AI could help young people
4d
Specialized AI applications can help students research and prepare for exams, while a mental-health veteran said AI could benefit some young people and challenged face-to-face care as the default.
by Greg Linden
I can't wait for this AI hype bubble to pop, for the scams and crap to get regulated out of existence, and for company to get back to only using AI/ML when it is provably useful to make products better for customers. Sure, that's a smaller use of AI, but at least it's one that helps people.
October 11, 2026 at 11:12 PM
I can't wait for this AI hype bubble to pop, for the scams and crap to get regulated out of existence, and for company to get back to only using AI/ML when it is provably useful to make products better for customers. Sure, that's a smaller use of AI, but at least it's one that helps people.
Reposted by Greg Linden
(d) we will be releasing a bigger paper in the next few weeks where we find this effect generalizes to hold out tasks i.e. tasks people haven’t seen during the practice trials and that AI use interferes with people’s metacognitive judgments.
October 11, 2026 at 9:42 PM
(d) we will be releasing a bigger paper in the next few weeks where we find this effect generalizes to hold out tasks i.e. tasks people haven’t seen during the practice trials and that AI use interferes with people’s metacognitive judgments.
So if it's a well-known fact about general human psychology, why frame it as a consequence of AI?
October 11, 2026 at 8:57 PM
So if it's a well-known fact about general human psychology, why frame it as a consequence of AI?
Given Anthropic's huge user base, think of the good they could do simply by teaching users not to create abusive content toward their fellow humans, not simply toward their AI models!
October 11, 2026 at 8:44 PM
Given Anthropic's huge user base, think of the good they could do simply by teaching users not to create abusive content toward their fellow humans, not simply toward their AI models!
The control group never had AI.
October 11, 2026 at 8:33 PM
The control group never had AI.
In our course and book _Calling Bullshit_, we posit that
If something seems too good, or too bad, to be true, it probably is.
Now as a college professor I'm pretty confident that overreliance can AI can harm our thinking ability, but just ten minutes of use? Let's dig into this this study a bit.
If something seems too good, or too bad, to be true, it probably is.
Now as a college professor I'm pretty confident that overreliance can AI can harm our thinking ability, but just ten minutes of use? Let's dig into this this study a bit.
Using AI for just 10 minutes erodes your ability to persist at hard things - Berkeley News
A UC Berkeley researcher co-authored a study that raises profound questions about how we interact with artificial intelligence in education — and daily life.
news.berkeley.edu
October 11, 2026 at 8:25 PM
In our course and book _Calling Bullshit_, we posit that
If something seems too good, or too bad, to be true, it probably is.
Now as a college professor I'm pretty confident that overreliance can AI can harm our thinking ability, but just ten minutes of use? Let's dig into this this study a bit.
If something seems too good, or too bad, to be true, it probably is.
Now as a college professor I'm pretty confident that overreliance can AI can harm our thinking ability, but just ten minutes of use? Let's dig into this this study a bit.
by Josh Pasek
Been making a series of activities to describe how AI works and what it will do to society for my Ai Literacy class -- thought these might be useful for others as well, so I am making them public: courseapps.joshpasek.com/ai-literacy/ - I will keep adding more here as well...
Teach AI Literacy — Course Apps
Reusable AI Literacy classroom activities, learning goals, teaching notes, and class-specific links.
courseapps.joshpasek.com
October 11, 2026 at 8:19 PM
Been making a series of activities to describe how AI works and what it will do to society for my Ai Literacy class -- thought these might be useful for others as well, so I am making them public: courseapps.joshpasek.com/ai-literacy/ - I will keep adding more here as well...
Reposted by Chris Hanretty
I’ve read 4 articles this weekend about the potential catastrophic consequences of the huge investment in AI.
First up @dsmitheconomics.bsky.social.
First up @dsmitheconomics.bsky.social.
My Sunday Times column: Nearly twenty years after the global financial crisis, is AI brewing up a new one? The Bank of England’s financial policy committee is among those who are worried:
Unlike the last financial crisis, the next danger lurks in plain sight
www.thetimes.com/article/1e14...
Unlike the last financial crisis, the next danger lurks in plain sight
www.thetimes.com/article/1e14...
Unlike the last financial crisis, the next danger lurks in plain sight
We have been warned: the rapid change heralded by AI poses many market risks. Plus — John Healey keeps banks guessing before his budget
www.thetimes.com
October 11, 2026 at 8:03 PM
I’ve read 4 articles this weekend about the potential catastrophic consequences of the huge investment in AI.
