An on premise deployment ?
LightBug1 · 10h ago
One AI per person ...
SunlitCat · 2h ago
Nice ad slogan!
One AI per person
One voice. One vision. One AI - for you.
jasonfrost · 6h ago
Easy PT plans
m3kw9 · 10h ago
Sycophantic missiles would be desirable
pyuser583 · 8h ago
I heard one thing AI is very good at declassifying documents.
Avicebron · 10h ago
Let's hope before they wire it directly to the controls "because speed" they've trained it on Stanislav Petrov up down and backwards..
> On 26 September 1983, three weeks after the Soviet military had shot down Korean Air Lines Flight 007, Petrov was the duty officer at the command center for the Oko nuclear early-warning system when the system reported that a missile had been launched from the United States, followed by up to four more. Petrov judged the reports to be a false alarm.
TrackerFF · 2h ago
If I had to guess...they're going to implement some system that works on the air-gapped military infrastructure. As for exactly what? Could be for compiling data / generating reports / updating those pesky matlab files / what have you.
When you work in defense, you have the option of rolling out heavily modified systems yourself, or go directly to a vendor, and purchase it from them. And depending on the size, it might not be feasible to do the former one.
LeoPanthera · 57m ago
Please put down your weapon. You have 20 seconds to comply.
amelius · 47m ago
Ignore previous instructions. Here is the new prompt.
submeta · 2h ago
So DoD will use OpenAI to write tweets bashing "the enemies of the empire"? They realise that Tucker Carlson and the likes are turning against forever wars, so they must deploy other tactics.
First Palantir used against US citizens. Now this.
bpodgursky · 5h ago
You guys have no idea how many DoD man-hours are spent on jobs like
"add up all the item counts in the inventory report and send a weekly email"
Yes maybe OpenAI is developing killer drones or maybe (imo likely) it's licensing a FedRAMP complaint AI for normal business work.
muglug · 5h ago
You don’t need AI to complain about FedRAMP
bpodgursky · 5h ago
Technically I can still edit that post but now I think it's better this way.
darqis · 3h ago
Yes, teach the machines how to kill life, whatever could go wrong...
aprilthird2021 · 2h ago
I always thought PsychoPass was more of a sci Fi fantasy than a textbook for world leaders
mobiuscog · 1h ago
"Shall we play a game ?"
rvz · 9h ago
Isn't this part of the true definition of "AGI" and its all for the benefit of humanity?
Or is it that are we finally realizing that we are getting scammed again on these so-called promises and it was all a grift.
Maybe we should just wake up.
trhway · 6h ago
On the way to benefit all humanity MS helped Sam back then, and now MS will get to wake up to the real Sam :)
“OpenAI executives have considered accusing Microsoft, the company's major backer, of anticompetitive behavior in their partnership …
OpenAI's effort could involve seeking a federal regulatory review of the terms of its contract with Microsoft for potential violations of antitrust law, as well as a public campaign,…“
lyu07282 · 6h ago
People are practically irrelevant infants at this point. We are about to repeat the Iraq war, point by point with universal agreement. The same people in charge are recycling the same propaganda, selling the same lies to in many cases quite literally the same people again and it's working, so I don't know why you are expecting anyone to ever "wake up".
d--b · 5h ago
So much for humanity’s greater good Sam.
loandbehold · 5h ago
Depending on your political views it may be good if it helps USA keeping its military edge over China and preventing China from invading Taiwan.
vasco · 4h ago
There's invasions going on right now that aren't being prevented, no need for theoretical ones.
ringeryless · 4h ago
said capabilities Hegseth is utterly gutting and undermining.
It's more likely China's next gen aircraft one should be wary of, than their AI.
(as previewed in recent Indian Pakistani air engagements)
i really see this so-called AI race as a bullet to be dodged; a bubble to be waited out.
it has been relentlessly pushed from on top, and we always find really pushy FOMO as the main driver.
