This semester, college students and postdocs throughout MIT had been invited to submit ideas for the first-ever MIT Ignite: Generative AI Entrepreneurship Competition. Over 100 groups submitted proposals for startups that make the most of generative synthetic intelligence applied sciences to develop options throughout a various vary of disciplines together with human well being, local weather change, schooling, and workforce dynamics.
On Oct. 30, 12 finalists pitched their ideas in entrance of a panel of skilled judges and a packed room in Samberg Conference Center.
“MIT has a responsibility to help shape a future of AI innovation that is broadly beneficial — and to do that, we need a lot of great ideas. So, we turned to a pretty reliable source of great ideas: MIT’s highly entrepreneurial students and postdocs,” mentioned MIT President Sally Kornbluth in her opening remarks at the occasion.
The MIT Ignite occasion is a part of a broader give attention to generative AI at MIT put forth by Kornbluth. This fall, throughout the Institute, researchers and college students are exploring alternatives to contribute their data on generative AI, figuring out new purposes, minimizing dangers, and using it for the good thing about society. This occasion — co-organized by the MIT-IBM Watson AI Lab and the Martin Trust Center for MIT Entrepreneurship, and supported by MIT’s School of Engineering and the MIT Sloan School of Management — impressed younger researchers to contribute to the dialogue and innovate in generative AI.
Serving as co-chairs for the occasion had been Aude Oliva, MIT director of the MIT-IBM Watson AI Lab and a principal investigator in the Computer Science and Artificial Intelligence Laboratory (CSAIL); Bill Aulet, the Ethernet Inventors Professor of the Practice at the MIT Sloan School of Management and director of the Martin Trust Center; and Dina Katabi, the Thuan (1990) and Nicole Pham Professor in the Department of Electrical Engineering and Computer Science, director of the Center for Wireless Networks and Mobile Computing, and a CSAIL principal investigator.
Twelve groups of scholars and postdocs had been competing for quite a few prizes, together with 5 MIT Ignite Flagship Prizes of $15,000 every, a particular first-year undergraduate pupil crew Flagship Prize, and runner-up prizes. All prizes had been supplied by the MIT-IBM AI Watson Lab. Teams had been judged on their challenge’s revolutionary purposes of generative AI, feasibility, potential for real-world impression, and the standard of presentation.
After the 12 groups showcased their know-how, its potential to deal with a difficulty, and the crew’s potential to execute the plan, a panel of judges deliberated. As the viewers waited for the outcomes, remarks had been made by Mark Gorenberg ’76, chair of the MIT Corporation; Anantha Chandrakasan, dean of the MIT School of Engineering and the Vannevar Bush Professor of Electrical Engineering and Computer Science; and David Schmittlein, the John C. Head III Dean and professor of promoting at the MIT Sloan School of Management. The pupil winners included:
MIT Ignite Flagship Prizes
eMote (Philip Cherner, Julia Sebastien, Caroline Lige Zhang, and Daeun Yoo): Sometimes figuring out and expressing feelings is tough, significantly for these on the alexithymia spectrum; additional, remedy may be costly. eMote’s app permits customers to establish their feelings, visualize them as artwork utilizing the co-creative technique of generative AI, and mirror on them by means of journaling, thereby aiding college counselors and therapists.
LeGT.ai (Julie Shi, Jessica Yuan, and Yubing Cui): Legal processes round immigration may be sophisticated and expensive. LeGT.ai goals to democratize authorized data. Using a platform with a big language mannequin, immediate engineering, and semantic search, the crew will streamline a chatbot for completion, analysis, and drafting of paperwork for corporations, in addition to enhance pre-screening and preliminary consultations.
Sunona (Emmi Mills, Selin Kocalar, Srihitha Dasari, and Karun Kaushik): About half of a health care provider’s day is consumed by medical documentation and scientific notes. To deal with this, Sunona harnesses audio transcription and a big language mannequin to rework audio from a health care provider’s go to into notes and have extraction, affording suppliers extra time in their day.
