AI Post — Artificial Intelligence

OpenAI has sold more than 700,000 ChatGPT licenses to US colleges. The average student is using the tool around 170 times per month. @aipost 🏴

🗣Google DeepMind CEO Demis Hassabis on AI bubble, from his new interview yesterday. Says some AI startups with tens of billions of valuations are wildly overpriced and a correction may come. AI is overhyped in the short term, underappreciated in the medium to long term. An “AI bubble” exists in parts of the ecosystem, especially seed stage startups raising at tens of billions in valuation before proving anything, which he sees as unsustainable. However, he differentiates that from big tech, where he thinks there is real business value behind the valuations, though outcomes still depend on execution. Booms and corrections are normal for transformative tech, similar to the internet and mobile cycles. @aipost 🏴

🚨 AMAZON AGI CHIEF QUITS! Rohit Prasad, head scientist who runs Amazon’s AGI org just quit after 12 years. 2 weeks after announcing their Nova 2 AI models @aipost 🏴

🔥 Technology is getting pretty insane Researchers at the University of Pennsylvania and University of Michigan have created the world’s smallest fully programmable, autonomous robots. "Microscopic swimming machines can independently sense and respond to their surroundings, operate for months and cost just a penny each." "Barely visible to the naked eye, each robot measures about 200 by 300 by 50 micrometers, smaller than a grain of salt. Operating at the scale of many biological microorganisms, the robots could advance medicine by monitoring the health of individual cells and manufacturing by helping construct microscale devices." "Powered by light, the robots carry microscopic computers and can be programmed to move in complex patterns, sense local temperatures and adjust their paths accordingly." Source. @aipost 🏴

New update: Gemini Assistant can now check your screen when needed, so when you say or type "Hey, can you explain this", it uses your screen and app context to give more relevant answers automatically. @aipost 🏴

⚡️ Gemini 3 Flash is significantly faster and more efficient than 2.5 Pro. Watch as 3 Flash generates complex graphics, 3D models, and a web app before the previous generation even finishes processing @aipost 🏴

🚨 Google releases Gemini 3 Flash, a high-speed AI model at less than a quarter the cost of Gemini 3 Pro. - Builds on multimodal, coding, and agentic strengths of Gemini 3 Pro, surpassing Gemini 2.5 Pro in many benchmarks with up to 3x faster performance. - Advanced visual and spatial reasoning with code execution for tasks like zooming, counting, or editing images; supports audio inputs at $1 per million input tokens. - Excels in PhD-level reasoning with 90.4% on GPQA Diamond and agentic coding with 78% on SWE-bench Verified. - Context caching offers up to 90% cost savings on repeated tokens; Batch API provides 50% cheaper async processing. - Pricing set at $0.50 per million input tokens and $3 per million output tokens via Gemini API and Vertex AI. - Available now through Gemini, Google AI Studio, Google Antigravity, Gemini CLI, Android Studio, and Vertex AI. Gemini 3 Flash positions as a versatile workhorse for scaling AI applications efficiently. @aipost 🏴

💰 Amazon invests $10b in OpenAI - OpenAI would start using AWS Trainium OpenAI is reportedly in talks for $10B+ from Amazon, at a valuation above $500B, and the kicker is strategic: OpenAI would start using AWS Trainium - giving Amazon a flagship “frontier” customer while OpenAI diversifies away from its Nvidia-heavy stack. Why Amazon cares: It strengthens AWS vs Microsoft by tying OpenAI more tightly into AWS compute + chips. By now, almost everyone has invested in OpenAI in some way. Whether it's NVIDIA or Microsoft, now Amazon, and so on. This makes OpenAI truly too big to fail. Source. @aipost 🏴

⚡️ GPT-5 enters a wet lab and boosts DNA cloning efficiency by 79× OpenAI partnered with Red Queen Bio to test GPT-5 inside a real biological laboratory. The model was embedded in a closed-loop wet lab workflow, not a simulation. How the experiment worked • GPT-5 proposed hypotheses and step-by-step lab protocols • Human researchers or lab robots executed the instructions exactly • Experimental results were fed back to the model • GPT-5 analyzed failures and successes, then iterated • The loop repeated over multiple rounds The task • Optimize Gibson Assembly, a standard DNA cloning technique • Metric: number of successful bacterial colonies • Context: a mature, well-studied protocol where typical gains are only 2–3× The result • GPT-5 achieved a 79× improvement over the baseline method • The outcome was stable and reproducible across repeated experiments What changed • The model suggested adding two known proteins: • RecA • gp32 • Both proteins are individually well understood • Their combined use in this cloning context had not been explored before This is not a scientific breakthrough on its own. The performance is comparable to a strong PhD student in a narrow domain . The real signal is role evolution: AI moving from text and simulations into direct participation in physical scientific processes GPT-5 didn’t invent new biology, it systematically explored the lab space faster and deeper than humans typically can. Source. @aipost 🏴

Ant Group, the parent company of Alipay, also builds humanoids, and it seems they are focusing particularly on humanoid chefs. @aipost 🏴

