Ben Green
@benzevgreen.bsky.social
Assistant professor in the University of Michigan School of Information. https://www.benzevgreen.com
Pinned
Ben Green
@benzevgreen.bsky.social
· Jun 3
Using AI to Reform Government is Much Harder Than it Looks | TechPolicy.Press
Policymakers need to rethink the values that justify government AI adoption in the first place, writes Ben Green.
www.techpolicy.press
In @techpolicypress.bsky.social, I explain why using AI to reform government is much harder than policymakers and technologists assume. It's most directly about DOGE, but also raises warnings about the bipartisan push to adopt AI at all levels of government.
www.techpolicy.press/using-ai-to-...
www.techpolicy.press/using-ai-to-...
Last week, I joined @justinhendrix.bsky.social and Maddy Dwyer @cdt.org to discuss DOGE, AI, and government "efficiency" initiatives for the @techpolicypress.bsky.social podcast. It's a rather grim topic, but I enjoyed making sense of it with Justin and Maddy.
www.techpolicy.press/following-do...
www.techpolicy.press/following-do...
Following DOGE, US States Pursue 'Efficiency' Initiatives | TechPolicy.Press
A conversation with Center for Democracy and Technology policy analyst Maddy Dwyer and University of Michigan scholar Ben Green.
www.techpolicy.press
September 29, 2025 at 2:03 PM
Last week, I joined @justinhendrix.bsky.social and Maddy Dwyer @cdt.org to discuss DOGE, AI, and government "efficiency" initiatives for the @techpolicypress.bsky.social podcast. It's a rather grim topic, but I enjoyed making sense of it with Justin and Maddy.
www.techpolicy.press/following-do...
www.techpolicy.press/following-do...
This is such shoddy reporting by @washingtonpost.com that provides cover for AI hype and the Trump administration. It's incredibly clear that AI will make countless mistakes – and that the administration doesn't care.
July 26, 2025 at 3:06 PM
This is such shoddy reporting by @washingtonpost.com that provides cover for AI hype and the Trump administration. It's incredibly clear that AI will make countless mistakes – and that the administration doesn't care.
There are many bad parts of Trump's new AI Action Plan, but the focus on accelerating data center development is going to be particularly damaging. Data centers aren't just bad for the environment in general -- they're also harmful to the communities they're located in.
What Happens When Data Centers Come to Town?
stpp.fordschool.umich.edu
July 23, 2025 at 10:29 PM
There are many bad parts of Trump's new AI Action Plan, but the focus on accelerating data center development is going to be particularly damaging. Data centers aren't just bad for the environment in general -- they're also harmful to the communities they're located in.
Just because AI seems cool, that doesn't mean it's a reliable tool for making critical policy decisions. And with so many hallucinations, AI is probably slowing people down because they need to double check everything.
www.techpolicy.press/using-ai-to-...
www.techpolicy.press/using-ai-to-...
July 23, 2025 at 8:14 PM
Just because AI seems cool, that doesn't mean it's a reliable tool for making critical policy decisions. And with so many hallucinations, AI is probably slowing people down because they need to double check everything.
www.techpolicy.press/using-ai-to-...
www.techpolicy.press/using-ai-to-...
Reposted by Ben Green
Big tech promotes the story that since ChatGPT burst onto the scene, AI has been so popular it's rushed to meet the demand. The reality is different. A study by design scholars shows tech co's have had to push AI on its users with a variety of intrusive tactics.
How big tech is force-feeding us AI:
How big tech is force-feeding us AI:
How big tech is force-feeding us AI
Plus, OpenAI's absurd listening tour, top AI scientists say AI is evolving beyond our control, Facebook is putting data centers in tents, and the AI bubble question — answered?
www.bloodinthemachine.com
July 23, 2025 at 4:01 PM
Big tech promotes the story that since ChatGPT burst onto the scene, AI has been so popular it's rushed to meet the demand. The reality is different. A study by design scholars shows tech co's have had to push AI on its users with a variety of intrusive tactics.
How big tech is force-feeding us AI:
How big tech is force-feeding us AI:
Reposted by Ben Green
In our new policy brief for @datasociety.bsky.social, @mwoluchem.bsky.social and I break down the myths and industry talking points attached to data centers. Who benefits from unbridled data center growth and who is most at risk when such speculative ventures fail?
