To the Stars with Data: July 25 2021
Resolution to the protein folding problem, solar sails, multi-modal AI, and more...
Hello Datanauts!
Do you ever freeze up when asked “what do you want to be in 10 years” or “what do you want to do with your life”? You aren’t alone.
Here’s some great advice I heard this week that makes me feel so much better:
Your long-term goals aren’t always declaring who you will be in the future. Instead, think of them as guides to direct your actions now. It’s okay (and expected) if they change as you change.
After hearing that, I felt a weight lift off my shoulders. I no longer feel immense pressure to have everything figured out and instead feel like I can just work on “me”.
Speaking of future selves, I hope you enjoy this week’s To the Stars with Data!
Joe
Thought-Provoking Data Viz
Quick Hits. This interactive map from the New York Times displays recent and ongoing wildfires in the western United States. The fires shown in the visualization are reported by the Wildland Fire Interagency Geospatial Services group and are mapped according to the NASA FIRMS satellite-based fire detection system.
My Thoughts. Within the past week, New York City (where I live) had one of the worst air qualities of any major metropolitan area on Earth. The reason? Wildfires from the western United States. Primarily due to the Bootleg Fire in Oregon, it’s likely that other wildfires in western North America contributed.
Earth News
🧬 Protein Folding AI Is Making a ‘Once in a Generation’ Advance in Biology
Quick Hits. DeepMind’s AlphaFold2 AI shocked the world in December when it could predict protein folding much more quickly than humans. Now, DeepMind has published AlphaFold2’s source code for anyone to access. They also released a database with the predicted protein structures for over 350,000 proteins (including all ~20,000 expressed by the human genome). In the coming months, they will complete the database with almost all ~100 million proteins known to humans.
My Thoughts. At the moment, almost all AI we produce is classified as Artificial Narrow Intelligence (ANI), or AI that’s great at one task. The “protein folding problem” is a great candidate for our ANI as it requires high computing power but a narrow intelligence. Other than the huge strides AlphaFold2 will allow in medicine, I’m particularly excited about such a perfect application of AI as a tool to augment our work.
Other Earth News
There was too much great news this week to pick just one article, so here are a few headlines:
🚗 Autonomous Vehicles (AVs).
Ford and Argo AI will launch AVs on Lyft in Miami by end of 2021
Mobileye has begun testing their AVs in New York City. Mobileye is the first AV company to obtain permits to operate in all 5 boroughs
Tokyo is using AVs to shuttle athletes in the Olympic Village
🩺 Biotech.
A new drug successfully killed breast cancer cells in mice
First new-generation artificial heart transplanted into human in US
🌧 Weather. Dubai made it rain by electrifying clouds with drones
Space News
🛰 NASA is sending a solar sail satellite to a nearby asteroid
Quick Hits. NASA is sending its NEA Scout to 1991 VG, a near-Earth asteroid (NEA). Notably, NEA Scout will navigate to the asteroid with a solar sail. Once the craft arrives at 1991 VG, it will capture high-quality pictures of the asteroid. This will be the first time solar sail propulsion is used for an interplanetary mission.
My Thoughts. Solar sails are so promising because they require no fuel. Due to this lack of fuel, crafts that utilize solar sails can navigate in unique ways compared to crafts that rely on other propulsion systems. One example includes positioning crafts at the poles of the sun instead of orbiting around the Ecliptic. Thus, solar sails could help us unlock new secrets of our solar system.
📡 Masten to create Moon GPS
Quick Hits. Masten Space Systems wants to create a GPS-like system on the moon as part of a contract awarded through the Air Force Research Laboratory’s AFWERX program. To accomplish this, they plan to plant beacons on the moon that broadcast their location to nearby craft, enabling location tracking. They aim to have a prototype of the beacons prepared in 2023.
My Thoughts. As the article notes, current spacecraft must be equipped with specialized equipment to navigate the moon. Considering that getting stuff off Earth costs a lot of money, reducing necessary components on spacecraft can significantly reduce costs. For context, the European Space Agency is also working on a GPS system for the moon, so we will likely see more updates on this topic soon.
Things that make you think
🤖 DeepMind’s “Perceiver” AI accepts multiple inputs
Quick Hits. DeepMind (Alphabet’s AI company) created an AI called Perceiver that can process multiple inputs (images, videos, point clouds, and more). Most AI only process one input type, such as images or text. Perceiver performs well with images, videos, and others, but performs poorly in point clouds.
My Thoughts. AI often seems so intelligent, but we are mostly still in a state of Artificial Narrow Intelligence (ANI)—AI that’s great at one task. In order to reach a state of human-level intelligence, we need to build AI that can process many types of input. Considering that Perceiver performs on par with other AI networks on many of its inputs, I’d consider Perceiver a big step in the direction of human-level intelligence.
Your Thoughts. Do you think AI processing multiple types of inputs is a big deal? At what point do you think an AI gains the intelligence of a human?
Future-Thinking Companies
🏠 Mighty Buildings. 3D-prints homes more sustainably than typical construction companies.
👚 Natural Fiber Welding, Inc. Manufactures plant-based textiles that are usually made from petroleum (oil).