Sonia Joseph’s Plan to Find the Answer to Everything

How one SPC member hopes to use AI to answer the most fundamental questions of existence.

Sonia Joseph’s Plan to Find the Answer to Everything

What’s the reason for our reality?

It’s an almost ludicrously ambitious question to ask seriously. But if you spend time chatting with SPC member Sonia Joseph, you quickly realize how serious she is about finding an answer. Her plan to resolve the question is no less daunting: by building artificial general intelligence (AGI), an AI system with all the flexibility and resourcefulness of human intelligence (and perhaps far more). Sonia hopes to learn why there is something instead of nothing through the creation of a new form of advanced intelligence.

These radical ambitions are grounded in Sonia’s upbringing. Throughout her 26 years of life, Sonia, who joined SPC in November 2021, has found herself at home in the spaces between contrasting worlds. She grew up between Boston and India, the daughter of an Indian Catholic family. Her father was an academic, her mother an engineer, while a grandmother she was close with had a PhD in British Indian feminist literature. These authorities filled her life equally with AI treatises by Norbert Weiner and legends of Rani of Jhansi, India’s Joan of Arc. Caught between the gravities of very different worlds, it makes a bit more sense why Sonia imagined philosopher machines as the solution to a problem that had bothered her since childhood.

“I can remember being confounded by reality when I was 3 years old. I would try to explain to my parents that I didn't understand why we exist, but coming from a toddler, my parents figured I’d grow out of it,” Sonia said.

She did not. She delved into reading, often spending 16 hours a day with her head buried in a book, or, later, in online forums dedicated to the same questions.

“I think everyone experiences this, but in my case, it was just extremely strong and it never left,” Sonia said. “And I'm still pretty metaphysically confused and anxious today to the point where it impacted my career.”

Sonia followed her confusion to Princeton University, where she studied machine learning and creative writing. Before graduating she wrote a thesis on semantic representation in the human brain in collaboration with the Princeton Neuroscience Institute and Google, while also authoring a pseudonymous memoir under the guidance of novelist and screenwriter Aleksandar Hemon.

“I've always been most comfortable between fields doing a kind of arbitrage because in some ways it feels more true than staying within a single field,” Sonia said before adding, “but it might just be my background and what I'm used to and comfortable with.”

Following a short stint at a startup following graduation, Sonia enrolled in a machine learning PhD program at Mila Quebec AI Institute. But her interest in company building was growing. Increasingly, a startup seemed like the best path to realize her goals, but the mechanics of success as a founder were unfamiliar. So she teamed up with her good friend Amelie Lasker to explore and eventually launch Alexandria, an internet publishing house that builds infrastructure for web3 books. She also joined SPC.

“SPC doesn't have the baggage of academia, which is often less accepting of new paradigms,” Sonia said. “It feels like a biocomputer that you can use to activate whatever company or project you want to create. There are fewer constraints, so you can pursue your interests in unconventional ways and have a better chance of creating something genuinely novel.”

Even as she learned about building startups with Alexandria and fellow SPC members, the nagging question remained. But the world seemed to be catching up. The AI research ecosystem—and, even more noticeably, the rapidly growing number of companies and products spinning out from it—had embraced language as the core training data for groundbreaking AI models like GPT-3. These models were exhibiting novel and unexpected capabilities simply from getting bigger, a phenomenon of the scaling laws that are a core part of Sonia’s research at Mila. By mid 2022, Silicon Valley had collectively experienced a vibe shift Sonia had long expected: AI, a field that had for so long over-promised and under-delivered, was finally moving.

Fortunately for all of us at SPC, members like Sonia were ready to help us keep up. She co-organized a recently-launched member learning forum on AGI. Over seven weeks, the forum explores the core challenges and opportunities of a technology that suddenly feels much more science than fiction. It’s yet another ambitious undertaking for Sonia and her fellow organizers—to the best of their knowledge, no other such curriculum on AGI exists.

“The term AGI is still controversial in the research world, so I’ve been looking at the technical, political, social, philosophical, and economic connotations behind how we are using this word – instead of buying into the current simple narratives that are floating about San Francisco,” Sonia explained. “I’m looking into company structures for a privatized research moonshot in the AI space that also has the upside of a for-profit company. Company structures for AI moonshots are important because you quickly connect the question of AI alignment to the incentive structure of capitalism.”

Sonia attributes much of her motivation to build an AGI supercomputer to books like “The Hitchhiker's Guide to the Galaxy” and Ken Liu’s short story about “The Bookmaking Habits of Select Species” – literature that imprinted on her during her early search for answers. At the same time, Sonia acknowledges mixed feelings about today’s notion of god-like supercomputers.

“The DeepMind founder is doing this—he's building the “Hitchhiker's Guide to the Galaxy” supercomputer that's going to solve all of science,” Sonia said. “In some ways I'm also aware this thinking is very arrogant. I don't think one epistemology ever leads to truth. You get to truth by using many, many epistemologies to approximate.”

Sonia clearly won’t be tackling the question of reality during the AGI forum’s seven weeks. Nor will the literature enabled by Alexandria settle questions that have flummoxed writers for millennia. But in the years to come, you are sure to find Sonia comfortably stationed between spaces, both physically and metaphysically, in search of that fundamental answer that so many of us seek.

If you're looking for a community to help you explore the most ambitious questions imaginable, consider applying to SPC!