Amazon and Hopkins are collaborating to create JHU + Amazon Initiative for Interactive AI (AI2AI), a joint research initiative focused on gaining deeper insights into artificial intelligence (AI) and natural language processing.
The initiative, centered in the Whiting School of Engineering, will be directed by Sanjeev Khudanpur, an associate professor of electrical and computer engineering with appointments in the Center for Language and Speech Processing, the Department of Computer Science and the Human Language Technology Center of Excellence.
Hopkins and the tech giant have worked together in the past. In 2017, Amazon established the Alexa Fund Fellowship, which funds conversational AI research at Hopkins and 17 other universities.
In an interview with The News-Letter, Khudanpur spoke about how AI2AI formalizes the University’s existing relationship with Amazon.
According to him, corporate heads at Amazon were initially reluctant to fund the project. However, the history between Hopkins and Amazon and the University’s previous AI research allowed the project to move forward.
Khudanpur recognizes that Amazon has similar partnerships with other universities, such as the Columbia Center of AI Technology and the Amazon-Virginia Tech Initiative for Efficient and Robust Machine Learning. Despite this, he claimed that the Hopkins-Amazon partnership is developing a unique kind of AI research: interactive AI.
In some kinds of technology, AI is the most intelligent piece of the device that interacts solely with its surroundings, performing actions such as identifying possible cancerous tumors or driving fully autonomous vehicles. However, Khudanpur explained that the Hopkins-Amazon initiative will focus on AI which involves human interaction.
“When you have a conversation, there is so much common grounding. If I said to someone, ‘Astronaut Kieran was really worried; the oxygen generator on moon base five was not functioning,’ we immediately know what the connection is: Kieran is probably human. He needs oxygen to live. The moon doesn’t have it. Think about if an AI can infer all of this,” he said. “This is the challenge with [interactive AI] that we hope to key in on with AI2AI.’”
Interactive AI is intimately associated with natural language processing, a branch of computer science and linguistics that attempts to understand how human language can be interpreted by a computer.
Khudanpur explained the importance of understanding human language to create AI.
“What makes humans different from all of the other animals that we know? It’s the way we collect and pass on knowledge from generation to generation,” he said. “Language holds human knowledge. And so of course, anything that’s going to be AI, it’s going to have to be able to process human language.”
The initiative itself will consist of four distinct nodes. The first supports fellowships for graduate students for experimental projects, similar to basic science research. The second funds applied research projects by Hopkins faculty that may be of interest to Amazon. The third sponsors projects by both Amazon and faculty to achieve a desired product. The fourth node involves outreach efforts to connect faculty to aspiring undergraduate researchers.
Khudanpur expressed excitement about the project’s potential opportunities across many departments.
“I hope that a fun, curiosity-driven project in year one becomes a faculty-driven project in year two, an Amazon-driven project in year three and eventually becomes a product out there,” he said. “In medicine, they call it translation. You want to translate your research into real, useful things. Translating our work into usefulness, even if it’s through a corporation, is a fundamentally good thing.”
Steven David, professor of Political Science, acknowledged that the future will be largely influenced by AI technology. Though AI is not his area of expertise, he highlighted the potential ethical dilemmas AI can pose in an email to The News-Letter.
“For all the good that AI has done (and can do) it can also have worrisome consequences,“ he wrote. “Whether it be in surveillance (government or private) or even the stuff of Science Fiction movies where machines present a threat to humankind, AI needs to controlled. I hope that Hopkins will employ effective and independent safeguards on the research that is envisioned.”
Junior Mark Tiavises, a computer science major, also commented on the ethics of the project in an email to The News-Letter.
“This funding is great for developing new AI technology,” he wrote. “I just hope that with the new research collaboration that they do take the time to discuss the ethical issues that could arise from it.”
Khudanpur recognizes the potential consequences of the partnership due to the many allegations against Amazon for its unethical labor practices. He explained that the team considered Amazon’s business practices when devising the initiative.
“In fact, there’s a funny thing that happened with what to name this thing, and whether Amazon should be part of the name; several people in the University actually got involved, all the way up to the President’s Office,” he said.
However, he acknowledged that these relationships are imperfect, such as when Google employees revolted against a project that attempted to bring Google’s imaging software to military drones.
“All I can say is that we go in with good intentions, and if we see something like [Google’s example], we’ll call it out,” he said. “We’re an academic institution, we don’t have any corporate strings in front of us. If something happens, we are happy to cut off this relationship.”
On Amazon’s controversial nature, Tiavises echoed the sentiments made by Khudanpur.
“I’m sort of ignorant towards Amazon’s labor practices, but I hope this is not something that permeates throughout Amazon,“ he wrote. “I have one friend who works as an intern at Amazon as a software engineer, and he seems to be enjoying his time.”