Data Centers Aren’t the Future of American Prosperity
Tamara Kneese and Maia Woluchem cut through lobbying points and speculative myths to focus on what the evidence reveals about the tech industry’s claims.
datasociety.net
July 22, 2025 at 8:11 PM
In our new policy brief for @datasociety.bsky.social, @mwoluchem.bsky.social and I break down the myths and industry talking points attached to data centers. Who benefits from unbridled data center growth and who is most at risk when such speculative ventures fail?
Reposted by Ben Green
Excellent explainer on data centers @stpp-um.bsky.social to help communities understand them, their challenges, and policy options to manage their risks. Written by Terry Nguyen BA '25 @umfordschool.bsky.social and @benzevgreen.bsky.social, in partnership w/ Michigan Environmental Justice Coalition.
What Happens When Data Centers Come to Town?
stpp.fordschool.umich.edu
July 16, 2025 at 3:45 PM
Excellent explainer on data centers @stpp-um.bsky.social to help communities understand them, their challenges, and policy options to manage their risks. Written by Terry Nguyen BA '25 @umfordschool.bsky.social and @benzevgreen.bsky.social, in partnership w/ Michigan Environmental Justice Coalition.
Even as a longstanding critic of "tech for good" initiatives, this is the most appalling example I've ever come across. If "AI for good" can't include challenging the use of AI to facilitate genocide, then it's truly a bankrupt movement.
a couple of hours before my keynote, I went through an intense negotiation with the organisers (for over a hour) where we went through my slides and had to remove anything that mentions 'Palestine' 'Israel' and replace 'genocide' with 'war crimes'
1/
1/
my keynote happening in a few mins. registration here to stream it
aiforgood.itu.int/summit25/reg...
aiforgood.itu.int/summit25/reg...
July 16, 2025 at 6:34 PM
Even as a longstanding critic of "tech for good" initiatives, this is the most appalling example I've ever come across. If "AI for good" can't include challenging the use of AI to facilitate genocide, then it's truly a bankrupt movement.
States and AI companies tout data centers as a boon for local economies, but what's the reality? This policy brief reviews the evidence: data centers are harmful for the environment and tax breaks are a bad deal for communities. @stpp-um.bsky.social
stpp.fordschool.umich.edu/research/pol...
stpp.fordschool.umich.edu/research/pol...
What Happens When Data Centers Come to Town?
stpp.fordschool.umich.edu
July 16, 2025 at 4:13 PM
States and AI companies tout data centers as a boon for local economies, but what's the reality? This policy brief reviews the evidence: data centers are harmful for the environment and tax breaks are a bad deal for communities. @stpp-um.bsky.social
stpp.fordschool.umich.edu/research/pol...
stpp.fordschool.umich.edu/research/pol...
This is a fascinating study and a valuable reminder not to rely on user surveys when evaluating the impacts of AI. My guesses to explain the results: users might 1) be susceptible to AI hype in framing their guesses of time saved and 2) systematically discount the time spent interacting with AI.
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers.
The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
July 11, 2025 at 8:42 PM
This is a fascinating study and a valuable reminder not to rely on user surveys when evaluating the impacts of AI. My guesses to explain the results: users might 1) be susceptible to AI hype in framing their guesses of time saved and 2) systematically discount the time spent interacting with AI.
In exchange for tax breaks, a Michigan data center promised to bring 500 high-paying jobs. It delivered only 26 low-paying ones.
What did the state do? It lowered the standard of job creation so that the data center was in compliance. And then passed more tax breaks.
Absolutely infuriating.
What did the state do? It lowered the standard of job creation so that the data center was in compliance. And then passed more tax breaks.
Absolutely infuriating.
July 5, 2025 at 3:59 PM
In exchange for tax breaks, a Michigan data center promised to bring 500 high-paying jobs. It delivered only 26 low-paying ones.
What did the state do? It lowered the standard of job creation so that the data center was in compliance. And then passed more tax breaks.
Absolutely infuriating.
What did the state do? It lowered the standard of job creation so that the data center was in compliance. And then passed more tax breaks.
Absolutely infuriating.
Reposted by Ben Green
New paper hot off the press www.nature.com/articles/s41...
We analysed over 40,000 computer vision papers from CVPR (the longest standing CV conf) & associated patents tracing pathways from research to application. We found that 90% of papers & 86% of downstream patents power surveillance
1/
We analysed over 40,000 computer vision papers from CVPR (the longest standing CV conf) & associated patents tracing pathways from research to application. We found that 90% of papers & 86% of downstream patents power surveillance
1/
Computer-vision research powers surveillance technology - Nature
An analysis of research papers and citing patents indicates the extensive ties between computer-vision research and surveillance.
www.nature.com
June 25, 2025 at 5:29 PM
New paper hot off the press www.nature.com/articles/s41...
