AI to Improve the World

(See the bottom — we’re looking for new partners.)

Back in October, on one of our recurring walk-and-talks around Oxford, Brody (a computational biologist) and I (a machine learning researcher) shared something that was missing from our PhD work:

“We want to use AI to improve the world around us — in ways nothing else can.”

Twisting through lamp-lit Oxford alleyways, accompanied by tolling church bells, we sketched out a verbal framework for an organization that would tackle important problems using AI. Something that would combine code, strategy, and impact.

RAIL — the Rhodes Artificial Intelligence Lab — was born.


Building an Action Lab

Here’s how we work. We partner with someone doing something meaningful for the world. RAIL creates an action team of PhD and masters students. We work with the partner’s data. Over 8 weeks, we code and build a solution to the problem.

On January 16, we launched our first 8-week cohort. Our team is made of 26 Rhodes Scholars: 13 PhD students, 13 masters’ students. 50% are AI engineers and the other 50% are strategists. We have AI researchers, geneticists, public policy students, trained doctors, social scientists, linguists, and more.

We’ve been blown away by what RAILers have accomplished across four projects:

  • Diagnosing Kids with Sleep Apnea: Team “Dreamworks” is using AI to diagnose sleep apnea in kids better, alongside Dr. Amal Isaiah at the University of Maryland School of Medicine. It costs $3,000 and several months for a diagnosis, and between 1–10% of the population has sleep apnea. RAIL’s results show that machine learning can do pretty accurate diagnosis, instantly, for free.
Dreamworks strategists interrogate the engineers’ results.
  • Triaging Patients Online in Kenya: Working with ConnectMed, a telehealth startup in Kenya, RAILers are creating an AI-powered virtual triage system to get patients the right type of care. The patient answers an adaptive 10-minute questionnaire, and the machine learning system recommends the best format of care for them individually based on patterns in a massive database of health records. Like Netflix recommendations for healthcare.
  • Predicting Heart Attacks: Ischaemic heart disease is the #1 cause of death worldwide, killing 8.8 million people in 2015. Using a massive amount of data, this team has been using multifacted health and personal data to predict adverse cardiac events before they happen. If successful, the solution could be scaled to millions of people.
  • Team Kadmos: Working with researchers at the University of Oxford Faculty of Classics, Team Kadmos — named after Poseidon’s grandson who brought the first alphabet to Greece — is building language models to reconstruct missing words in Ancient Greek inscriptions. Kadmos is leveraging AI to give us a glimpse into a past we’ve never been able to see before — and could be applied to new languages, manuscripts, and histories.
Preparing to present results and share methods at our mid-term retreat.

It’s not easy. Building these solutions means early morning meetings, late nights coding, sometimes-sketchy Skype calls, pouring into models you’ve never learned, hours cleaning data you’ve never seen. And sometimes the models don’t work. And sometimes the data is more annoying – or not there. But this is part of the game, and the end result justifies the adventure.

RAILers — and the partners who help — are truly exceptional people.


Our Impact

We exist to generate impact, solve hard problems, and have fun doing it. In a few weeks, when we’ve written the last lines of code, tidied up the papers, and finished testing our models, we’re confident that each project above will have the ability make an impact on the lives of thousands (to millions) of patients, researchers, kids, and normal people.

That’s what we live for.

But we never predicted some of the great things that would happen along the way:

  • We saw team members who’d never touched a line of code teach themselves to program.
  • We had partners encourage us to write papers and send them to top journals.
  • We had projects generate ideas for startups — in the near future, we’d love to support RAIL alum to launch and scale ideas born from our projects.

3 Key Lessons We’ve Learned

We are building a serious, capable, exciting AI lab. We’ve built our core learnings into the heart of RAIL:

  1. The potential for AI to tackle important social challenges is huge.
  2. Extremely smart people + technology + creativity + structure = scalable impact.
  3. Learn as much as you can during the process. Foster new ideas.

Our Vision

We’re building RAIL into an engine that creates advanced technology solutions to real world problems. Along the way, we’re forming a community of extremely smart, forward-thinking, future-proof leaders.

We’re one part action lab, one part research lab, one part factory.

Magic happens when you cluster future-proof leaders together.

We hope, in the time ahead, to create and grow startup and research ideas from our work. We also hope to create a more diverse RAIL, increasing the number of women and minorities.

Maybe our drive to combine powerful technology with real, meaningful problems is best expressed by how we sign off on our emails:

“Ad Astra”

To the stars.


Partner With Us for a Project

We are excited to call for partners for our next round of RAIL projects. Teams begin on April 24th.

If you are doing something meaningful in the world, have data, and an interesting challenge — we want to work with you. You can talk to us directly at partners [at] rhodeslab.com, or you can go to our website here. You can email me directly at logan [at] rhodeslab.com.

We take projects in every area, but for this iteration we are especially interested in energy, bioinformatics, education, and government.