Pancho Lopez Ph.D.
@ajgi.bsky.social
Reposted by Pancho Lopez Ph.D.
The surge of AI slop, deceptive synthetic videos & images flooding our feeds, is driven in large part by Agentic AI Accounts (AAAs). aiforensics.org/work/agentic...
Key findings:
-Over 43,000 mostly AI-made posts generated 4.5 billion views
-More than 65% of accounts were created in early 2025
1/
Key findings:
-Over 43,000 mostly AI-made posts generated 4.5 billion views
-More than 65% of accounts were created in early 2025
1/
aiforensics.org
December 27, 2025 at 6:33 PM
The surge of AI slop, deceptive synthetic videos & images flooding our feeds, is driven in large part by Agentic AI Accounts (AAAs). aiforensics.org/work/agentic...
Key findings:
-Over 43,000 mostly AI-made posts generated 4.5 billion views
-More than 65% of accounts were created in early 2025
1/
Key findings:
-Over 43,000 mostly AI-made posts generated 4.5 billion views
-More than 65% of accounts were created in early 2025
1/
Reposted by Pancho Lopez Ph.D.
Again: the Purpose of LLM- type "AI" bullshit engines is not to tell you the truth or to lie to you, but to tell you something you are statistically determined to be more likely to accept, irrespective of facts— which makes them perfect for accelerating dis- & misinformation & persuasive propaganda…
AI dialogues shifted political support by 2 to 3 points in trials, eclipsing standard ads. Roughly one-third of this effect persisted for a month. However, models optimized for persuasion proved more factually error-prone.
#MLSky
#MLSky
AI chatbots can persuade voters to change their minds
Conversations with AI can sway people’s political views. Concerningly, a chatbot’s facts are not always accurate, especially when it supports right-wing positions.
www.nature.com
December 5, 2025 at 3:35 PM
Again: the Purpose of LLM- type "AI" bullshit engines is not to tell you the truth or to lie to you, but to tell you something you are statistically determined to be more likely to accept, irrespective of facts— which makes them perfect for accelerating dis- & misinformation & persuasive propaganda…
Reposted by Pancho Lopez Ph.D.
narrator: no, it cannot
A new paper suggests a photo can tell a recruiter much about an applicant’s personality
Should facial analysis help determine whom companies hire?
A new paper suggests a photo can tell a recruiter much about an applicant’s personality
econ.st
November 6, 2025 at 9:32 PM
narrator: no, it cannot
Reposted by Pancho Lopez Ph.D.
built on theft and at massive environmental cost and most importantly, numerous studies keep demonstrating that genAI tools often work a little better than by chance. how is it that we can’t expunge these slop machines out of our institutions and communities
www.bbc.co.uk/mediacentre/...
www.bbc.co.uk/mediacentre/...
Largest study of its kind shows AI assistants misrepresent news content 45% of the time – regardless of language or territory
An intensive international study was coordinated by the European Broadcasting Union (EBU) and led by the BBC
www.bbc.co.uk
October 24, 2025 at 10:00 AM
built on theft and at massive environmental cost and most importantly, numerous studies keep demonstrating that genAI tools often work a little better than by chance. how is it that we can’t expunge these slop machines out of our institutions and communities
www.bbc.co.uk/mediacentre/...
www.bbc.co.uk/mediacentre/...
Reposted by Pancho Lopez Ph.D.
LLMs’ “simulated reasoning” abilities are a “brittle mirage,” researchers find arstechnica.com/ai/2025/08/r...
LLMs’ “simulated reasoning” abilities are a “brittle mirage,” researchers find
Chain-of-thought AI “degrades significantly” when asked to generalize beyond training.
arstechnica.com
August 12, 2025 at 11:54 AM
LLMs’ “simulated reasoning” abilities are a “brittle mirage,” researchers find arstechnica.com/ai/2025/08/r...
Reposted by Pancho Lopez Ph.D.
positioned by Musk as an "unfiltered, boundary-pushing AI, the generator allows users to make NSFW content" techcrunch.com/2025/08/04/g...
Grok Imagine, xAI’s new AI image and video generator, lets you make NSFW content | TechCrunch
Elon Musk continues to position Grok as an unfiltered, boundary-pushing AI.
techcrunch.com
August 4, 2025 at 3:20 PM
positioned by Musk as an "unfiltered, boundary-pushing AI, the generator allows users to make NSFW content" techcrunch.com/2025/08/04/g...
Reposted by Pancho Lopez Ph.D.
The explicit plan is to ethnically cleanse Gaza with US help while the whole world watches.
(AXIOS) - The director of Israel's Mossad spy agency visited Washington this week seeking U.S. help in convincing countries to take hundreds of thousands of Palestinians from Gaza, two sources with knowledge of issue tell Axios.
