Graham Webster | Lightnews

Graham Webster @gwbstr.com · 4d

On the longest night of
the year,

the days get brighter
from here.

Graham Webster @gwbstr.com · 6d

also worse off: freedom of expression, rule of law

Dare Obasanjo @carnage4life.bsky.social · 6d

Democrats really scored an own goal with the TikTok ban. The company has gone from being owned by China to being owned by China and Trump allies.

If the concern was TikTok pushing problematic politics, Dems are actually worse off.

TikTok Deal Done And It’s Somehow The Shittiest Possible Outcome, Making Everything Worse

There were rumblings about this for a while, but it looks like the Trump TikTok deal is done, and it’s somehow the worst of all possible outcomes, amazingly making all of the biggest criticis…

www.techdirt.com

Victor Shih @vicshih.bsky.social · 6d

MIIT's priority AI techs 2025 "complex reasoning models, embodied intelligence models, edge-side models; large-scale computing center, multi-cluster computing power scheduling and coordination, computing center energy consumption management 1/n

Dare Obasanjo @carnage4life.bsky.social · 6d

Democrats really scored an own goal with the TikTok ban. The company has gone from being owned by China to being owned by China and Trump allies.

If the concern was TikTok pushing problematic politics, Dems are actually worse off.

TikTok Deal Done And It’s Somehow The Shittiest Possible Outcome, Making Everything Worse

There were rumblings about this for a while, but it looks like the Trump TikTok deal is done, and it’s somehow the worst of all possible outcomes, amazingly making all of the biggest criticis…

www.techdirt.com

Graham Webster @gwbstr.com · 6d

- Gives a new inflection to the flexibly used term “agent“

- Seems like the type of thing one might try to use open-weight (and for now likely Chinese) models for, lest a closed vendor decide to cut off the gestapo‘s API key.

404 Media @404media.co · 7d

NEW: ICE has paid hundreds of thousands of dollars to a company that makes “AI agents” to rapidly track down targets.

AI Solutions 87 did not respond to a request for comment on how its AI agents work.

ICE Contracts Company Making Bounty Hunter AI Agents

AI Solutions 87 says on its website its AI agents “deliver rapid acceleration in finding persons of interest and mapping their entire network.”

www.404media.co

QuoProQuid @quoproquid.bsky.social · 6d

i recognize it's just a performance arts center and we've all collectively stopped giving a shit but changing the signage of the kennedy center is explicitly prohibited by law

404 Media @404media.co · 7d

NEW: ICE has paid hundreds of thousands of dollars to a company that makes “AI agents” to rapidly track down targets.

AI Solutions 87 did not respond to a request for comment on how its AI agents work.

ICE Contracts Company Making Bounty Hunter AI Agents

AI Solutions 87 says on its website its AI agents “deliver rapid acceleration in finding persons of interest and mapping their entire network.”

www.404media.co

Graham Webster @gwbstr.com · 8d

taste*

Graham Webster @gwbstr.com · 8d

This kind of thing is so deeply disappointing from keepers of scholarship. It may not be my tasted but I can truly understand a lot of businesses experimenting. But the non-profit source-of-authority academic knowledge people? Yikes.

Is there a term for societal "AI psychosis"?

Dr. Casey Fiesler @cfiesler.bsky.social · 9d

The ACM Digital Library, where a LOT of computing-related research is published (I'd say at least 75% of my own publications), is now not only providing (without consent of the authors and without opt-in by readers) AI-generated summaries of papers, but they appear as the *default* over abstracts.

Screenshot of a paper entry:
Fictional Failures and Real-World Lessons: Ethical Speculation Through Design Fiction on Emotional Support Conversational AI
Authors: Faye Kollig, Jessica Pater, Fayika Farhat Nova, Casey Fiesler
(There are tabs with "abstract" and "summary" and "summary" is selected.)

Graham Webster @gwbstr.com · 8d

This has been today’s installment of AI takes that satisfy no one.

Graham Webster @gwbstr.com · 8d

For those who know the model outputs are often wrong, it’s similar to those who know the details on Wikipedia are often wrong. When the stakes are low, convenience wins, and it’s not worth being irate.

The big problem is where stakes are high or people don’t internalize the frequent errors.

Graham Webster @gwbstr.com · 8d

I’m often at a loss when people who know better proudly report they ”looked up” some information on ChatGPT. This is in some cases a retreat from former commitments to reliable sources, but in most cases it seems more likely to just reflect how little people cared about accuracy all along.

Dr. Casey Fiesler @cfiesler.bsky.social · 9d

The ACM Digital Library, where a LOT of computing-related research is published (I'd say at least 75% of my own publications), is now not only providing (without consent of the authors and without opt-in by readers) AI-generated summaries of papers, but they appear as the *default* over abstracts.

Stanford CISAC @stanfordcisac.bsky.social · 9d

Have Chinese AI models pulled ahead of their global counterparts? In a new issue brief from @stanfordhai.bsky.social, CISAC research scholar @gwbstr.com and colleagues analyze China’s open-weight model ecosystem and its global policy implications.
hai.stanford.edu/policy/beyon...