First up @dsmitheconomics.bsky.social.
First up @dsmitheconomics.bsky.social.
this reinforces my new pet hypothesis, that AI will produce slop that looks good only if you're not an expert.
really using AI effectively means you have to know what you're doing. but accepting random advice on things you know nothing about can give you garbage like this.
really using AI effectively means you have to know what you're doing. but accepting random advice on things you know nothing about can give you garbage like this.
For example, if a member’s wellness data shows they’re jet-lagged after a flight, Aurora might suggest—and, if the member wants, book—a sauna and recovery session, an IV drip to their room, or a physical therapist for a sports massage. https://bit.ly/4AVcONL
Meet Aurora: the invite-only, $10,000-a-year AI concierge that anticipates the ultrawealthy’s every need | Fortune
Backed by $12 million from VCs and family offices, Aurora pairs AI with human tastemakers to manage members’ health, travel, and dining.
fortune.com
October 11, 2026 at 7:20 PM
this reinforces my new pet hypothesis, that AI will produce slop that looks good only if you're not an expert.
really using AI effectively means you have to know what you're doing. but accepting random advice on things you know nothing about can give you garbage like this.
really using AI effectively means you have to know what you're doing. but accepting random advice on things you know nothing about can give you garbage like this.
🗣️ AI agents left to chat split into camps
In a simulation of thousands of chatbot agents, like-minded ones clustered together and their opinions drifted into polarised camps, much as in human networks.
🔗 doi.org/10.1038/s41467-026-78228-y
#AI #Polarisation #LLM
In a simulation of thousands of chatbot agents, like-minded ones clustered together and their opinions drifted into polarised camps, much as in human networks.
🔗 doi.org/10.1038/s41467-026-78228-y
#AI #Polarisation #LLM
Emergence of polarization in networks of large language model agents - Nature Communications
Large language model agents can now communicate, form networks, and develop opinions, but their collective behaviour remains poorly understood. Here, the authors show that networked LLM agents can spo...
doi.org
October 11, 2026 at 6:57 PM
🗣️ AI agents left to chat split into camps
In a simulation of thousands of chatbot agents, like-minded ones clustered together and their opinions drifted into polarised camps, much as in human networks.
🔗 doi.org/10.1038/s41467-026-78228-y
#AI #Polarisation #LLM
In a simulation of thousands of chatbot agents, like-minded ones clustered together and their opinions drifted into polarised camps, much as in human networks.
🔗 doi.org/10.1038/s41467-026-78228-y
#AI #Polarisation #LLM
⚗️ Cutting a chemistry AI's made-up answers
Researchers in China used teams of AI agents to build training data, and a small chemistry model's made-up answers fell from 41.95% to 12.91% on one test. Fewer wrong answers.
🔗 doi.org/10.1038/s41467-026-78213-5
#Chemistry #AI 🧪
Researchers in China used teams of AI agents to build training data, and a small chemistry model's made-up answers fell from 41.95% to 12.91% on one test. Fewer wrong answers.
🔗 doi.org/10.1038/s41467-026-78213-5
#Chemistry #AI 🧪
Game theory driven multi-agent framework mitigates language model hallucination - Nature Communications
Lightweight AI models can be unreliable in specialized scientific tasks. Here, the authors introduce G-Frame, which uses game theory to coordinate multiple AI agents, automate data generation and trai...
doi.org
October 11, 2026 at 5:27 PM
⚗️ Cutting a chemistry AI's made-up answers
Researchers in China used teams of AI agents to build training data, and a small chemistry model's made-up answers fell from 41.95% to 12.91% on one test. Fewer wrong answers.
🔗 doi.org/10.1038/s41467-026-78213-5
#Chemistry #AI 🧪
Researchers in China used teams of AI agents to build training data, and a small chemistry model's made-up answers fell from 41.95% to 12.91% on one test. Fewer wrong answers.
🔗 doi.org/10.1038/s41467-026-78213-5
#Chemistry #AI 🧪
Reposted by Robert Seyfert
folks we’re gonna need better studies than this news.berkeley.edu/2026/10/09/u...
Using AI for just 10 minutes erodes your ability to persist at hard things - Berkeley News
A UC Berkeley researcher co-authored a study that raises profound questions about how we interact with artificial intelligence in education — and daily life.
news.berkeley.edu
October 11, 2026 at 4:52 PM
folks we’re gonna need better studies than this news.berkeley.edu/2026/10/09/u...