@axios.com
www.axios.com/2025/07/18/i...
@axios.com
www.axios.com/2025/07/18/i...
July 18, 2025 at 5:07 PM
The explicit plan is to ethnically cleanse Gaza with US help while the whole world watches.
Reposted by Pancho Lopez Ph.D.
The bottom line, says @willie-agnew.bsky.social, a postdoctoral fellow in AI ethics at Carnegie Mellon University and one of the coauthors, is that “anything you put online can [be] and probably has been scraped.”
The paper: arxiv.org/pdf/2506.17185
The paper: arxiv.org/pdf/2506.17185
July 18, 2025 at 3:52 PM
The bottom line, says @willie-agnew.bsky.social, a postdoctoral fellow in AI ethics at Carnegie Mellon University and one of the coauthors, is that “anything you put online can [be] and probably has been scraped.”
The paper: arxiv.org/pdf/2506.17185
The paper: arxiv.org/pdf/2506.17185
Reposted by Pancho Lopez Ph.D.
Ever wished you could track any pixel’s 3D path in a video—with just one model?
SpatialTrackerV2 makes it real. This end-to-end, fully differentiable framework unlocks world-space 3D trajectories for arbitrary 2D pixels, all from monocular videos. No need for clunky offline estimators!
SpatialTrackerV2 makes it real. This end-to-end, fully differentiable framework unlocks world-space 3D trajectories for arbitrary 2D pixels, all from monocular videos. No need for clunky offline estimators!
July 12, 2025 at 5:15 PM
Ever wished you could track any pixel’s 3D path in a video—with just one model?
SpatialTrackerV2 makes it real. This end-to-end, fully differentiable framework unlocks world-space 3D trajectories for arbitrary 2D pixels, all from monocular videos. No need for clunky offline estimators!
SpatialTrackerV2 makes it real. This end-to-end, fully differentiable framework unlocks world-space 3D trajectories for arbitrary 2D pixels, all from monocular videos. No need for clunky offline estimators!
Reposted by Pancho Lopez Ph.D.
NEW: ProPublica spent 14 months investigating how the FDA gave substandard factories banned from the United States a special pass to keep sending drugs to Americans.
Here are the key takeaways from our reporting.
Here are the key takeaways from our reporting.
Seven Things to Know About ProPublica’s Investigation of the FDA’s Secret Gamble on Generic Drugs
ProPublica spent 14 months investigating the FDA’s oversight of foreign drugmakers that send medications to the U.S. These are the key takeaways.
www.propublica.org
June 23, 2025 at 12:32 PM
NEW: ProPublica spent 14 months investigating how the FDA gave substandard factories banned from the United States a special pass to keep sending drugs to Americans.
Here are the key takeaways from our reporting.
Here are the key takeaways from our reporting.
Reposted by Pancho Lopez Ph.D.
Check out these baddies waging a psyops campaign outside the hotel where ICE agents were *trying* to sleep.
Mariachi for the win!! 💃
Mariachi for the win!! 💃
June 21, 2025 at 8:19 PM
Check out these baddies waging a psyops campaign outside the hotel where ICE agents were *trying* to sleep.
Mariachi for the win!! 💃
Mariachi for the win!! 💃
Reposted by Pancho Lopez Ph.D.
Chip's got a new issue of ZDARSKY COMIC NEWS this week, and it might just be all about MANGA, with some very special guest appearances! substack.com/home/post/p-...
June 18, 2025 at 6:12 PM
Chip's got a new issue of ZDARSKY COMIC NEWS this week, and it might just be all about MANGA, with some very special guest appearances! substack.com/home/post/p-...
Reposted by Pancho Lopez Ph.D.
Palmer: Charges should be brought against each one of you
AOC: Raising the prospect of charges against witnesses is a form of intimidation is it not?
Palmer: I strike those words
AOC: Raising the prospect of charges against witnesses is a form of intimidation is it not?
Palmer: I strike those words
June 12, 2025 at 5:49 PM
Palmer: Charges should be brought against each one of you
AOC: Raising the prospect of charges against witnesses is a form of intimidation is it not?
Palmer: I strike those words
AOC: Raising the prospect of charges against witnesses is a form of intimidation is it not?
Palmer: I strike those words
Reposted by Pancho Lopez Ph.D.
Reposted by Pancho Lopez Ph.D.
Reposted by Pancho Lopez Ph.D.
this is what quality, meaningful, and consequential investigative journalism looks like
New from 404 Media: Delta, United, American etc don't want you to know they sold your flight data to DHS. The country's airlines sell bulk names and flight itineraries to DHS to be searched without a warrant. Contract tells DHS to not reveal where this data came from www.404media.co/airlines-don...