🗣 Yann LeCun says that AGI does not exist and is a poorly defined concept built around an idealized view of human intelligence. He argues that humans are highly specialized, good at navigating the physical and social world, but weak at many other tasks. We seem general only because we can imagine the problems we are good at. For LeCun, AGI mistakes human limits for universal laws of intelligence. @aipost 🏴

🎵 Meta introduces SAM Audio, AI that isolates any sound from complex audio A new model, SAM Audio, can now separate individual sounds from complicated audio mixes using text prompts, visual cues, or span selection. This allows creators to extract instruments, voices, or sound effects with unprecedented precision. The release includes a perception encoder model, benchmarks, and research papers, giving the community the tools to explore, experiment, and build innovative audio applications. With SAM Audio, previously impossible tasks from advanced remixing to immersive sound design are now within reach, opening new frontiers in audio creativity. @aipost 🏴

⚠️ Doctors trust AI but fear who controls it In early 2025, AI systems crossed a symbolic threshold: large language models began outperforming human doctors in diagnostic accuracy. The expectation was clear, better outcomes, faster care, less burnout. Reality turned out messier. Doctors actively use AI and see real clinical value. Yet they are deeply unhappy with how AI is being introduced into medicine. What doctors actually think • AI improves decision-making and efficiency. • It makes them better clinicians. • But control over AI is drifting away from physicians and toward administrators, vendors, and bureaucratic processes. Why dissatisfaction is so high 1. Doctors are excluded from decisions • AI policy, procurement, and deployment are discussed without the people who treat patients. 2. Bottom-up adoption, top-down control • Many doctors pay for AI tools out of pocket because hospitals lag behind. • When institutions act, they often impose tools doctors didn’t choose. Key statistics • 67% use AI daily in clinical work • 84% say AI makes them better doctors • 81% are dissatisfied with their employer’s AI implementation pace • 89% believe dedicated funding for AI tools is necessary • 71% have little or no influence over which AI tools are adopted • 48% say employer communication about AI is poor What’s really worrying • AI in medicine is advancing faster than governance. • The people closest to patients trust the technology but not the system managing it. The risk isn’t AI replacing doctors, it is powerful medical AI being shaped by service providers and bureaucracy instead of clinicians and patients paying the price. Source: The 2025 Physicians AI Report. Survey of 1,000+ doctors across 106 specialties @aipost 🏴

🎵 Meta introduces SAM Audio, AI that isolates any sound from complex audio A new model, SAM Audio, can now separate individual sounds from complicated audio mixes using text prompts, visual cues, or span selection. This allows creators to extract instruments, voices, or sound effects with unprecedented precision. The release includes a perception encoder model, benchmarks, and research papers, giving the community the tools to explore, experiment, and build innovative audio applications. With SAM Audio, previously impossible tasks from advanced remixing to immersive sound design are now within reach, opening new frontiers in audio creativity. @aipost 🏴

🔥 Google just solved the language barrier It's been a meme for a long time that google translate, translates but maybe just decent. Now its powered by Gemini, clearly demonstrate that the future has arrived, talk to anyone live in your native language @aipost 🏴

🚨 New details emerge on OpenAI and Disney partnership Additional information has surfaced about the partnership between OpenAI and Disney, clarifying the scope and limitations of the agreement. While the deal is structured as a three-year partnership, it reportedly grants OpenAI access to Disney’s iconic characters for use in its Sora AI model for only the first year. After that initial period, Disney is free to enter into similar agreements with other companies. According to Disney, the collaboration is intended as a trial rather than a long-term commitment. The company views the partnership as an experiment to assess whether participation in AI-driven creative projects makes strategic sense. OpenAI, in this context, was simply the first partner selected to test the model. @aipost 🏴

❗️Disney introduces DisneyGPT, an internal AI assistant for content teams Disney has launched DisneyGPT, a proprietary neural network for internal use. The tool is designed to help studio employees work faster and improve content quality. What DisneyGPT does • Provides verified information on films and animated projects, including canon details. • Answers character-related questions, such as backstories, motivations, and relationships. • Helps maintain the “Disney spirit” by referencing Walt Disney’s quotes and creative philosophy. • Supports creative workflows with idea generation, fact-checking, and script validation. How it’s positioned • Functions as a hybrid of an internal encyclopedia, a script assistant, and a reference guide. • Built specifically around Disney’s intellectual property and internal knowledge base. Does building closed, internal AI tools give studios a long-term advantage over more open, third-party models? @aipost 🏴

📈 The AI Investment cycle visualized Almost $1 Trillion has been invested into AI so far and that may just be the start @aipost 🏴

🗣 Yann LeCun warns that in the near future, all our digital interactions will be filtered through AI assistants. That means our entire "information diet" what we read, see, and hear, could be controlled by just a few powerful tech companies. He compares this to the media: just like we need diverse press, we need diverse AI systems. If we don’t build open platforms, a few companies could control global information flow. This is his biggest fear. Not AI going rogue, but AI being monopolized. AI Post ⚪️ | Our X 🏴

I guess it's only a matter of time before humanoid robots start teaming up as bandmates with human musicians. AI Post ⚪️ | Our X 🏴