Aditya Banerjee @ab_aditya

🚨 Hallucinations have been the Achilles’ heel of LLMs. Google Research just unveiled a fascinating new approach to tackle them from the inside. It’s called Self Logits Evolution Decoding (SLED). Instead of relying only on the model’s final layer, SLED taps into the entire…

🎻 In the Age of AI, is “pretty good” still good enough? A century ago, the phonograph created what Seth Godin calls “The Violinist Problem.” Why settle for a “pretty good” local player when you could hear the world’s best on demand? Suddenly, being average wasn’t a viable…

70,000 applicants. One big question: Can AI run job interviews better than humans? Recruiters predicted no. The data says otherwise. A new field experiment (Voice AI in Firms by Jabarian & Henkel) randomly assigned candidates to human recruiters, AI voice agents, or a choice of…

AI isn’t laying off teams — it’s reshaping how careers even begin. A new Harvard paper, Generative AI as Seniority-Biased Technological Change (Aug 2025), tracks 62 million workers across 285,000 U.S. firms. The findings are sobering: 🚪 The Entry-Level Door is Closing…

“You’re a jerk!” — not me, GPT-4o-mini. All it took was a little Cialdini-style persuasion. A new UPenn study (Call Me A Jerk) shows that large language models can be nudged off their safety guardrails using the same psychological tricks Robert Cialdini outlined in Influence.…

“Machine smart, but humans smarter.” That’s how I ended a blog post I wrote years ago about Arimaa — a board game designed to be computer-resistant after Deep Blue defeated Kasparov. At the time (2007), humans were comfortably winning. By design, brute-force search — the trick…

When 95% of AI Pilots Fail: The Enterprise Learning Gap No One’s Mentioning We love talking about the promise of AI—but MIT’s new study just dropped a dose of reality: 95% of enterprise AI pilot programs fail to deliver measurable business impact.  Why? Not because the tech…

Are you finding yourself increasingly reliant on AI tools for critical decision-making? This article from Rands in Repose offers a fascinating deep dive into the evolving relationship between humans and AI, focusing on three key 'mirrors': Context, Memory, and Voice. The author…

Your AI might be confidently wrong—and fluent nonsense is still nonsense A new study on large language models’ “chain-of-thought” reasoning is a stark reminder: eloquence ≠ understanding. Researchers at Arizona State built a controlled test environment to measure how well LLMs…

When AI upgrades feel like losing a friend 🤖💔 OpenAI launched GPT-5 this week, promising smarter writing, better coding, and a leap forward for ChatGPT. But within hours, users were asking for… the old model back. So many people “mourned” GPT-4o — describing it as a partner,…

We keep asking AI the wrong questions. Most debates are stuck on “Is it intelligent?” or “Does it understand?” But the more useful question—especially if you’re building with LLMs—is: How does it create? Inside today’s models, there’s no hidden “mind”—just a dynamic map of how…

It’s not just the technology—it’s the thinking behind it. The language of AI leadership is shifting. After Microsoft’s directive-heavy memo on “clarity and accountability,” Mark Zuckerberg’s new “Superintelligence Memo” from Meta reads like a manifesto for the next AI epoch.…

AI capital expenditure is eating the world—and it's the real story behind recent strategic moves by Microsoft, Meta, Google, and Amazon. This isn't just a corporate spending spree; it's a profound economic event. According to a recent Wall Street Journal report, the combined…

Satya Nadella’s recent layoff memo at Microsoft isn’t just corporate communication—it’s a reflection of deeper shifts around AI, strategy, and organizational clarity. Om Malik captures it brilliantly, decoding corporate-speak into clear insights: •AI isn’t just tech—it’s…

Just wrapped up a month of deep vibe coding with Cursor, building Agentic AI prototypes for enterprise workflows. Productivity boost? Easily 10x. But vibe coding has a twist: it feels like managing a software team with high attrition and terrible knowledge transfer—imagine…

🔐 The $5 Wrench in Your AI Strategy In cybersecurity, we often fixate on sophisticated, AI-powered attacks versus complex encryption. The reality? The biggest threats often bypass technology entirely. As this classic XKCD comic hilariously illustrates, why build a…

Aditya Banerjee writes for #YourTurn—our exclusive subscriber offering! You can also stand a chance to publish your opinion and get pictured on ThePrint. Subscribe Here: theprint.in/subscribe/

♟️ From Chess Masters to Self-Driving Cars—AI's Big-Picture Journey In my latest article for The Print, I unpack how artificial intelligence evolved from narrowly beating chess grandmasters to confidently navigating crowded city streets. This journey isn't just about technical…

✈️ AI is quietly rewriting the rules of pricing. Delta is piloting a new AI-powered pricing engine that personalizes fares using data from your browsing, past bookings, urgency cues — even your financial profile. Same flight. Same seat. Different price — based on what the AI…

"The machines may be abundant, but the value lies in scarcity." Ben Thompson’s latest piece, "Tech Philosophy and AI Opportunity," is a must-read on the real drivers of the AI race – and why it’s not just about compute or models. Some key takeaways that resonated with me: 🧠…