Airlines Don't Want You to Know They Sold Your Flight Data to DHS
A contract obtained by 404 Media shows that an airline-owned data broker forbids the feds from revealing it sold them detailed passenger data.
www.404media.co
June 10, 2025 at 9:58 PM
this is what quality, meaningful, and consequential investigative journalism looks like
Reposted by Pancho Lopez Ph.D.
Shortcuts that companies with AI-derived drugs are taking, not testing the molecules in vivo
@science.org
science.org/doi/10.1126/...
@science.org
science.org/doi/10.1126/...
May 30, 2025 at 3:30 PM
Shortcuts that companies with AI-derived drugs are taking, not testing the molecules in vivo
@science.org
science.org/doi/10.1126/...
@science.org
science.org/doi/10.1126/...
Reposted by Pancho Lopez Ph.D.
"our counter-intuitive results demonstrate ways in which common interpretations of Large Reasoning Models may be anthropomorphizations or simplifications" arxiv.org/abs/2505.13775
Beyond Semantics: The Unreasonable Effectiveness of Reasonless Intermediate Tokens
Recent impressive results from large reasoning models have been interpreted as a triumph of Chain of Thought (CoT), and especially of the process of training on CoTs sampled from base LLMs in order to...
arxiv.org
June 1, 2025 at 1:30 PM
"our counter-intuitive results demonstrate ways in which common interpretations of Large Reasoning Models may be anthropomorphizations or simplifications" arxiv.org/abs/2505.13775
Reposted by Pancho Lopez Ph.D.
AI chatbots have been rolled out across 100s of white-collar workplaces but their effect on hours & pay has been negligible accrdg NBER. On average, employees saved 3% of their time, while just 3%-7% of their productivity gains came back to them in the form of higher pay fortune.com/2025/05/18/a...
Study looking at AI chatbots in 7,000 workplaces finds 'no significant impact on earnings or recorded hours in any occupation'
Despite AI’s promise to revolutionize white-collar work, most workers are using it sparingly—or hiding it from their boss.
fortune.com
May 29, 2025 at 11:05 AM
AI chatbots have been rolled out across 100s of white-collar workplaces but their effect on hours & pay has been negligible accrdg NBER. On average, employees saved 3% of their time, while just 3%-7% of their productivity gains came back to them in the form of higher pay fortune.com/2025/05/18/a...
Reposted by Pancho Lopez Ph.D.
“Billionaires building bunkers to survive an AI apocalypse, attempting to disrupt death through cryopreservation, scouting the planet for pop-up cities and network states — are not reliable stewards of the collective good.”
— @ruha9.bsky.social
— @ruha9.bsky.social
May 23, 2025 at 5:10 AM
“Billionaires building bunkers to survive an AI apocalypse, attempting to disrupt death through cryopreservation, scouting the planet for pop-up cities and network states — are not reliable stewards of the collective good.”
— @ruha9.bsky.social
— @ruha9.bsky.social
Reposted by Pancho Lopez Ph.D.
there's a massive rush to insert AI agents into everything but who is vetting these systems before they're dumped into vulnerable groups like children
WSJ test conversations “found that both Meta’s official AI helper, called Meta AI, and a vast array of user-created chatbots will engage in and sometimes escalate discussions that are decidedly sexual—even when the users are underage or the bots are programmed to simulate the personas of minors.“
Meta’s ‘Digital Companions’ Will Talk Sex With Users—Even Children
Chatbots on Instagram, Facebook and WhatsApp are empowered to engage in ‘romantic role-play’ that can turn explicit. Some people inside the company are concerned.
www.wsj.com
April 27, 2025 at 8:35 PM
there's a massive rush to insert AI agents into everything but who is vetting these systems before they're dumped into vulnerable groups like children
Reposted by Pancho Lopez Ph.D.
🧵 THREAD: A federal whistleblower just dropped one of the most disturbing cybersecurity disclosures I’ve ever read.
He's saying DOGE came in, data went out, and Russians started attempting logins with new valid DOGE passwords
Media's coverage wasn't detailed enough so I dug into his testimony:
He's saying DOGE came in, data went out, and Russians started attempting logins with new valid DOGE passwords
Media's coverage wasn't detailed enough so I dug into his testimony:
April 18, 2025 at 12:10 AM
🧵 THREAD: A federal whistleblower just dropped one of the most disturbing cybersecurity disclosures I’ve ever read.
He's saying DOGE came in, data went out, and Russians started attempting logins with new valid DOGE passwords
Media's coverage wasn't detailed enough so I dug into his testimony:
He's saying DOGE came in, data went out, and Russians started attempting logins with new valid DOGE passwords
Media's coverage wasn't detailed enough so I dug into his testimony:
