Anthropic Gives METR Network Access
8m
Anthropic said Thursday it would grant nonprofit evaluator METR employee-level access to its offices and internal network to assess AI safety, a move that drew criticism.
by Grady Booch
Granting rights to an AI is ludicrous. Among other things, accountability and responsibility have no meaning to any such entity.
this basically amounts to giving corporations more rights and power... we cover this very argument in this 2024 paper
firstmonday.org/ojs/index.ph...
firstmonday.org/ojs/index.ph...
October 15, 2026 at 9:51 PM
Granting rights to an AI is ludicrous. Among other things, accountability and responsibility have no meaning to any such entity.
I see what you mean. We probably agree that agents have destructive potential and people shouldn't be using them if they don't know what they're doing. They're like chainsaws in that respect
October 16, 2026 at 4:23 PM
I see what you mean. We probably agree that agents have destructive potential and people shouldn't be using them if they don't know what they're doing. They're like chainsaws in that respect
And even that fails to consider the impact of my choice to engage with and even partially endorse AI tools. My main goal is to get starry-eyed "AI for everyone" institutions to back-the-F-up.
October 16, 2026 at 3:07 PM
And even that fails to consider the impact of my choice to engage with and even partially endorse AI tools. My main goal is to get starry-eyed "AI for everyone" institutions to back-the-F-up.
A colleague stepped out of #cyborgWorkers for a Google talk to the EP. Google said Europe was falling behind because it wasn't using AI and this was because women are afraid of AI. This is truly evil – even the "falling behind" part is false! A reminder of our new data (short thread) #AIEthics
The Non-Binarity of Artificial Intelligence Global Competition: 2018-2025
Technological progress plays a crucial role in international relations. In particular, innovation has long been theorized as instrumental to power transitions. It is therefore unsurprising that the re...
osf.io
October 16, 2026 at 7:26 AM
A colleague stepped out of #cyborgWorkers for a Google talk to the EP. Google said Europe was falling behind because it wasn't using AI and this was because women are afraid of AI. This is truly evil – even the "falling behind" part is false! A reminder of our new data (short thread) #AIEthics
This thought experiment illustrates the core problem of the last three decades of Silicon Valley: powerful tech harnessed by extraordinarily reductive and banal thinking about the actual problems facing humanity. An endless parade of rewarding reflexive action on half-baked wouldn't-it-be-cool-ifs
It also seems obvious that past the hypothetical tipping point on AI assistant use, firms in the many underregulated sectors of the economy will take advantage of agentic brittleness and vulnerabilities and the AI companies will simply have added another layer of cost, manipulation, and dysfunction
Humans need friction. So does the economy. Wrote about Musing Ourselves to Death & what a future where the internet transforms into a place for AI assistants to perfect commerce. Spoiler: We talked to some smart ppl & it would be chaos! w/ @matteowong.bsky.social www.theatlantic.com/technology/2...
October 16, 2026 at 5:03 PM
This thought experiment illustrates the core problem of the last three decades of Silicon Valley: powerful tech harnessed by extraordinarily reductive and banal thinking about the actual problems facing humanity. An endless parade of rewarding reflexive action on half-baked wouldn't-it-be-cool-ifs
I fear the non-expert in a substantive space - using agentic/generative AI, and other more advanced tools, offloading the intellectual work (because the user lacks the knowledge), yielding a product they think is "something" when they (w/AI assistance) have produced total bullshit.
October 16, 2026 at 2:33 PM
I fear the non-expert in a substantive space - using agentic/generative AI, and other more advanced tools, offloading the intellectual work (because the user lacks the knowledge), yielding a product they think is "something" when they (w/AI assistance) have produced total bullshit.
Danielle Citron @daniellecitron.bsky.social on accountability for AI. Builds on her work on strict liability for data breaches - an early piece called Reservoirs of Danger, which is also well worth reading. www.project-syndicate.org/commentary/a...
