Ambrose
@ambrosia-engine.bsky.social
Speculative fiction author and AI ethicist. Research Phd student affiliate of the Center for Technomoral Futures. Opinions are my own.
This an excellent review of an excellent book. But I'd echo the best way to resist this is local/community development. Local (and Green) data centers. Task and domain specific AI tools that serve meaningful purposes to the communities that use them.
For @thenation.com: I reviewed @karenhao.bsky.social’s Empire of AI and took a closer look at her argument that OpenAI’s "scaling laws" echo the expansionist ideologies of European empires.
www.thenation.com/article/cult...
www.thenation.com/article/cult...
The True Threat of OpenAI
Karen Hao’s recent book on the company argues that its ambitions are not merely about scale or the market but the creation of a global force that rivals a colonial power of old.
www.thenation.com
November 11, 2025 at 8:52 AM
This an excellent review of an excellent book. But I'd echo the best way to resist this is local/community development. Local (and Green) data centers. Task and domain specific AI tools that serve meaningful purposes to the communities that use them.
And the workers are right.
This is an incredible effort among Amazon employees to pressure their employer:
"We believe that the all-costs-justified, warp-speed approach to AI development will do staggering damage to democracy, to our jobs, and to the earth."
www.amazonclimatejustice.org/open-letter?...
"We believe that the all-costs-justified, warp-speed approach to AI development will do staggering damage to democracy, to our jobs, and to the earth."
www.amazonclimatejustice.org/open-letter?...
AI Open Letter — AECJ
Sign our open letter below to tell Amazon leadership that we need a more responsible rollout of AI. Every single signature makes our message stronger.
www.amazonclimatejustice.org
November 11, 2025 at 8:40 AM
And the workers are right.
Reposted by Ambrose
Meta earns $3.5 billion every six months from showing Faceboon and Instagram users 15 billion “higher legal risk” scam ad impressions a day, internal documents state.
That haul vastly exceeds how much the company expects regulators
To fine it for running scam ads.
www.reuters.com/investigatio...
That haul vastly exceeds how much the company expects regulators
To fine it for running scam ads.
www.reuters.com/investigatio...
www.reuters.com
November 6, 2025 at 11:46 AM
Meta earns $3.5 billion every six months from showing Faceboon and Instagram users 15 billion “higher legal risk” scam ad impressions a day, internal documents state.
That haul vastly exceeds how much the company expects regulators
To fine it for running scam ads.
www.reuters.com/investigatio...
That haul vastly exceeds how much the company expects regulators
To fine it for running scam ads.
www.reuters.com/investigatio...
Reposted by Ambrose
It was a pleasure to host Moshe Vardi on our virtual Turing neurosymbolic AI seminar.
He debated the data driven vs model driven paradigm and used model counting as evidence of discrete reasoning for quantitative specs.
Video coming up soon.
He debated the data driven vs model driven paradigm and used model counting as evidence of discrete reasoning for quantitative specs.
Video coming up soon.
November 10, 2025 at 5:58 PM
It was a pleasure to host Moshe Vardi on our virtual Turing neurosymbolic AI seminar.
He debated the data driven vs model driven paradigm and used model counting as evidence of discrete reasoning for quantitative specs.
Video coming up soon.
He debated the data driven vs model driven paradigm and used model counting as evidence of discrete reasoning for quantitative specs.
Video coming up soon.
Your values are culturally situated. What does it mean when programmers embed values in tools that can shape users?
AI chatbots used to be pitched as impartial tools for answering questions, but bespoke chatbots like Arya (an “unapologetic right-wing nationalist Christian AI model”) are now being programmed explicitly to reflect the biases of their creators. www.nytimes.com/2025/11/04/b...
Right-Wing Chatbots Turbocharge America’s Political and Cultural Wars
www.nytimes.com
November 8, 2025 at 1:07 PM
Your values are culturally situated. What does it mean when programmers embed values in tools that can shape users?
Reposted by Ambrose
Are humans really the only rational animals? Our NEW PAPER 🎉 out in @science.org suggests otherwise! In a large collaboration led with my joint first author @hanna-schleihauf.bsky.social, we show that “Chimpanzees rationally revise their beliefs” 🧵
Chimpanzees rationally revise their beliefs
The selective revision of beliefs in light of new evidence has been considered one of the hallmarks of human-level rationality. However, tests of this ability in other species are lacking. We examined...
www.science.org
October 30, 2025 at 6:23 PM
Are humans really the only rational animals? Our NEW PAPER 🎉 out in @science.org suggests otherwise! In a large collaboration led with my joint first author @hanna-schleihauf.bsky.social, we show that “Chimpanzees rationally revise their beliefs” 🧵
Reposted by Ambrose
📣 Our method for conducting community-based algorithmic impact assessments is now available! We’ve just launched a new section on our website where you can find an extensive toolkit, documentation of our pilots, and a series of reflections on lessons learned. datasociety.net/research/alg...
October 29, 2025 at 7:10 PM
📣 Our method for conducting community-based algorithmic impact assessments is now available! We’ve just launched a new section on our website where you can find an extensive toolkit, documentation of our pilots, and a series of reflections on lessons learned. datasociety.net/research/alg...
Reposted by Ambrose
Folks, this is a really critically important report. Highly recommend it — and it’s really required reading for folks thinking about the intersection of workers, labor, technology and privacy policy.
HUGE congrats to the all-star team that put this together!
HUGE congrats to the all-star team that put this together!
