Lightnews — Scholar-powered news

Aviral Prakash @heyaviral.bsky.social · Jun 9

Attempting to fix a broken model with AI will likely only magnify its flaws.

Aviral Prakash @heyaviral.bsky.social · Jun 9

Per Davenport & Mittal, "AI-fuelled" success is predicated on pre-existing robust fundamentals (model, customer relations, product).

AI's role is acceleration, not creation ex nihilo. Thus, any AI strategy must be preceded by a rigorous assessment of core business health.

Aviral Prakash @heyaviral.bsky.social · Jun 9

Just finished a new book on Sunchaser - "All-in on AI"

Learnt a core principle for AI adoption: AI is not a panacea for business deficiencies; it is an amplifier of existing competence—or incompetence.

Lessons From – AI Snake Oil

Aviral Prakash @heyaviral.bsky.social · Jun 1

What's the most over-the-top AI claim you've seen lately that made your "snake oil" senses tingle? Let's call out the hype! 👇

Aviral Prakash @heyaviral.bsky.social · Jun 1

Lessons From - AI Snake Oil

sunchaser.life/2025/05/29/l...

AI Snake Oil is a bold exposé on the misleading promises of artificial intelligence. Authors Narayanan and Kapoor dissect the rampant hype, revealing how many AI technologies are inaccurately marke…

Aviral Prakash @heyaviral.bsky.social · Jun 1

Often, it's not an oopsie; companies might *knowingly* overhype their AI. So, always dig deeper! Ask: "Cool, but what *kind* of AI is it?" (Generative? Predictive?). Specifics beat vague promises every time. Stay sharp! #Sunchaser #AISkeptic #ResponsibleAI

Aviral Prakash @heyaviral.bsky.social · Jun 1

Seeing a lot of "AI-powered" magic out there? Let's talk "AI Snake Oil" 🐍 – a term for AI that "does not and cannot work as advertised" (thanks, Narayanan & Kapoor!). This means it might just be bad, or the problem is too big for *any* current AI.

Aviral Prakash @heyaviral.bsky.social · May 28

How can everyday users push for less biased AI? What questions should we ask before trusting an AI tool? Let's discuss practical steps! 👇

Aviral Prakash @heyaviral.bsky.social · May 28

Bottom line: AI can easily become a super-spreader for existing societal biases. We HAVE to demand & build AI that's fair & represents everyone. What's your biggest concern? #Sunchaser #AIEthics #ResponsibleAI

Read more: buff.ly/nO542WZ

Ethan Mollick’s “Co-Intelligence” dives headfirst into the disorienting world of Generative AI, arguing it’s reshaping workplace dynamics and creativity. As AI blurs the lines bet…

Aviral Prakash @heyaviral.bsky.social · May 28

Even the "human feedback" (RLHF) part meant to fix it can add more bias, because the human raters aren't immune! Plus, there are ethical questions about that work itself. 🤔

Aviral Prakash @heyaviral.bsky.social · May 28

Let's talk AI Bias – it's a HUGE deal. LLMs learn from us & the internet... which means they learn our prejudices, stereotypes & the views of a limited slice of humanity. 😬

Aviral Prakash @heyaviral.bsky.social · May 27

This 'emergence' makes AI both exciting and a bit unpredictable. What's one question you have about where this is heading? Let's brainstorm! 👇

Aviral Prakash @heyaviral.bsky.social · May 27

So they act more like 'connection machines' (kinda like our brains!) than typical software. Super powerful, a bit wild, and totally fascinating. What AI 'emergence' has wowed you?

Read more here: buff.ly/nO542WZ

Ethan Mollick’s “Co-Intelligence” dives headfirst into the disorienting world of Generative AI, arguing it’s reshaping workplace dynamics and creativity. As AI blurs the lines bet…

Aviral Prakash @heyaviral.bsky.social · May 27

Think of them as autocomplete geniuses. BUT then, poof – "Emergence"! They start doing things they were never, ever programmed for – like showing 'empathy' or mastering games. 🤯

Aviral Prakash @heyaviral.bsky.social · May 27

Curious how AI like ChatGPT gets so smart? It's not magic, it's "unsupervised learning"! They basically read everything, finding patterns and word connections themselves. No cheat sheets! 🤓

Aviral Prakash @heyaviral.bsky.social · May 26

Beyond text generation, where do you see the core principles of the Transformer architecture having the most profound future impact? Discuss the possibilities. 👇

Aviral Prakash @heyaviral.bsky.social · May 26

This allowed models (now LLMs) to grasp long-range dependencies and context with unprecedented accuracy, enabling tools like ChatGPT. It marked AI's shift from niche applications to mainstream consciousness, kickstarting the current Generative AI discourse. A true paradigm shift.

buff.ly/nO542WZ

Ethan Mollick’s “Co-Intelligence” dives headfirst into the disorienting world of Generative AI, arguing it’s reshaping workplace dynamics and creativity. As AI blurs the lines bet…

Aviral Prakash @heyaviral.bsky.social · May 26

2017 was arguably the most critical year in modern AI's trajectory. Google's 'Attention Is All You Need' paper introduced the "Transformer" architecture. Its core innovation, the 'attention mechanism,' fundamentally broke from prior sequential processing limitations (RNNs/LSTMs).

Aviral Prakash @heyaviral.bsky.social · May 25

As AI accelerates and its capabilities become less predictable, how must our frameworks for education, policy, and ethical oversight adapt in real-time? Share your insights. 👇

sunchaser.life/2025/05/22/l...

Ethan Mollick’s “Co-Intelligence” dives headfirst into the disorienting world of Generative AI, arguing it’s reshaping workplace dynamics and creativity. As AI blurs the lines bet…

Aviral Prakash @heyaviral.bsky.social · May 25

This rapid, often surprising trajectory signifies a transformative era. It necessitates a dual strategy: courageous exploration of potential alongside rigorous caution and societal dialogue regarding its integration and implications. #AI #Acceleration #CognitiveAI #AI_Ethics #TuringTest #Sunchaser

Aviral Prakash @heyaviral.bsky.social · May 25

The AI progress curve is not just rising; it's steepening. We're observing a shift from AI as tools to AI exhibiting human-like cognitive capabilities, evidenced by success in complex professional exams (law, medicine) and progress towards benchmarks like the Turing and Lovelace tests.

Lessons From – The Pathless Path

Aviral Prakash @heyaviral.bsky.social · May 24

How much of the 'Organization Man' ethos do you believe still lingers in today's corporate environments, despite the growing appeal of the 'Pathless Path'? Share your observations and insights. 👇

Aviral Prakash @heyaviral.bsky.social · May 24

The Pathless Path champions a more autonomous, purpose-driven approach to work and life. Food for thought on how far we've come, and where we're headed. #WorkHistory #OrganizationMan #PathlessPath #Sunchaser

Learn more: sunchaser.life/2025/05/15/l...

Paul Millerd’s “The Pathless Path” invites readers to explore a life centered on authentic creation rather than societal expectations. It challenges the conventional “Default Path…

Aviral Prakash @heyaviral.bsky.social · May 24

The contemporary "Pathless Path" movement can be seen as a significant counter-narrative to this historically entrenched norm of conformity and the all-encompassing corporate identity.