@RailwaySeva

6-0 just got real 😉🥰

Top AI Papers of The Week (July 28 - August 3): - GEPA - Graph-R1 - AlphaEarth - Self-Evolving Agents - Hierarchical Reasoning Model - Efficient Attention Mechanisms - Geometric-Mean Policy Optimization Read on for more:

🚨NVIDIA CEO just redpilled the MIT study claiming AI makes people dumber > “I haven’t read the MIT research lol” > “I literally use AI every day and my cognitive skills are actually advancing” > most of his job as CEO = asking good questions > “I’m not asking a machine to teach

Upgraded from Llama 3 to Qwen3 as my go-to model for research experiments, so I implemented qwen3 from scratch: https://t.co/PZDxKyow2v Trade-off: Qwen3 0.6B is deeper (28x vs 16x layers) & slower than the wider Llama 3 1B but more memory efficient due to fewer params

Link to Paper:- https://t.co/ljvfbz9f39

Think of it like a mental muscle. If you stop exercising it, it gets weaker. Is our reliance on AI leading to a future of "cognitive atrophy"? Let's discuss!

The LLM group, on the other hand, showed the least brain activity. This suggests they were "offloading" their thinking to the AI. While this might seem efficient, the study warns of a "cognitive debt" – a potential decline in our thinking abilities over time.(4/n)

The study compared three groups: one using an LLM (like ChatGPT), one using a search engine, and one using only their brains. The results? The "Brain-only" group showed the highest levels of brain connectivity, meaning they were thinking the hardest. 🧠⚡️ (3/n)

Is your AI assistant a helpful tool or a cognitive crutch? A fascinating new paper from MIT Media Lab suggests that while AI can make tasks like essay writing easier, it might come at a long-term cost to your brain. (2/n) #AI #CognitiveScience #ChatGPT

Is your AI assistant making you smarter, or just taking on your "cognitive debt"? A recent MIT study has some shocking answers. 🤯 (1/n)

🎙 𝗧𝗼𝗱𝗮𝘆’𝘀 𝗦𝗽𝗲𝗰𝗶𝗮𝗹 𝗔𝗜/𝗠𝗟 𝗢𝗽𝗲𝗻 𝗗𝗶𝘀𝗰𝘂𝘀𝘀𝗶𝗼𝗻! Join us this evening as we host 𝗛𝗮𝗿𝘀𝗵 𝗝𝗼𝘀𝗵𝗶, AI & Product Leader at 𝗔𝗜𝗢𝗡, for a thought-provoking session on 𝗛𝗘𝗡𝗔. @AILucknow @cataluna84 @TFUGlucknow

Link to Paper: https://t.co/a4sElqm60S

🔍 Apple’s point? We’re not close to AGI. These models match patterns, not reason. The industry is celebrating fake intelligence.

📊 Apple found: * ✅ Low complexity: Regular models win * ⚖ Medium: "Reasoning" models help * 🚫 High: Everything breaks 🏗 Models did well on memorized puzzles (Tower of Hanoi) ❌ Failed at unseen logic problems (River Crossing)

💥 Result: * Models collapsed at higher complexity (0% accuracy) * Gave up faster, used fewer tokens — even with unlimited compute * Failed even with exact solution steps 🧠❌

🚨 Apple just called out the AI reasoning hype🚨 📌 Tested top models like Claude, DeepSeek-R1, and OpenAI’s o3-mini 🧩 Used *fresh puzzle games* (not memorized math problems)

AAKASH THE G.O.A.T.

Notice the uniforms.

शं नो वरुणः