BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Memento EPFL// BEGIN:VEVENT SUMMARY:U.S. DOD DARPA - Artificial Intelligence Quantified (AIQ) DTSTART;VALUE=DATE:20240625 DTSTAMP:20260411T111310Z UID:8affc2a1698776a20ad21f62550a7ad635452340481cc0e20c97d67b CATEGORIES:Call for proposal DESCRIPTION:The Defense Advanced Research Projects Agency (DARPA) is solic iting innovative proposals in the technical areas of assessing and underst anding the capabilities of artificial intelligence (AI) to enable mathemat ical guarantees on performance of generative AI. Proposed research should investigate innovative approaches that enable revolutionary advances in sc ience\, devices\, or systems. Specifically excluded is research that prima rily results in evolutionary improvements to the existing state of practic e.\n\nArtificial Intelligence Quantified (AIQ) will develop technology to assess and understand the capabilities of AI to enable guaranteed performa nce. The program will test the hypothesis that mathematical methods\, comb ined with advances in measurement and modelling\, will allow guaranteed qu antification of AI capabilities. Specifically\, the program will address t hree interrelated capability levels:\n\n specific problem level\,\n classe s of problem level\, and\n natural class level\, aiming to address the qua ntification and assessment challenges at each level.\n\n\nTwo technical ar eas (TA) of Interest:\n\n\n The goal of TA1 is to provide rigorous foundat ions for understanding and guaranteeing capabilities across levels\; teams proposing for TA1 are expected to be led by individuals with deep technic al expertise\, such as pure or applied mathematics\, theoretical computer science\, or statistics\, or other relevant expertise and demonstrate rele vance to AI.\n  \n The goal of TA2 is to develop methods for evaluating A I models\, integrating and evaluating TA1 results at scale using appropria te research datasets\; teams proposing for TA2 are expected to comprise co mputational\, cognitive\, and/or behavioral scientists with expertise in A I evaluation.\n\nA single proposal may address only one of the TA\; no com bined proposals for TA1 or TA2 will be accepted.\n\nAmount/duration:\nProj ects are expected to have a duration of 36 months\, divided into two phase s of 18 months each. The Funding Opportunity announcement does not mention a specific maximum budget\, the cost realism is one of the evaluation cri teria.\n\nDeadline:\n\n\n Proposers Day: June 14\, 2024\n [REQUIRED] Propo sal Abstract: June 25\, 2024 at 12:00 PM.\n Full proposal: August 13\, 202 4 at 12:00 PM\n\nEligibility / Restrictions:\nEPFL researchers are encoura ged to participate.  For this call\, the U.S. government expects fundamen tal research\, encouraging opensource regimes as well as a collaborative a nd transparent environment.\nThe ReO is available to verify that both the conditions of the funder and interests of EPFL are respected.\n\nHow to Ap ply:\nProposers are required to submit a two-page abstract. Proposers will receive a written response to their abstract either encouraging or discou raging a full proposal submission. This written response will include a ra tionale for the decision.\nProposers may only submit a full proposal if th eir abstract received an encourage response.\n\n Funder Website: https://w ww.darpa.mil/about-us/offices/i2o?ppl=collapse\n Link to call: https://www .grants.gov/search-results-detail/354691\n Application portal link: Abstra cts and proposals must be received via DARPA's BAA Website (https://baa.da rpa.mil)\n\nFor Further Info:\n\n\n Funder contact details: The BAA Coordi nator for this effort may be reached at: AIQ@darpa.mil\n For any EPFL rela ted questions contact the Research Office\n LOCATION: STATUS:CONFIRMED END:VEVENT END:VCALENDAR