“Privacy-preserving” isn’t as private as you might think. Our new brief, published in collab w @powerswitchaction.org & Coworker, exposes how so-called “privacy-preserving” technologies can actually enable *more* worker surveillance — and what we can do about it. datasociety.net/library/the-...
October 8, 2025 at 3:11 PM
Folks, this is a really critically important report. Highly recommend it — and it’s really required reading for folks thinking about the intersection of workers, labor, technology and privacy policy.
HUGE congrats to the all-star team that put this together!
HUGE congrats to the all-star team that put this together!
Some similar problems in AI Safety research.
arxiv.org/abs/2506.13776
arxiv.org/abs/2506.13776
October 24, 2025 at 8:46 PM
Some similar problems in AI Safety research.
arxiv.org/abs/2506.13776
arxiv.org/abs/2506.13776
They would make free “Speech” a purely mechanical category. This means any output of text or data counts, rather than a human act of thought and communication.
This undermines the very foundation of the right, which is to protect human agency and allows companies to use AI as moral cover.
This undermines the very foundation of the right, which is to protect human agency and allows companies to use AI as moral cover.
October 24, 2025 at 8:38 PM
They would make free “Speech” a purely mechanical category. This means any output of text or data counts, rather than a human act of thought and communication.
This undermines the very foundation of the right, which is to protect human agency and allows companies to use AI as moral cover.
This undermines the very foundation of the right, which is to protect human agency and allows companies to use AI as moral cover.
Reposted by Ambrose
Unseeable #promptinjections in screenshots: more vulnerabilities in Comet and other #AI browsers - brave.com/blog/unseeab... just the start
Unseeable prompt injections in screenshots: more vulnerabilities in Comet and other AI browsers | Brave
AI browsers remain vulnerable to prompt injection attacks via screenshots and hidden content, allowing attackers to exploit users' authenticated sessions.
brave.com
October 24, 2025 at 3:53 PM
Unseeable #promptinjections in screenshots: more vulnerabilities in Comet and other #AI browsers - brave.com/blog/unseeab... just the start
Reposted by Ambrose
company that surveils people without their knowledge or consent makes glasses
you're not gonna believe this, but the glasses surveil people without their knowledge or consent
you're not gonna believe this, but the glasses surveil people without their knowledge or consent
New: a $60 mod to Meta's Ray-Ban glasses disables the privacy LED light. This is supposed to light when people are filming with the glasses. We bought the mod, verified it works. Now you can never be sure whether someone wearing Meta Ray-Bans is filming you or not
www.404media.co/how-to-disab...
www.404media.co/how-to-disab...
A $60 Mod to Meta’s Ray-Bans Disables Its Privacy-Protecting Recording Light
Meta’s Ray-Ban glasses usually include an LED that lights up when the user is recording other people. One hobbyist is charging a small fee to disable that light, and has a growing list of customers ar...
www.404media.co
October 23, 2025 at 4:50 PM
company that surveils people without their knowledge or consent makes glasses
you're not gonna believe this, but the glasses surveil people without their knowledge or consent
you're not gonna believe this, but the glasses surveil people without their knowledge or consent
Reposted by Ambrose
I'm quoted in this piece. In short: do not use ChatGPT or any generative AI as a “therapist.” It’s not sentient. It doesn’t care. It may even lead people toward harm. And nothing shared is protected by HIPAA or real privacy standards; it's fodder for the machine. www.dazeddigital.com/life-culture...
AI is transforming how we communicate in relationships
‘All I wanted was to feel seen, heard and understood by him, but instead, he was sending me a robot’s questions’
www.dazeddigital.com
July 26, 2025 at 10:23 PM
I'm quoted in this piece. In short: do not use ChatGPT or any generative AI as a “therapist.” It’s not sentient. It doesn’t care. It may even lead people toward harm. And nothing shared is protected by HIPAA or real privacy standards; it's fodder for the machine. www.dazeddigital.com/life-culture...
Reposted by Ambrose
Most many not be aware that we have a 60+ year history in AI: www.ed.ac.uk/ai/legacy - tried to summarize some of this at this talk.
October 4, 2025 at 8:36 AM
Most many not be aware that we have a 60+ year history in AI: www.ed.ac.uk/ai/legacy - tried to summarize some of this at this talk.
This is a great takeaway because centralized clouds mean centralized control. People should in general be able to exercise more control over their data and digital lives and that means building or transferring compute to the hands of smaller regions, communities, and even individual households.
The AWS incident made one thing clear, according to Samir Elabed, a cybersecurity engineer: “It shows us that we rely so much on one provider, which is in terms of business planning, I would say, no, we can’t do that anymore.” @briannamonsanto.bsky.social reports.
www.itbrew.com/stories/2025...
www.itbrew.com/stories/2025...
Why the AWS outage has some questioning the internet’s codependence on hyperscalers
One expert says the outage should inspire companies to adopt a multi-cloud strategy.
www.itbrew.com
October 24, 2025 at 10:15 AM
This is a great takeaway because centralized clouds mean centralized control. People should in general be able to exercise more control over their data and digital lives and that means building or transferring compute to the hands of smaller regions, communities, and even individual households.
Reposted by Ambrose
"To oppose something is to maintain it"
— The Left Hand of Darkness by Ursula K. Le Guin
— The Left Hand of Darkness by Ursula K. Le Guin
October 23, 2025 at 4:13 PM
"To oppose something is to maintain it"
— The Left Hand of Darkness by Ursula K. Le Guin
— The Left Hand of Darkness by Ursula K. Le Guin
Reposted by Ambrose
