AI startup co-founded by Google veterans sets sights on humanity’s toughest challenges

• 'Quái kiệt AI' người Việt rời Google, mở startup
• 'Quái kiệt' Lê Viết Quốc: Làm AI cần sáng tạo, đừng vội vã chạy theo

Earlier this week, the formidable quartet of Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Le Viet Quoc launched Discovery Loop.

Writing on X, Dean outlined the firm's overarching purpose: "Discovery Loop will aim for public benefit goals, with a mission to automate machine learning, science, and engineering to accelerate discoveries and progress."

The co-founders, from left: Oriol Vinyals, Sanjay Ghemawat, Jeff Dean, and Le Viet Quoc. Photo courtesy of Discovery Loop

Further details posted on the founders' official website and social media channels shed light on the startup's ambitious core mechanism.

Dean explained that the "loop" in the company's name reflects its core imperative: automating the entire scientific workflow, from proposing and executing experiments to analyzing the results.

This design aims to "solve some of the hardest problems in science and engineering by running thousands of experiments in parallel," he said.

On X, the startup added that its long-term goal is to engineer systems capable of tackling key elements of the 14 grand challenges laid out in 2008 by a panel of world-leading scientists and engineers to advance human welfare.

Known as the National Academy of Engineering Grand Challenges, the agenda identifies 14 essential engineering benchmarks for the 21st century, spanning directives from "reverse-engineering the brain" to "preventing nuclear terror".

Designed to mobilize researchers around existential priorities like clean energy, environmental sustainability, healthcare, global security, and quality of life, the project was proposed among others by Google co-founder Larry Page.

Fresh out of stealth mode, Discovery Loop has already received massive attention across Silicon Valley.

Top-tier venture capital firms Radical Ventures and Khosla Ventures co-led the initial funding round, alongside prominent backers like Lightspeed, Kleiner Perkins, and Doerr Capital.

TechCrunch reported that the enterprise also secured financial backing directly from Alphabet, Google's parent company.

A pedigree of AI pioneers

As Business Insider points out, the venture has ignited widespread interest largely due to its executive lineup, which comprises key architects of Google's technological foundation.

Headlining the roster is Dean, Google's 30th employee, whose legendary career stretches from establishing Google's dominance in web search to steering its foundational generative AI efforts.

Over a 27-year tenure at Google, he played a pivotal role in engineering search, news, translation, and advertising architecture.

He co-founded Google Brain, a unit that AI pioneer Geoffrey Hinton praised for transforming fundamental neural network research into world-changing applications. Dean was also the mastermind behind Google's Tensor Processing Unit chips, and his former deputies include Anthropic CEO Dario Amodei and OpenAI co-founder Ilya Sutskever.

Ghemawat has been one of Dean's most enduring research partners, with a partnership pre-dating their Google tenure to their time at Digital Equipment Corporation.

Inside Google, their bond grew so close that Dean's children refer to Ghemawat as "Uncle Sanjay," according to a profile by The New Yorker.

Ghemawat and Dean were the first engineers at Google to reach Level 11, holding the title of Senior Fellow, and both rank among the most cited figures in distributed systems research.

Wired once hailed Ghemawat as "one of the most important software engineers of the Internet age."

Vinyals previously served as a principal scientist at Google DeepMind. Specializing in machine learning and reinforcement learning, he led key breakthroughs underlying Google Translate, and his landmark AlphaStar project made headlines by defeating top professional players in the real-time strategy game StarCraft II.

Dr. Le Viet Quoc was likewise a cornerstone of Google Brain since its inception. Born in 1982 in Thua Thien Hue Province in central Vietnam, Quoc was involved in one of Google Brain’s early projects, co-developing a neural network that ran on 1,000 computers and 16,000 processors. This experiment proved that AI could learn independently after being trained on 10 million unlabeled YouTube video frames, according to The New York Times.

MIT Technology Review named Quoc in its prestigious list of Innovators Under 35 in recognition of his contributions to global artificial intelligence research.

Over the years, he has regularly returned to Vietnam to mentor emerging researchers and speak on advancements in AI, science, and technology.

Analysis by research firm Futuresearch suggests that, given the founders' track record and stated objectives, Discovery Loop will likely concentrate heavily on AI chip design, assigning a 38% probability to a major milestone by 2028.

Secure cyberspace infrastructure and AI-driven scientific discovery tools follow, each assessed at a 28% likelihood.

Industry outlet Emergent notes that the decision by top Google veterans to branch out independently signals massive potential in AI-automated research.

Discovery Loop's public-benefit corporate structure sets it apart from traditional AI ventures: While typical AI firms operate under standard venture capital models aiming for commercial buyouts, a public-benefit entity is better structured to partner with academic institutions and government labs that champion open science.

"Operating as a public-benefit corporation demonstrates a commitment to democratizing scientific discovery rather than chasing purely commercial returns," Emergent observed. "If successful, Discovery Loop could pioneer a paradigm where AI systems actively participate in research, compressing years of work into months and unlocking breakthroughs unattainable through manual experimentation alone."