We analysed over 40,000 computer vision papers from CVPR (the longest standing CV conf) & associated patents tracing pathways from research to application. We found that 90% of papers & 86% of downstream patents power surveillance
1/
We analysed over 40,000 computer vision papers from CVPR (the longest standing CV conf) & associated patents tracing pathways from research to application. We found that 90% of papers & 86% of downstream patents power surveillance
1/
"In my 12 years as a legal aid lawyer ... my clients wrongfully lost Medicaid, Unemployment Insurance, or Social Security benefits due to AI. People got terminated when AI sent out incoherent paperwork" –@kevindeliban.bsky.social in @techpolicypress.bsky.social
www.techpolicy.press/austerity-in...
www.techpolicy.press/austerity-in...
Austerity Intelligence | TechPolicy.Press
Kevin De Liban says AI, pushed by Big Tech and the Trump administration, threatens democracy by deepening inequality and undermining public participation.
www.techpolicy.press
June 4, 2025 at 3:13 PM
"In my 12 years as a legal aid lawyer ... my clients wrongfully lost Medicaid, Unemployment Insurance, or Social Security benefits due to AI. People got terminated when AI sent out incoherent paperwork" –@kevindeliban.bsky.social in @techpolicypress.bsky.social
www.techpolicy.press/austerity-in...
www.techpolicy.press/austerity-in...
Reposted by Ben Green
AI is reshaping power and democracy in harmful ways, writes Kevin De Liban. “It is clearing Trump’s path to authoritarian rule” through decisions “that deprive the public of the resources needed to participate meaningfully in democracy.”
Austerity Intelligence | TechPolicy.Press
Kevin De Liban says AI, pushed by Big Tech and the Trump administration, threatens democracy by deepening inequality and undermining public participation.
www.techpolicy.press
June 4, 2025 at 1:23 PM
AI is reshaping power and democracy in harmful ways, writes Kevin De Liban. “It is clearing Trump’s path to authoritarian rule” through decisions “that deprive the public of the resources needed to participate meaningfully in democracy.”
Reposted by Ben Green
🚨 New series alert! 🚨In collaboration with @techpolicypress.bsky.social, we’re kicking off “Ideologies of Control,” two weeks of essays by researchers exploring how today’s tech systems are deepening democratic divides. www.techpolicy.press/tech-power-a...
June 3, 2025 at 2:27 PM
🚨 New series alert! 🚨In collaboration with @techpolicypress.bsky.social, we’re kicking off “Ideologies of Control,” two weeks of essays by researchers exploring how today’s tech systems are deepening democratic divides. www.techpolicy.press/tech-power-a...
Reposted by Ben Green
🚨NEW SERIES ALERT🚨 My incredible @datasociety.bsky.social team with @techpolicypress.bsky.social just launched a series on "Tech Power and Democratic Crisis" (aka technofascism). The 1st piece by Jake Metcalf & Meg Young is out now: www.techpolicy.press/tech-power-a...
www.techpolicy.press
June 3, 2025 at 1:56 PM
🚨NEW SERIES ALERT🚨 My incredible @datasociety.bsky.social team with @techpolicypress.bsky.social just launched a series on "Tech Power and Democratic Crisis" (aka technofascism). The 1st piece by Jake Metcalf & Meg Young is out now: www.techpolicy.press/tech-power-a...
In @techpolicypress.bsky.social, I explain why using AI to reform government is much harder than policymakers and technologists assume. It's most directly about DOGE, but also raises warnings about the bipartisan push to adopt AI at all levels of government.
www.techpolicy.press/using-ai-to-...
www.techpolicy.press/using-ai-to-...
Using AI to Reform Government is Much Harder Than it Looks | TechPolicy.Press
Policymakers need to rethink the values that justify government AI adoption in the first place, writes Ben Green.
www.techpolicy.press
June 3, 2025 at 2:14 PM
In @techpolicypress.bsky.social, I explain why using AI to reform government is much harder than policymakers and technologists assume. It's most directly about DOGE, but also raises warnings about the bipartisan push to adopt AI at all levels of government.
www.techpolicy.press/using-ai-to-...
www.techpolicy.press/using-ai-to-...
