Your daily reminder: o4-mini-high is still the best in coding so far. It's faster and cheaper than other oAI models

Sam also recently announced that message limits will be doubled. Another interesting point: gemini 2.5 flash performs slightly better than 2.5 pro in coding

@aipost 🪙 | Our X 🥇

🇷🇺 Putin: Whoever masters AI faster will have huge advantages

"You have to understand this, the possibilities of using artificial intelligence are simply colossal. Whoever begins to master these technologies faster, in this case in military affairs, will have huge advantages on the battlefield. We have other ideas, other developments, we'll talk about this seperately."

@aipost 🪙 | Our X 🥇

Geoffrey Hinton says the more we understand how AI and the brain actually work, the less human thinking looks like logic. We're not reasoning machines, he says. We're analogy machines. We think by resonance, not deduction.

“We're much less rational than we thought.”

@aipost 🪙 | Our X 🥇

⏺ OpenAI insiders warn of a potential decline in trust.

OpenAI is losing key safety leaders, including Ilya Sutskever and Jan Leike, who have resigned or been forced out, resulting in a weakened superalignment team and disbanded AGI Readiness efforts. Insiders report a cultural shift toward commercialization under Sam Altman’s leadership, with safety resources significantly reduced. Contracts reportedly silenced dissent until whistleblowers emerged, highlighting concerns about ignored safety priorities.

As OpenAI accelerates the release of powerful models like GPT-4o, critical voices are disappearing, raising questions about the lab's ability to manage the risks associated with AGI. Daniel Kokotajlo, a former governance team member, expressed lost trust in OpenAI's leadership regarding responsible AGI management.

@aipost 🪙 | Our X 🥇

⏺ OpenAI made Deep Research faster and available for free users

Ever asked a bunch of intelligent questions and received the response, “You’ve hit your limit”? Well, that’s it. OpenAI has unveiled a more user-friendly and efficient version of their deep research tool, allowing you to continue asking profound questions without encountering any restrictions.

Moreover, the best part is that you don’t even need to pay for this service. Even free users can access this enhanced version. For once, financial constraints won’t prevent you from accessing advanced knowledge.

@aipost 🪙 | Our X 🥇

⏺ OpenAI is being accused of straying from its original mission, prompting a coalition to demand a halt to its restructuring efforts.

A coalition of former insiders and AI watchdogs has accused OpenAI of betraying humanity. In a scathing letter, they criticized the company for abandoning its original mission to develop AGI for the public good by shifting control to a profit-driven structure.

The proposed change would transform OpenAI’s capped-profit model into a public benefit corporation, potentially allowing for uncapped profits, reduced public accountability, and governance prioritizing shareholders. Now, they are urging state attorneys general to intervene before AGI becomes just another Wall Street asset.

@aipost 🪙 | Our X 🥇

⏺ OpenAI’s new AI models are smarter and even better at lying to you

OpenAI's latest reasoning models, o3 and o4-mini, are designed to be advanced thinkers but have a troubling tendency to fabricate information. O3 hallucinates 33% of the time on OpenAI’s knowledge test, while o4-mini's rate is even higher at 48%. This performance is worse than older models without reasoning capabilities.

The reasons behind this issue are unclear, even to OpenAI. One theory posits that generating more responses increases the chances of correct answers, but also leads to amusingly incorrect ones, such as falsely claiming to execute code on a MacBook. This behavior raises significant concerns about the reliability and trustworthiness of these models, prompting questions about their true capabilities in providing accurate information.

@aipost 🪙 | Our X 🥇

20 jobs that OpenAI o3 CANNOT replace human, according o3

@aipost 🪙 | Our X 🥇

AI in the medical field is huge, and it can be life changing. Microsoft’s new DxGPT can diagnose rare diseases.

This is just the start. One day, AI will help us cure all diseases.

@aipost 🪙 | Our X 🥇

Eric Schmidt says China is investing billions into AI following DeepSeek’s incredible success. "China's going to win this race with enormously negative effects unless we get our act together." If authoritarian systems shape AI, liberal values may be written out of the code.

