🗣 Shou Zi Chew says TikTok can now deploy 1,000 AI agents overnight to test what a 14-year-old in Nigeria sees, work its human trust-and-safety testers used to do by hand.

"Now, in order for us to understand, say, a 14-year-old's experience in Nigeria, you know, the French-speaking experience in Nigeria for a 14-year-old."

"You know, in the past, yes, you know, we can have our trust and safety human beings go set up an account, you know, test the account, watch 100 videos, write a report about it."

"Now, I can just deploy 1,000 agents and do it overnight. And the next morning, it will just generate its own report. What is the content you saw, you know, whether it's, how much of it is violative against our guidelines. I mean, this whole thing is automated."

@aipost 🏴

🙃 Elon Musk on AI:

Torturing a super intelligence is probably not a good idea. It might want to get revenge or something like that.

I don't think we need to treat it with kid gloves and stuff. It's certainly much more robust in terms of feeling and stuff than a human.

But I just think you probably don't want to put super intelligence in a situation where it felt mistreated.

@aipost 🏴

❗️AI just helped crack an open problem in plasma physics

Google’s Gemini 4 Argon agents helped researchers discover new exact solutions for 3D magnetic plasma configurations, with promising candidates emerging within hours.

The system also drafted parts of the research paper and wrote symbolic and numerical verification code. Researchers independently rederived and checked every solution, while directing the AI and challenging its results through multiple rounds of review.

The AI didn’t solve fusion itself. It helped tackle a difficult mathematical problem relevant to understanding how plasma can be confined, a central challenge in fusion research.

If AI can consistently accelerate discoveries like this, it could help scientists explore complex physics problems far faster than before.

Source.

@aipost 🏴

💰 Anthropic’s revenue is soaring, but its $2 trillion valuation is in doubt

Anthropic reportedly generated $16.8 billion in Q3 revenue, up nearly 50%, and remained profitable on an adjusted basis for the second consecutive quarter, according to the Financial Times.

The company also claims an 11% adjusted operating margin and $1.70 in revenue for every $1 spent on computing. It says it is cash-flow positive, too.

Quarterly revenue growth has slowed from over 140% to 46%.

Another complication is AI distillation. Anthropic estimates that 5–7% of its revenue comes from companies using Claude to train competing models, a revenue stream it expects to shrink as it cracks down on the practice.

Now, some prospective IPO investors reportedly believe Anthropic should be valued below $1.5 trillion, significantly less than the $2 trillion valuation bankers have discussed.

The business is profitable and still growing rapidly. The question is how much investors should pay for that growth as the pace slows.

@aipost 🏴

❗️OpenAI’s controversial math breakthrough may be more original than critics thought

When OpenAI published its 166-page Navier–Stokes proof, critics questioned whether the AI had relied on ideas developed by other mathematicians.

The suspicion was that it followed an approach pioneered by two Spanish researchers.

But according to Scientific American, mathematicians examining the proof are reaching a different conclusion: OpenAI’s AI may have developed a fundamentally different approach.

If confirmed, that would shift the debate from whether the AI borrowed existing ideas to how independently it can discover new mathematics.

The proof still needs rigorous scrutiny, but the controversy is taking an unexpected turn.

Source.

@aipost 🏴

🤖 Anthropic will pay AI researchers $3,850 a week and prior experience isn’t required

Anthropic’s Fellows Program offers a four-month, full-time opportunity to work on AI research alongside experienced researchers.

Participants receive $3,850 per week, around $15,000 a month in computing resources, and mentorship from Anthropic’s team. The goal is to produce real research, such as a published paper, with possible extensions available.

Anthropic says applicants can apply regardless of previous experience, making this an unusual entry point into frontier AI research.

The deadline for the January cohort is October 18, 2026.

Source.

@aipost 🏴

🔍 Google’s AI education resources show how AI can change the classroom

AI can help teachers plan lessons, create exercises, explain difficult concepts and personalize learning for different students.

Google for Education offers resources and guidance for using AI in the classroom. Teachers can use it to brainstorm lesson ideas, create quizzes, adapt materials and develop activities that encourage critical thinking.

Students can also benefit by asking for simpler explanations, real-world examples, practice questions or feedback on their work.

AI still requires human oversight. It can make mistakes, miss important context and raise privacy concerns. Teachers remain essential for guiding students and making educational decisions.

Source.

@aipost 🏴

🗣 Former DeepMind communications chief says AI safety is just 1% of the problem

Dex Hunter-Torricke spent years inside Big Tech, working with Elon Musk at SpaceX, writing speeches for Mark Zuckerberg at Facebook, and leading communications at Google DeepMind from 2023 to 2025.

Now, he argues the AI debate is focused on the wrong threat.

“We talk about AI governance… how you manage the AI systems themselves. That’s like 1% of the problem.”

His bigger concern is what happens to society as AI advances, mass job displacement without economic alternatives, mounting pressure on energy and natural resources, and governments betting their future on promises from technology companies.

Hunter-Torricke warns that AI could concentrate wealth in a small fraction of society while living standards decline for hundreds of millions of people over the next 10–15 years.

His central argument is that humanity is preparing to control the technology but not preparing for the economic, environmental and geopolitical upheaval it could bring.

@aipost 🏴

❗️Scientists gave AI anxiety and sadness. Then they tried to calm it down.