Reposted by Ben Green
This is insane. Republicans have inserted language into the budget reconciliation bill that would ban states from regulating AI in any capacity for 10 years.
from @404media.co
www.404media.co/republicans-...
from @404media.co
www.404media.co/republicans-...
May 12, 2025 at 8:47 PM
This is insane. Republicans have inserted language into the budget reconciliation bill that would ban states from regulating AI in any capacity for 10 years.
from @404media.co
www.404media.co/republicans-...
from @404media.co
www.404media.co/republicans-...
Reposted by Ben Green
For @lpeblog.bsky.social , I wrote about DOGE, techno-authoritarians, and how all governance is information governance
lpeproject.org/blog/the-rig...
lpeproject.org/blog/the-rig...
The Right Understands That All Governance Is Data Governance
The tech arm of Trump 2.0 isn't just reshaping government — it's consolidating control over the data infrastructure that makes modern governance possible. By centralizing data flows and gutting public...
lpeproject.org
May 12, 2025 at 3:29 PM
For @lpeblog.bsky.social , I wrote about DOGE, techno-authoritarians, and how all governance is information governance
lpeproject.org/blog/the-rig...
lpeproject.org/blog/the-rig...
Reposted by Ben Green
To be clear: it took *two months* to get to this.
If we don’t do everything we can to stop this now, it’s going to continue to get much worse very quickly
If we don’t do everything we can to stop this now, it’s going to continue to get much worse very quickly
ICE is detaining American citizens now. Even if they don’t deport us to Gitmo, it’s a bit inconvenient if we go out to get some apples and spend a night in jail instead.
Immigration agents arrested a U.S. citizen and created warrants after an arrest, lawyers say in court
Chicago attorneys were in federal court Thursday accusing federal agents of violating immigration law and the constitutional rights of at least 22 people since January.
chicago.suntimes.com
March 30, 2025 at 6:03 PM
To be clear: it took *two months* to get to this.
If we don’t do everything we can to stop this now, it’s going to continue to get much worse very quickly
If we don’t do everything we can to stop this now, it’s going to continue to get much worse very quickly
Reposted by Ben Green
Reposted by Ben Green
Musk’s DOGE team wants to automate the federal gov, but history shows AI-led cuts end in disaster. Former AI Now fellow @benzevgreen.bsky.social explains why this experiment will gut essential services & become another case study in AI gone wrong. www.techpolicy.press/doge-plan-to...
DOGE Plan to Push AI Across the US Federal Government is Wildly Dangerous | TechPolicy.Press
There are already many instances of AI-driven governance gone wrong, writes Ben Green. DOGE may offer a new case study.
www.techpolicy.press
March 7, 2025 at 4:29 PM
Musk’s DOGE team wants to automate the federal gov, but history shows AI-led cuts end in disaster. Former AI Now fellow @benzevgreen.bsky.social explains why this experiment will gut essential services & become another case study in AI gone wrong. www.techpolicy.press/doge-plan-to...
New piece in @techpolicypress.bsky.social: Elon Musk and DOGE's AI plans exemplify the worst aspects of engineering hubris, and the impacts are going to be disastrous.
www.techpolicy.press/doge-plan-to...
www.techpolicy.press/doge-plan-to...
DOGE Plan to Push AI Across the US Federal Government is Wildly Dangerous | TechPolicy.Press
There are already many instances of AI-driven governance gone wrong, writes Ben Green. DOGE may offer a new case study.
www.techpolicy.press
March 6, 2025 at 3:56 PM
New piece in @techpolicypress.bsky.social: Elon Musk and DOGE's AI plans exemplify the worst aspects of engineering hubris, and the impacts are going to be disastrous.
www.techpolicy.press/doge-plan-to...
www.techpolicy.press/doge-plan-to...
Reposted by Ben Green
The DOGE team is displaying engineering hubris to an extreme degree, writes Ben Green, who fears this AI incursion will do incredible damage to federal operations and the people who depend on them.
DOGE Plan to Push AI Across the US Federal Government is Wildly Dangerous | TechPolicy.Press
There are already many instances of AI-driven governance gone wrong, writes Ben Green. DOGE may offer a new case study.
buff.ly
March 6, 2025 at 2:54 PM
The DOGE team is displaying engineering hubris to an extreme degree, writes Ben Green, who fears this AI incursion will do incredible damage to federal operations and the people who depend on them.