If You Don’t Have a PhD, You’re Not an AI Startup

Meta AI chief Yann LeCun thinks it might be. In a recent chat with Zerodha co-founder Nikhil Kamath, LeCun emphasised how a PhD equips entrepreneurs with the skills to innovate while avoiding false assumptions about their ideas. 

“In a complex field like AI, knowing what’s possible and what’s not is invaluable,” he said.

India’s AI ecosystem, however, is still catching up. Only 8% of Indian AI startups are founded by PhDs, a stark contrast to global trends. Industry veterans like Amit Sheth of the AI Institute at the University of South Carolina argued that deep technical expertise gives startups a definitive edge. His students’ ventures, like AppZen, thunk.ai, Objective, Inc., and Clinical AI Assistance, stand as a testament to this claim. 

“While a PhD isn’t mandatory, it sharpens your ability to execute complex innovations,” Sheth affirmed.

But the gap isn’t just academic. Investors’ preference for quick returns often sidelines deep-tech ventures, a challenge voiced by founders like Vishnu Vardhan, who lamented the lack of genuine support for foundational AI innovation. Startups like Subtl.ai and Sarvam AI, powered by founders with academic rigour, highlight the potential for transformation when research meets entrepreneurship. 

Globally, pioneers such as Ilya Sutskever and François Chollet have leveraged academic insights to disrupt AI’s trajectory, inspiring Indian founders to aim higher.

So, what’s holding us back? Is it time for Indian entrepreneurs and investors to shift gears and embrace the long game? Dive deeper into this fascinating story and its implications for India’s AI landscape here.

AI is Discovering Millions of Materials—But Can it Deliver?

AI is churning out discoveries at a mind-blowing pace: 200 million protein structures from Google DeepMind’s AlphaFold, 772 million from Meta’s ESMFold, and Orbital Materials’ claims of 10x performance gains in carbon removal. The promise? Revolutionary breakthroughs in medicine, clean tech, and more. The reality? We’re far from seeing these innovations in action.

AI can uncover the future, but until it can deliver at scale, the hype remains just that—hype. The bigger question is—can AI move from discovery to real-world impact? Read to find out. 

AIM Predictions 2025

Like every year, AIM is at it again with bold predictions for 2025, envisioning breakthroughs in agentic AI transforming enterprises, humanoid robots entering homes, generative 3D modeling surging, and edge computing empowering devices. All this while China's AI dominance continues to grow. Here's hoping some of these come true! More details here. 

In other news, 

• Google DeepMind introduced ‘Socratic learning’, a framework where AI self-improves in closed systems by engaging in language games inspired by Wittgenstein's philosophy and Socratic questioning. 

• Google introduced PaliGemma 2, a family of vision language models optimised for task transfer across fields like OCR, molecular structure recognition, and music transcription. 

• OpenAI launched a $200 ChatGPT Pro plan featuring the advanced o1 model with multimodal input, improved reasoning, and developer-centric tools, alongside a 12-day release spree teasing innovations like GPT 4.5 and an AI agent called Operator. 

• Microsoft introduced Copilot Vision in Edge that can analyses web pages in real time to assist with tasks like shopping and learning while prioritising user privacy.