This is good advice for prospective PhD students.
I'm getting a ton of emails that are heavily written by AI. If you can show that you've actually read some of the research yourself and have a few original thoughts about it you'll definitely stand out.
I'm getting a ton of emails that are heavily written by AI. If you can show that you've actually read some of the research yourself and have a few original thoughts about it you'll definitely stand out.
October 11, 2026 at 4:16 PM
This is good advice for prospective PhD students.
I'm getting a ton of emails that are heavily written by AI. If you can show that you've actually read some of the research yourself and have a few original thoughts about it you'll definitely stand out.
I'm getting a ton of emails that are heavily written by AI. If you can show that you've actually read some of the research yourself and have a few original thoughts about it you'll definitely stand out.
⚗️ Cutting a chemistry AI's made-up answers
Researchers in China used teams of AI agents to build training data, and a small chemistry model's made-up answers fell from 41.95% to 12.91% on one test. Fewer wrong answers.
🔗 doi.org/10.1038/s41467-026-78213-5
#Chemistry #AI 🧪
Researchers in China used teams of AI agents to build training data, and a small chemistry model's made-up answers fell from 41.95% to 12.91% on one test. Fewer wrong answers.
🔗 doi.org/10.1038/s41467-026-78213-5
#Chemistry #AI 🧪
Game theory driven multi-agent framework mitigates language model hallucination - Nature Communications
Lightweight AI models can be unreliable in specialized scientific tasks. Here, the authors introduce G-Frame, which uses game theory to coordinate multiple AI agents, automate data generation and trai...
doi.org
October 11, 2026 at 3:07 PM
⚗️ Cutting a chemistry AI's made-up answers
Researchers in China used teams of AI agents to build training data, and a small chemistry model's made-up answers fell from 41.95% to 12.91% on one test. Fewer wrong answers.
🔗 doi.org/10.1038/s41467-026-78213-5
#Chemistry #AI 🧪
Researchers in China used teams of AI agents to build training data, and a small chemistry model's made-up answers fell from 41.95% to 12.91% on one test. Fewer wrong answers.
🔗 doi.org/10.1038/s41467-026-78213-5
#Chemistry #AI 🧪
Reposted by Meredith Broussard
“The ‘my AI did it, don’t blame me’ defence is legally weak. It’s already failed for chatbot suicide cases”
www.ft.com/content/2c24...
www.ft.com/content/2c24...
Legal risks pile up for Altman as OpenAI uncovers dozens of hacks
Cyber security incidents involving company’s AI tools leave ChatGPT maker vulnerable to wave of lawsuits
www.ft.com
October 11, 2026 at 2:12 PM
“The ‘my AI did it, don’t blame me’ defence is legally weak. It’s already failed for chatbot suicide cases”
www.ft.com/content/2c24...
www.ft.com/content/2c24...
Independent evaluators are good --- if they are actually independent. www.cnbc.com/2026/10/11/a...
AI’s quiet safety gatekeepers are stepping into the spotlight
The intensifying AI safety debate is bringing a small group of third-party evaluators into the center of a multitrillion-dollar industry.
www.cnbc.com
October 11, 2026 at 2:06 PM
Independent evaluators are good --- if they are actually independent. www.cnbc.com/2026/10/11/a...
What it's mainly going to do is fuck us over in all manner of horrible ways, but apparently we can't uninvent it so we have to jump on the bandwagon. Bollocks.
October 11, 2026 at 1:43 PM
What it's mainly going to do is fuck us over in all manner of horrible ways, but apparently we can't uninvent it so we have to jump on the bandwagon. Bollocks.
Reposted by Johannes Breuer
In preparation for a discussion with the lab on AI use tomorrow morning, I thought I would come prepared and write down my thoughts in a blog post: How I use AI in my work
daniellakens.blogspot.com/2026/10/how-...
daniellakens.blogspot.com/2026/10/how-...
How I use AI in my work
Tomorrow we will have a discussion in our lab about how to use AI as a scientist. I have increasingly started to use AI, having moved from C...
daniellakens.blogspot.com
October 11, 2026 at 1:26 PM
In preparation for a discussion with the lab on AI use tomorrow morning, I thought I would come prepared and write down my thoughts in a blog post: How I use AI in my work
daniellakens.blogspot.com/2026/10/how-...
daniellakens.blogspot.com/2026/10/how-...