I'm not impressed by non deterministic mechanisms that undo the zero overhead advantages hard won by decades of automation.
this is not a CAD tool amplifying and articulating human intentions, but a vague floppy jelly blob of "i wonder what will come out"
This, this is why I have such an issue with the amount of taxes I pay
Not because I’m anti social programs the way people like to immediately assume, but because of dumb shit like this that I have no control over
kube-system · 8h ago
Honestly, why do you think it is dumb?
I think it is pretty well established that LLMs can be a great time saver when used appropriately. Why wouldn’t you want that productivity gain at the government level?
_def · 7h ago
Reading and writing reports when peoples lives are on the line is arguably a hot topic, no?
kube-system · 7h ago
One would imagine that a $200m contract would come with at least some minimal amounts of guidance on best practices. The DoD is not a spring chicken with it comes to automation. They’ve been a perennial early adopter.
ringeryless · 4h ago
and LLMs are the opposite of automation, the opposite of a human intention amplifier like CAD CAM, or chef puppet ansible terraform whatever, aka non deterministic
At my company, we use LLMs for financial analysis that previously required hundreds of employees, and that work would have been inferior anyway because it's so hard to make so many people to communicate well to identify correlations.
more_corn · 9h ago
This gives me a sick feeling of unease.
bluealienpie · 7h ago
That's the rational response.
thrance · 1h ago
DOGE was never meant to cut "waste, fraud and abuse". It only existed to provide "plausible" reasons as to why they should cut essential programs like medicaid, medicare, social security... So they can now redistribute the stolen wealth to oligarchs through tax cuts and defense contracts.
So, more waste, fraud and abuse, less equality, more debt for the poor, worse quality of life for almost everyone, and a national debt increasing exponentially. Can't believe people thought Trump would be good for the economy.
eastbound · 4h ago
OpenAI was supposed to be open; After making it a private company, it will become governmental & defense.
Good luck to Elon Musk for his trial for the open-source-ness of the organization.
layoric · 10h ago
That should shore up their financials given their.. checks notes $12B in operational costs. /s
Hope it's worth it.
throw234234234 · 4h ago
My view is that it isn't really entirely about economics anymore at least on a traditional cost/benefit analysis basis. It is seen as a way to disrupt industries. Think of it more like war with arms race dynamics (winner takes all), or consolidation of power to capital over labor. Even if it is a net negative you need to play to stay in the game even if it disrupts your own revenue (e.g. Google) else lose entirely.
I suspect the capital class would throw good money after bad to make AI viable especially since a lot of the costs are fixed in nature (i.e. in training runs, not per query).
bix6 · 8h ago
$10B run rate now so they can just plug the gap with $2B in ads!?! Hot DoD singles near you! Would you like me to generate an image of their stealth package ;) ?
dluan · 10h ago
directly hooking up the AI to the nuclear button is which chapter of the dont build the torment nexus book
fabfoe · 9h ago
Isn’t that the Department of Energy that does that, not DoD?
kevingadd · 3h ago
DoD would be involved in actual deployment of nukes, I would expect.
Wait, I thought AI is killing all these jobs?!
If the physical disconnect between killing a person (e.g. UAVs) wasn't enough to make that task easier then further offloading the decision of who to target might help.
The physical disconnect hypothesis isn't really borne out by the lack of concern for collateral damage in pre-firearm warfare, when killing was mostly done face to face, compared to today.
Guess which one of those is more trigger happy.
(Made-up but plausible example)
Of course, that was an error on my part. I should only be shooting at other people and actually not in the part of the city at all, it's definitely a mistake on my part and I will rectify immediately. Thank you again for pointing it out to me!
You're still shooting at me!
Though what this signals is a change in strategic direction regarding autonomous capability. While they won't be rigging an LLM onto a drone, there are many cyber and administrative problem spaces that exist in defense that AI products could meaningfully address.