UltraNeuro (Mahdi Ramadan, Adam Gosztolai, Alaa Khaddaj, and Samara Khater): For about one in seven adults, spinal twine damage, stroke, or illness will induce motor impairment and/or paralysis. UltraNeuro’s neuroprosthetics will assist sufferers to regain a few of their each day skills with out invasive mind implants. Their know-how leverages an electroencephalogram, sensible sensors, and a multimodal AI system (muscle EMG, pc imaginative and prescient, eye actions) educated on hundreds of actions to plan exact limb actions.
UrsaTech (Rui Zhou, Jerry Shan, Kate Wang, Alan He, and Rita Zhang): Education right now is marked by disparities and overburdened educators. UrsaTech’s platform makes use of a multimodal massive language mannequin and diffusion fashions to create classes, dynamic content material, and assessments to help lecturers and learners. The system additionally has immersive studying with AI brokers for energetic studying for on-line and offline use.
First-Year Undergraduate Student Team MIT Ignite Flagship Prize
Alikorn (April Ren and Ayush Nayak): Drug discovery accounts for important biotech prices. Alikorn’s massive language model-powered platform goals to streamline the method of making and simulating new molecules, utilizing a generative adversarial community, a Monte-Carlo algorithm to vet essentially the most promising candidates, and a physics simulation to find out the chemical properties.
Runner-up Prizes
Autonomous Cyber (James “Patrick” O’Brien, Madeline Linde, Rafael Turner, and Bohdan Volyanyuk): Code safety audits require experience and are costly. “Fuzzing” code — injecting invalid or surprising inputs to disclose software program vulnerabilities — could make software program considerably safer. Autonomous Cyber’s system leverages massive language fashions to mechanically combine “fuzzers” into databases.
Gen EGM (Noah Bagazinski and Kristen Edwards): Making knowledgeable socioeconomic improvement insurance policies requires proof and knowledge. Gen EGM’s massive language mannequin system expedites the method by analyzing and analyzing literature, after which produces an proof hole map (EGM), suggesting potential impression areas.
Mattr AI (Leandra Tejedor, Katie Chen, and Eden Adler): Datasets which might be used to coach AI fashions typically have problems with range, fairness, and completeness. Mattr AI addresses this with generative AI with a big language mannequin and steady diffusion fashions to enhance datasets.
Neuroscreen (Andrew Lu, Chonghua Xue, and Grant Robinson): Screening sufferers to doubtlessly be a part of a dementia scientific trial is expensive, typically takes years, and largely outcomes in an ineligibility. Neuroscreen employs AI to extra shortly assess sufferers’ dementia causes, resulting in extra profitable enrollment in scientific trials and therapy of situations.
The Data Provenance Initiative (Naana Obeng-Marnu, Jad Kabbara, Shayne Longpre, William Brannon, and Robert Mahari): Datasets which might be used to coach AI fashions, significantly massive language fashions, typically have lacking or incorrect metadata, inflicting concern for authorized and moral points. The Data Provenance Initiative makes use of AI-assisted annotation to audit datasets, monitoring the lineage and authorized standing of information, enhancing knowledge transparency, legality, and moral issues round knowledge.
Theia (Jenny Yao, Hongze Bo, Jin Li, Ao Qu, and Hugo Huang): Scientific analysis, and on-line dialogue round it, typically happens in silos. Theia’s platform goals to carry these partitions down. Generative AI know-how will summarize papers and assist to information analysis instructions, offering a service for students in addition to the broader scientific neighborhood.
After the MIT Ignite competition, all 12 groups chosen to current had been invited to a networking occasion as a direct first step to creating their ideas and prototypes a actuality. Additionally, they had been invited to additional develop their ideas with the help of the Martin Trust Center for MIT Entrepreneurship by means of StartMIT or MIT Fuse and the MIT-IBM Watson AI Lab.
“In the months since I’ve arrived [at MIT], I’ve learned a lot about how MIT folks think about entrepreneurship and how it’s really built into everything that everyone at the Institute does, from first-year students to faculty to alumni — they are really motivated to get their ideas out into the world,” mentioned President Kornbluth. “Entrepreneurship is an essential element for our goal of organizing for positive impact.”