Why AI Investments Fall Short (and an AI immigrant who came to the US with $1,000)
Novelty Search with quantum physicist Dr. Radhika Dirks
“It is human nature to be searching for novelty. AI’s Novelty Search mirrors evolution and innovation. Humans always figure a way out of where we are: out of dystopia and out of utopia! We were bound to find our way into space. We are explorers at heart!” – Dr. Radhika Dirks
At XLABS, we believe in leveraging the huge progress in AI, quantum computing, and neurotech to build unambiguously good moonshots which amplify humans.
AI investments have been focused on incremental gains. Why? Because the algorithms themselves are not geared towards breakthroughs.
Dr. Kenneth O. Stanley (former head of Uber’s AI, current head of Peter Theil, and Elon Musk’s OpenAI) questioned why all AI algorithms minimize errors towards an objective.
Does nature have an objective in evolution? Do creators have an objective in breakthroughs. Or are they curious and in discovery mode, open to what might present.
How can we be open and curious when we are minimizing errors towards an objective?
If all of AI is built on objective-focused, error minimization, it’s no wonder AI investments have done OK, but not produced any big home runs (unicorns).
Worse, this AI is destined to fail at some of the more challenging feats. Walking requires falling to learn oscillation, yet AI considers falling an error, so it takes longer for robots to learn to walk.
What’s the solution?
Perhaps novelty search: tell the computer to try something new, something it hasn’t tried before. Search for novelty.
Novelty search is particularly successful at mimicking evolution (thus mapping cancer, viruses and RNA) and innovation - areas primed for big breakthroughs.
This is where you can then find breakthrough companies coming from the AI space.
Let’s dive into an innovator who took a novel path, Dr. Radhika Dirks, and learn her story.
Q: You came to the US from India with $1,000 to attend college and are now a global expert in AI. Why do you think you became who you are?
A: I was born in India. As I approached college, I realized that engineering schools in India teach what is already known (much like objective-focused AI). I’ve always been more interested in uncovering new knowledge versus excelling at what the world already knows. Growing up (and even now) my patron saints were Schrodinger, Vinci, Turing, Godel, and Emily Dickinson, truth-seekers who “broke” fields to create new ones or waltzed effortlessly between subjects.
In a country like India where parents and the whole society can be extremely controlling of a person’s life, my parents encouraged my curiosity, my ability to excel across fields, and most importantly, let me fully be me.
This freedom was the biggest gift I could have asked for.
I had the freedom to fail and the support to walk an atypical path, a path my parents still don’t fully comprehend but cheer me on wholeheartedly. Growing up as a mega curious person who is gritty enough to dive into insane depths of details and willing to actually do stuff, in a country where nothing makes sense on a given day, from Government policy to people blowing up for no reason, will shape you into an unreasonable kind of person, the kind who wonders what makes people, societies and the universe this way.
I came to the US to find the answers and reshape it all.
Q: In your career, you have developed powerful algorithms, which are a mix of novelty search and other techniques. What has your work in AI taught you or “decoded” about the world?
A: Our AI was called Seldn, named after the famous fictional mathematician from the Asimov novels.
Seldn taught us that building an AI that mimics societies (instead of mimicking the human brain) is actually possible by combining complexity physics, a field related to chaos theory, with machine learning.
Seldn could predict societal ‘black swans’, ie, highly unpredictable events which seem obvious in retrospect.
Q: Meaning?
A: It was unclear if society and societal shifts are complex (i.e. they can be modeled), or are just insanely complicated and unable to be modeled. What we found is that societies are indeed complex and miraculously can be modeled.
We were able to predict black swan events months in advance, such as oil price shocks and even the rise of ISIS.
Q: What is XLabs about?
A: Focusing our AI on outcomes too early limits the development of the AI engine, in the way a young child focusing on one thing too early would stunt her development. Variety and curiosity speed development.
Q: Why build an AI moonshot factory with XLabs?
A: I built my own expertise in order to create XLabs and amplify humans. My journey gave me the ability to connect the dots and imagine building moonshots.
I’m grateful my ceaseless pursuit of novelty and futuristic technology has brought together everything needed to solve moonshots — from incredible insights to brilliant technologists and the people who want to bring about humanity’s next phase, where intelligence isn’t just in our devices, it’s in us and our way of life!
Q: How are XLabs’ projects different from those of other AI startups?
Most of AI, even in Silicon Valley, is focused on mimicking what humans already do easily.
From computer vision, to understanding language, to navigating space and objects. We don’t care about that.
XLabs cares about building AI to do what humans cannot: extremely hard problems like protein folding, solving cancer, creating new intelligent portals, and more.
Q: How can Novelty Search help you achieve breakthroughs and moonshots?
A: Novelty Search codifies embracing novelty.
It asks the computer to try something new in its next step, something it has not tried in previous steps. The result is a focus on what’s interesting. It mirrors intellectual curiosity in humans. Novelty Search is the foundation for our culture at XLabs. We believe a culture of curiosity and embracing interestingness creates the most fertile ground for innovative breakthroughs.
Q: Where can Novelty Search take us?
A: One of the most promising areas is in human biology, genetics, and using that knowledge to discover pathways of disease. For instance, in cancer, it could find a match between the genetic signature of a tumor and the best molecular signature of treatment, be it proteins, chemicals, or other therapeutics and treatments.
Q: How is Novelty Search useful for achieving difficult scientific discoveries?
A: Many AI algorithms aim to reach an objective while minimizing errors. This leads to incremental improvements but rarely breakthroughs. Novelty Search allows for failure which significantly accelerates the pace of innovation.
Q: When is Novelty Search perhaps less desirable as an approach?
A: The best scientific breakthroughs balance constrained complexity (complex with but with guardrails).
Make the search space too big, and it never finishes trying new things. Make the search space too small and there are not enough interesting things to try (then an objective based algorithm to achieve an incremental improvement might perform better).
Thank you!
More on Dr Radhika Dirks:
A quantum physicist-turned-entrepreneur, Dr. Radhika Dirks is CEO and founder at XLabs, a next-gen artificial intelligence company mining nature’s algorithms to launch moonshots that amplify humans.
Prior to XLabs, Dr Dirks was CEO of Seldn, an AI that accurately predicted global socio-economic disruptions in 12 countries.
She was also a founding member of Shell Technology Ventures, Shell’s venture arm.
Prior to leading deals and spinouts for the $500M energy fund, Dr. Dirks helped move a multibillion project from R&D to commercialization at Shell.
She has also cofounded & as COO headed operations of Rotary Gallop, a data science firm in fintech, which combined game theory with machine learning.
Dr. Dirks has been named as one of the top rising women entrepreneurs in the US.
She has a Ph.D. in quantum computing and a M.S. in nanotechnology from the University of Illinois at Urbana-Champaign.
She has authored 22 peer-reviewed publications and has more than 290 citations on next-gen computing and technology.
She is responsible for having built the world’s best source of entangled photons for quantum computing and inventing a new protocol for quantum teleportation.
Dr. Dirks frequently advises Heads of States and Fortune 100 companies on the future of disruptive technologies.
She is also lecturer at Singularity University for their AI & quantum executive programs.
Her life’s work is centered on bringing novel technology to market. She blames science fiction novels for her obsession with technology.
Where can readers find out more?