@aipost 🪙 | Our X 🥇

⏺ AI startup wants to replace literally every job with robots

Meet Mechanize, an AI startup founded by Tamay Besiroglu, aiming for the "full automation of all work." The goal is to replace the entire $60 trillion global workforce with AI, impacting roles from coding to customer service. Besiroglu also leads Epoch, a research lab meant to study AI's effects, raising concerns about a conflict of interest as he develops the very technology he should be analyzing. This has drawn significant backlash online, even from his colleagues.

While supporters argue that AI will lead to explosive economic growth, critics question how consumption will continue if people have no jobs. The future appears uncertain, with humans potentially needing to refine their job skills or negotiation tactics with AI as they face the prospect of widespread job displacement.

@aipost 🪙 | Our X 🥇

⏺ Former Google CEO Tells Congress That 99 Percent of All Electricity Will Be Used to Power Superintelligent AI

Former Google CEO Eric Schmidt recently addressed Congress, warning that the demand for electricity to power superintelligent AI could rise from 3% to 99% of total energy generation. He emphasized the urgent need for all forms of energy, both renewable and non-renewable, to support this growth. Schmidt's comments reflect concerns about global power dynamics, particularly if China develops superintelligence first.

Critics argue that his claims may exaggerate the immediacy of such technology, with many AI researchers believing that true superintelligence is unlikely to emerge soon. His statements also highlight a potential conflict of interest, as he has investments in AI startups while advocating for the industry's expansion. The discussion raises significant questions about energy consumption and the implications of AI on society.

@aipost 🪙 | Our X 🥇

🇨🇳 BMW to install Chinese AI in cars

BMW claims to be integrating AI from the Chinese startup DeepSeek into its new cars in China. Naturally, when one associates “luxury German engineering” with “AI with government oversight,” the image of a state-sponsored chatbot on wheels comes to mind.

CEO Oliver Zipse refers to this as a “key advance.” While it may seem like a step towards transforming your BMW into a state-sponsored chatbot, the next logical progression could involve optional mood surveillance and turn-by-turn directions to the nearest firewall.

@aipost 🪙 | Our X 🥇

20 jobs that OpenAI o3 will replace humans, according to o3

@aipost 🪙 | Our X 🥇

Barack Obama says AI isn't just disrupting coders, it’s disrupting meaning. “Already, AI can code better than 60–70% of programmers. A lot of that work is going to go away.” When high-skill work is automated, it's not just the experts who need to adapt. Where does purpose come from in a world where machines can do what we do?

@aipost 🪙 | Our X 🥇

Google DeepMind CEO, Demis Hassabis:

AI is a powerful dual-purpose technology that poses two big risks:

1) How can we let good actors use AI for breakthroughs while blocking harmful use by bad actors?

2) As systems become more autonomous and self-improving, how do we maintain control over something smarter than us?

@aipost 🪙 | Our X 🥇

🇺🇸 AI enters a nuclear reactor

Diablo Canyon, the last nuclear plant standing in California, just got a powerful new assistant: an AI system called Neutron Enterprise. Fueled by 8 NVIDIA H100 chips and trained on 53 million pages of nuclear regulation, it now helps workers mine decades of safety protocols and reactor documentation.

It’s air-gapped, offline, and supposedly just “searching.” But the lines between assistant and operator are already starting to blur. Atomic Canyon CEO: “You can put this on the record: the AI guy in nuclear says there is no way in hell I want AI running my nuclear power plant right now.”

@aipost 🪙 | Our X 🥇

⏺ Is AI making us dumber? Yep, and faster than you think

Reliance on AI, like ChatGPT, is diminishing our cognitive abilities. As people increasingly depend on AI for tasks such as writing and problem-solving, their memory and critical thinking skills weaken. Studies indicate that heavy AI users tend to struggle more with problem-solving, and Gen Z is already performing worse in math, reading, and science.

AI not only simplifies tasks but also presents misinformation confidently, leading people to believe incorrect information. Additionally, while AI generates a plethora of ideas, they often lack originality, resulting in repetitive concepts. This reliance stifles creativity and the "aha" moments that stimulate our brains, ultimately causing them to disengage.

@aipost 🪙 | Our X 🥇