A personal assistant that knows the best time to wake you up. A machine that reads radiology scans together with a pathologist. More and more artificial intelligence (AI) applications have become a reality or are on their way. Daniela Rus, director of MIT's CSAIL (computer science and artificial intelligence laboratory), said the technology has also created challenges. If we can handle them well, AI will improve our lives significantly.
Rus shared her thoughts in a speech titled "AI and humanity: Opportunity and perils." It was part of Building a Better World, a master series organized by the Epoch Foundation and MIT Sloan School of Management. The online series is scheduled to release on January 7, 2022.
Several technological marvels that had previously been seen in science-fiction movies have become part of real life, thanks to AI. For example, Rus said, research in human-robot interaction enables a machine to monitor a person's muscle activity and adapt to that person's needs as if by magic.
The technology could benefit the medical field. Rus said a new AI-based approach was tasked with reviewing images of lymph node cells and diagnosing cancer. When the AI system and a pathologist worked together, the error rate was only 0.5%, much lower than that of an unassisted human.
"Imagine a future where every practitioner, even those working in small practices in rural settings, have access to these systems. An overwhelmed doctor may not have the time to review every new study and clinical trial. But working in tandem with the systems, the doctor will offer patients the most cutting-edge diagnosis and treatment options," Rus said.
AI has tremendous potential. Rus asked the audience to imagine having a wearable computer assistant like the fictional superhero Iron Man does, but with the focus on improving the user's situational awareness, health, and everyday life.
The director said exosuits could provide an extra pair of eyes that warn people of any threats. Furthermore, embedded sensors could monitor everything from a person's posture to digestion, and could even help the user make the right adjustments to avoid back pain.
AI has brought technical and societal challenges
The progress mentioned above will not happen overnight. Rus said the improvement is enabled by advances in three interconnected fields: robotics, AI, and machine learning. Robotics gives machines the ability to move. AI gives machines decision-making power. Machine learning enables systems to learn from known data and to make predictions regarding unknown data.
In addition, the increasing adoption of AI has created various technical and societal challenges. For example, Rus said to get major breakthroughs, more funding is required for more computational infrastructure. Massive, high-quality data sets are essential to ensure that algorithms reliably demonstrate good performance.
While the democratization of AI will improve our lives, Rus said we must anticipate and respond to the economic inequality it could create, including displacing human workers from the jobs they do today. Beyond that, as more data are fed into AI systems, the risk to privacy will grow, as will the opportunities for authoritarian governments to leverage these tools to curtail freedom and democracy.
"These problems aren't like the pandemic," Rus said. "We know they're coming. We can set out to find a solution at the intersection of policy, technology, and business – in advance. Now."
She offered six principles for an ethical approach: responsibility, equitable standards, traceability, reliability, governance, and sustainability. She said human beings should exercise appropriate levels of judgment and remain responsible for the development and outcomes of AI systems. Equitable standards must be enforced impartially, without fear or favor. Deliberate steps are needed to avoid biases and their consequences.
Rus said traceability means the AI engineering discipline should be sufficiently advanced to allow troubleshooting and technical experts should possess an appropriate understanding of the technology and the development processes of AI systems.
Reliability refers to an AI's reliable behavior in a well-defined domain of use. Governance means the technology should fulfill its intended function while safeguards detect any unintended disruptions.
Finally, Rus said we need to pay attention to the sustainability of all technological advances, including how they might impact our climate and the carbon footprint they will create.