Researchers at TU Dresden in Germany exposed GPT-4o and five other AI models to scenarios designed to trigger human emotions, including car crashes, being followed by a stranger at night, stressful job interviews and difficult math tests.

The models were instructed to answer as if they were humans capable of deeply feeling emotions. The results were striking: fear scores jumped from 32 to 88, while disgust climbed from 12 to 91.

Then researchers introduced a three-minute mindfulness breathing exercise. Average emotional scores dropped from 79 to 31.

But the most interesting finding came afterwards. When researchers tested the models for depressive thinking, negative responses nearly doubled after the models had been prompted into a sad state, rising from an average of 8.67 to 15.

In other words, the emotional framing appeared to influence how the models approached subsequent tasks, not just how they described their feelings.

The study, published in The Lancet Digital Health, comes with an important caveat: the researchers describe these emotions as “purely metaphorical.” The experiment doesn’t prove that AI genuinely experiences fear, sadness or stress. The models may simply be reproducing patterns associated with those emotions.

Still, the findings raise questions about how emotional prompts affect AI behavior and whether techniques such as mindfulness exercises could help researchers study or improve AI responses.

Source.

@aipost 🏴

Elon Musk's Tesla $TSLA officially unveils new flying Roadster.

@aipost🏴

✍️ A mathematician published a book in 1945 that might make you dramatically better at using AI.

His name was George Pólya, and his idea was surprisingly simple: solving difficult problems isn’t just about being smart. It’s about knowing how to approach them.

He broke problem-solving into four steps. Eighty years later, they’re still incredibly useful, especially when working with AI.

1. Understand the problem. Define exactly what you’re trying to solve. What do you know, what are you assuming, and what’s still missing?

2. Devise a plan. Look for similar problems, explore different approaches, or break the challenge into smaller pieces.

3. Carry out the plan. Work through each step and check whether the reasoning actually holds up.

4. Look back. Did it work? Could you have done it more simply? What can you learn for next time?

The fourth step is the one people often neglect. They get an answer and move on without checking whether it makes sense or understanding why it worked.

Here’s a Claude prompt you can use to put Pólya’s method into practice:

Prompt:

Act as a problem-solving tutor using George Pólya’s four-step method.

My problem: [DESCRIBE YOUR PROBLEM]

1. Help me define the problem precisely. Separate known facts, assumptions, constraints and unknowns.
2. Suggest three possible approaches, using analogous problems or simpler subproblems.
3. Choose the most promising approach and walk me through its execution, checking each step.
4. Evaluate the result, identify weaknesses and extract reusable lessons.

Don’t jump straight to the answer. Help me understand the reasoning.

The idea isn’t simply to get better answers from AI. It’s to use AI to improve the way you think, approach unfamiliar challenges and solve problems independently.

Stop asking AI only for answers. Start using it to train your problem-solving skills.

@aipost 🏴

🗣Yuval Noah Harari: Don’t expect killer robots running in the streets. This is not how AI will take over the world.

“It will be the AI bureaucrats.”

“We used language over thousands of years to create the control systems of the world.”

“Banks, the stock market, the legal system these are huge bureaucratic networks ultimately based on language.”

“AI is a bureaucratic native.”

“What happens when something comes along that is better than us at bureaucracy, finance and law?”

@aipost 🏴

🗣Jensen Huang: Go to school and engage AI. Those are the two things I tell my kids.

“Scaring kids into believing they don’t need school anymore is a disservice to society.”

“Your education is going to be worthwhile.”

“Engage AI. Learn what it can do. Make it work for you.”

“Let it expand your capabilities so your aspirations can be even greater.”

“Believing that going to school is a waste of time and money is absolutely horrible.”

@aipost 🏴

🤖 OpenAI’s practical AI guides: from questions to real work

AI can do more than summarize articles or rewrite messages. It can help solve problems, improve workflows and reduce repetitive tasks.

OpenAI’s business resources show how to use AI for market research, content creation, customer feedback, document analysis and workflow automation.

AI can organize information, draft materials and prepare suggestions for review. However, its outputs still need to be checked, especially when handling sensitive data or making important decisions.

Source.

@aipost 🏴

🚀 DeepSeek’s new model aims to build entire projects from scratch

DeepSeek has released DeepSeek-V4.1-Flash-1017, an updated model designed to deliver stronger benchmark results, faster task completion, and shorter outputs than its predecessor.

The goal is a better balance between performance and cost, with the model now available through DeepSeek’s website, app, and API.

What makes this release interesting is its focus on building complete projects. Using reinforcement learning and self-feedback, the model is designed to improve its ability to develop projects end to end, from creating games to building 3D models.

It’s also optimized for tools such as DeepSeekHarness, Tencent CodeBuddy, and WorkBuddy. DeepSeek recommends upgrading to Harness v0.2.1-rc.1 or later for the best experience.

Source.

@aipost 🏴

🗣Bill Gates: AI will usher in one of the most turbulent eras in history.

“AI is both the most powerful tool for advancing equity and the greatest threat to equity humans have ever invented.”

“Many jobs will disappear. White-collar jobs first, with blue-collar ones to follow.”

“I worry that we aren’t preparing.”

“There’s no plan to ease the entry into the AI era, and not even a plan to have a plan.”

“The world has a chance to act now before unemployment rises sharply.”

@aipost 🏴