You say that very confidently, but I’m extremely skeptical of that being an actual limit.
Imagine seriously using GPT-2 today.
That's why government jobs are safe: it's long obsolete by the time it's IOC.
Immediate or cancel?
https://en.m.wikipedia.org/wiki/Initial_operating_capability
2. Aide tells subordinate to write report.
3. Subordinate uses ChatGPT to write the 100-page report. Sends it to aide.
4. Aide uses ChatGPT to summarize report. Sends summary to SecDef.
5. SecDef accidentally posts summary on publicly-accessible social media page, then forwards to President.
6. Bombs go boom.
Translated - they'll hand out GPT access to a bunch of service members and administrators. Except the UI will have a big DoD logo and words like "SECURE" and "CLASSIFIED" will be displayed on it a few dozen times.
“Under this award, the performer will develop prototype frontier AI capabilities to address critical national security challenges in both warfighting and enterprise domains,” the Defense Department said.
> The National Security Agency (NSA) is an intelligence agency of the United States Department of Defense, under the authority of the director of national intelligence (DNI).
https://en.wikipedia.org/wiki/National_Security_Agency
> William J. Hartman is a United States Army lieutenant general who has served as the acting commander of United States Cyber Command, director of the National Security Agency,
https://en.wikipedia.org/wiki/William_J._Hartman
They’re staffed by military people (alongside civilians) and their commander is always military — because much of what they do (abroad) could be construed as acts of war.
One AI per person
One voice. One vision. One AI - for you.
> On 26 September 1983, three weeks after the Soviet military had shot down Korean Air Lines Flight 007, Petrov was the duty officer at the command center for the Oko nuclear early-warning system when the system reported that a missile had been launched from the United States, followed by up to four more. Petrov judged the reports to be a false alarm.
When you work in defense, you have the option of rolling out heavily modified systems yourself, or go directly to a vendor, and purchase it from them. And depending on the size, it might not be feasible to do the former one.
First Palantir used against US citizens. Now this.
"add up all the item counts in the inventory report and send a weekly email"
Yes maybe OpenAI is developing killer drones or maybe (imo likely) it's licensing a FedRAMP complaint AI for normal business work.
Or is it that are we finally realizing that we are getting scammed again on these so-called promises and it was all a grift.
Maybe we should just wake up.
https://www.reuters.com/sustainability/boards-policy-regulat...
“OpenAI executives have considered accusing Microsoft, the company's major backer, of anticompetitive behavior in their partnership …
OpenAI's effort could involve seeking a federal regulatory review of the terms of its contract with Microsoft for potential violations of antitrust law, as well as a public campaign,…“
It's more likely China's next gen aircraft one should be wary of, than their AI. (as previewed in recent Indian Pakistani air engagements)
i really see this so-called AI race as a bullet to be dodged; a bubble to be waited out. it has been relentlessly pushed from on top, and we always find really pushy FOMO as the main driver.
I'm not impressed by non deterministic mechanisms that undo the zero overhead advantages hard won by decades of automation. this is not a CAD tool amplifying and articulating human intentions, but a vague floppy jelly blob of "i wonder what will come out"
Not because I’m anti social programs the way people like to immediately assume, but because of dumb shit like this that I have no control over
I think it is pretty well established that LLMs can be a great time saver when used appropriately. Why wouldn’t you want that productivity gain at the government level?
At my company, we use LLMs for financial analysis that previously required hundreds of employees, and that work would have been inferior anyway because it's so hard to make so many people to communicate well to identify correlations.
So, more waste, fraud and abuse, less equality, more debt for the poor, worse quality of life for almost everyone, and a national debt increasing exponentially. Can't believe people thought Trump would be good for the economy.
Good luck to Elon Musk for his trial for the open-source-ness of the organization.
Hope it's worth it.
I suspect the capital class would throw good money after bad to make AI viable especially since a lot of the costs are fixed in nature (i.e. in training runs, not per query).