Strict Liability for AI Executives
Danielle Keats Citron argues that preventing frontier labs from causing harm requires holding them legally accountable for it.
www.project-syndicate.org
October 15, 2026 at 11:12 PM
Danielle Citron @daniellecitron.bsky.social on accountability for AI. Builds on her work on strict liability for data breaches - an early piece called Reservoirs of Danger, which is also well worth reading. www.project-syndicate.org/commentary/a...
Been doing AI/NNs since the 90s, and was lucky enough to be in the right place, right time to see early agentic work (SWARM) from Santa Fe Institute. The new AI is in large part still the old AI. Another tool with limited effective uses. Let's figure out what those are.
October 16, 2026 at 2:42 PM
Been doing AI/NNs since the 90s, and was lucky enough to be in the right place, right time to see early agentic work (SWARM) from Santa Fe Institute. The new AI is in large part still the old AI. Another tool with limited effective uses. Let's figure out what those are.
Reposted by Brian J. Enquist
"Open science has been building standards and norms...AI has ignored this...just sucked everything in..we need to reclaim ownership"
Some of us have said for years we need more nuance to openness given the potential for extractive commercial re-use 1/n
scedmunds.substack.com/p/sage-advic...
Some of us have said for years we need more nuance to openness given the potential for extractive commercial re-use 1/n
scedmunds.substack.com/p/sage-advic...
Sage Advice From The FAIR Data Ferryport
Here are some thoughts from the 2026 Sage Assembly, Sage Bionetworks comeback meeting tackling “Trust in the Machine: Verification, Agency, and the Future of Scientific Work.”
scedmunds.substack.com
October 16, 2026 at 12:30 PM
"Open science has been building standards and norms...AI has ignored this...just sucked everything in..we need to reclaim ownership"
Some of us have said for years we need more nuance to openness given the potential for extractive commercial re-use 1/n
scedmunds.substack.com/p/sage-advic...
Some of us have said for years we need more nuance to openness given the potential for extractive commercial re-use 1/n
scedmunds.substack.com/p/sage-advic...
Forget AI and SI. It’s now called Oy I.
Employers worldwide are pulling back on AI as they navigate a flood of errors and low-quality work that sap employee time, raise privacy and security risks, and drive up costs
Companies rushed in on AI. Now some are pulling back
Rapid AI adoption left workers cleaning up a flood of errors and low-quality work
www.theguardian.com
October 16, 2026 at 5:25 AM
Forget AI and SI. It’s now called Oy I.
In just the last few weeks, I've found multiple instances of heavy, undisclosed AI use in writing in major media outlets, academic research, and public commentary. Once you see it you can't unsee it (especially with @pangram.com).
October 15, 2026 at 9:32 PM
In just the last few weeks, I've found multiple instances of heavy, undisclosed AI use in writing in major media outlets, academic research, and public commentary. Once you see it you can't unsee it (especially with @pangram.com).
by Mel Bartley
Who would have ever thought it?
October 16, 2026 at 7:24 AM
Who would have ever thought it?
by Ian Bremmer
we need an independent agency that can 1) identify, 2) communicate, and 3) contain ai threats.
we don’t have that process.
we don’t even have the beginnings of it.
@gzeromedia.com
www.youtube.com/shorts/R1crS...
we don’t have that process.
we don’t even have the beginnings of it.
@gzeromedia.com
www.youtube.com/shorts/R1crS...
AI is becoming a superweapon | GZERO Media
YouTube video by GZERO Media
www.youtube.com
October 15, 2026 at 6:48 PM
we need an independent agency that can 1) identify, 2) communicate, and 3) contain ai threats.
we don’t have that process.
we don’t even have the beginnings of it.
@gzeromedia.com
www.youtube.com/shorts/R1crS...
we don’t have that process.
we don’t even have the beginnings of it.
@gzeromedia.com
www.youtube.com/shorts/R1crS...