Hold the Code #18
Happy finals week, NU!
Hello!
Welcome to Hold the Code, #18.
In this edition, we review stories about an AI incident database, applications of AI to Parkinson's treatment, new research by MIT on using photonics to reinvent AI, and an opinion piece on how China's use of AI.
At Hold The Code, we'd like to wish all of our Northwestern student readers luck on finals this week — and extend our sincerest thank you for being part of our community through this quarter.
Until next week!
Living in the Hall of Shame
Just like car accidents and airplane crashes are tracked to help manufacturers improve the safety and design of their vehicles, the AI Incident Database tracks harmful or potentially harmful AI systems. Currently, it contains 100 incidents, including some well-known ones like #16: Google’s photo-organizing system that tagged Black people as “gorillas.”
This “AI Hall of Shame” was launched in late 2020 by the Partnership on AI, a nonprofit that researches the negative impacts of AI technology, and includes incidents where “AI systems caused or nearly caused real-world harm.” Here’s a breakdown of notable companies who have found their AI creations in the database:
Microsoft has had 2 incidents recorded
Amazon has had 7 incidents recorded
Google has had 16, more than any other company
Sean McGregor, a machine learning engineer, and creator of the AI Incident Database says it is necessary because AI technology allows machines to have a massive impact on people’s daily lives, but the culture of software engineering does not encourage safety.
McGregor states that “My fellow engineers will have an idea that is quite smart, but you need to say ‘Have you thought about how you’re making a dystopia?’ ”
The Doctor's Order: Treating Parkinson's With AI
Every morning Rory Cellan-Jones wakes up, straps 3 sensor devices onto his arm, and switches on a tablet computer. He then spends 20 minutes going through a series of exercises — the data from which is recorded by the sensors and imputed into an AI system. He has Parkinson's Disease, and he's one of the latest patients to begin incorporating AI into his treatment plan.
Parkinson's disease is a brain disorder that leads to shaking, stiffness, and difficulty with walking, balance, and coordination. At this time, there is sadly no cure for Parkinson's, but many are optimistic about the possibility of using AI to monitor and treat patients.
How can AI help?
For now, the primary benefit of using AI is its diagnostic capabilities. By inputting data from sensors, AI systems can learn about patients' symptoms and create ratings aligned with the UPDRS (the unified Parkinson's Disease Rating Score) to calibrate the severity of a patient's symptoms.
Yet, looking ahead, some researchers believe AI devices can actually be used beyond monitoring. A new project supported by a one-million-dollar grant from the National Institute of Health Research explores shooting tiny amounts of current into the patient's muscle to control tremors.
Read Rory's full interview here.
Lighting Up the AI Future
Lightelligence is using photons to illuminate new possibilities in artificial intelligence
There is no computing without energy
"Computing" is performing arithmetic operations. Electronic chips in our computers perform arithmetic operations by switching on and off tens and sometimes hundreds of logic gate transistors (think switches that control the electricity flow in the computer chip). Every time a logic gate transistor switches, power is consumed. A typical computer can perform literally billions of switches every second – that's a lot of energy (and quite an electricity bill)!
AI’s appetite for energy
Artificial intelligence requires a huge amount of computing power. AI computations take up a large percentage of the computing capacity of data centers and a huge amount of energy.
Picture tens of thousands of servers, running continuously, burning millions of dollars worth of electricity.
Lighting up a new path in AI
Lightelligence, an MIT startup, is set out to change this. Instead of using electronic logic gate transistors, their computer chip uses the interference of photons to carry out arithmetic calculations.
"Think of the optical fibers spread across the bottom of the Pacific Ocean, and the light propagating through thousands of kilometers without losing much power. Lightelligence is bringing this concept for long-distance communication to on-chip compute.” — Yichen Shen, CEO of Lightelligence
The benefits?
Lower power consumption.
Lower energy costs.
Increased computing capacity and reduced costs at data centers
Notably, the Lightelligence chips also enable faster calculations, which could be crucial to advancing AI sectors that require fast AI decision speeds, such as in self-driving vehicles. Using #coolphysics, could Lightelligence be lighting up new possibilities in artificial intelligence?
Read the full article here.
Weekly Feature: AI Lessons from China's COVID response
A recently published piece in Fortune claims that China's COVID response highlights AI at its best.
Soon after the Wuhan outbreak began, China's AI companies began refining their algorithms to fight what would become the global pandemic. AI-driven chatbots were created to connect China's rural communities with health care professionals, and machine-learning algorithms were immediately used for pharmaceutical research. In the realms of health monitoring, medical imaging, robotics, and human-computer interaction, AI firms deployed innovative technologies with incredible speed.
Four powerful ways that AI was used
Several state-owned and private sector A.I. firms in China—such as 4 Paradigm, Potevio, Airdoc, and Beijing SEEMMO Technology—created A.I.-based surveillance systems to remotely monitor patients.
Others—such as Beijing Infervision, Beijing Kunlun Medical, Keya Imaging, United Imaging, and Yitu Technology—incorporated A.I. into medical imaging technology to detect COVID-19 cases faster.
A third set of firms: Wuzhu Technology, TMiRob, AUBO Robotics, Keenon Robotics, and Shanghai Mumu Robot, helped minimize exposure to the virus by developing robots that could provide disinfection services, temperature screenings, and contactless meal deliveries in hospitals and medical facilities.
A fourth group: Beijing Unisound, iFlytek, Futong Dongfang, and Yunji Technology, reduced the risk of public transmission by developing voice-based A.I. systems to ensure less contact with surfaces and humans.
What makes China’s AI application so unique?
According to Francois Candelon, "What stands out isn’t how China is using A.I. to tackle the pandemic, but how deep and specialized its health care data, algorithms, and A.I. research are becoming in the process."
Candelon further asserts that industry-specific vertical innovation is critical for sustained success with A.I., and China’s ability to kick-start that cycle in several industries may enable it to take over the leadership of the global A.I. industry in the not-too-distant future.
Written by: Larina Chen, Molly Pribble, and Lex Verb