Reimagining the Future of AI for National Security


The DARPA Information Innovation Office (I2O) will host “AI Forward,” an initiative to explore future directions for AI. DARPA invites prospective attendees to apply to attend (see RFI here). As a part of the application, individuals will be asked to submit an executive summary outlining a compelling research direction toward trustworthy AI with applications for national security.


In the past decade, we have seen impressive advances in the field of Artificial Intelligence (AI), particularly in data-driven machine-learning technologies. However, AI that functions with suitable levels of trustworthiness for many national security applications does not yet exist. To be trusted, such a system must reliably operate competently, interact appropriately with people, and meet pressing national security needs in an ethical manner. Creating such systems will require significant advances in (at least) three key areas:

  • Foundational theory, to understand the art of the possible, bound the limits of particular system instantiations, and inform guardrails for AI systems in challenging domains such as national security;
  • AI engineering, to predictably build systems that work as intended in the real world and not just in the lab; and
  • Human-AI teaming, to enable systems to serve as fluent, intuitive, trustworthy teammates to people with various backgrounds.

These thrusts emerged as important directions in the course of drafting DARPA’s current “AI Analytic Framework” and are consistent with the National AI R&D Strategic Priorities. In addition, we need to dramatically improve the efficiency of machine-learning (ML), in terms of both data and compute, to make ML-based systems cost-effective for many national security applications.

About the Workshops

AI Forward will consist of two workshops (one in-person and one virtual) in summer 2023. During these highly interactive and collaborative workshops, participants will work together to envision new research directions that could lead to trustworthy AI suitable for national security applications. Participation is limited to approximately 100 people per workshop. Each workshop will include discussions and guided brainstorming sessions with industry, academic, and government leaders and experts on current and future research directions in AI.

The workshops are intense, interactive, and free-thinking. They aim to stimulate thinking about promising new research directions that span the many diverse domains of AI, including human language technology, large language models, generative models, the philosophy and ethics of AI, and embodied cognition with an eye to future national-security related applications. Participants will be expected to engage constructively in dialogue with each other, the facilitators, and the Subject Guides (experts in the field who will help to frame and guide discussions) to develop collaborative research themes to help define possible future DARPA activities. The goal is to bring individuals from different disciplines together to interact and engage in free-thinking on first principles, to learn from one another, and to create an integrated vision for future research thematic areas.

Through the course of the workshop, participants will form small groups, pitch and refine themes (based on input from Subject Guides) for interdisciplinary topic areas that advance scientific questions about the future of AI and how to solve the problem of capable, trustworthy AI that partners well with humans in the context of challenging, real-world problems.

AI Forward will be facilitated by KnowInnovation, a company that specializes in working with science groups to accelerate innovation and achieve strategic outcomes. Participants can expect highly interactive virtual and in-person events. The majority of time will be spent in solution-focused discussions after a few targeted plenary lectures designed to set the stage.

Event Outcomes

The ultimate output of the workshops will be the identification of approximately 40 promising areas for future research. For each area, participants will explore items like open issues, potential research questions, desired impact, and needed expertise to make progress. Participants will make a presentation (2-3 minutes) for each area and provide a companion white paper (2-3 pages). DARPA will consider these presentations and white papers in forming its AI strategy going forward. As a part of that strategy, DARPA will likely create two (2) new Artificial Intelligence Exploration projects (AIEs) based on the outputs of the two AI Forward Workshops.

Participants will be able to submit their work for consideration to these AIEs, which will be announced broadly to solicit proposals.

Who Should Apply

Each workshop will bring together experts in AI Theory, Computer Vision, Human-Machine Symbiosis, Knowledge Representation & Reasoning, Machine Learning, Meta-Cognition, Large Language Models, Generative AI, Theory of Machine Learning, and more. It is anticipated that the interdisciplinary collaborations formed during these events will result in the identification of new and promising areas for future research. Participants will explore each area for open issues, potential research questions, desired impact, and needed expertise to make progress.

Organizers will select approximately 100 participants from the applicant pool for each workshop on the basis of the interests, expertise, and other characteristics solicited in the application, including date and format preference. Most participants are expected to be academic faculty or industry and government researchers, however, original thinkers from outside academia, industry, and government are also encouraged to apply. When selecting participants, organizers will strive to balance a range of diverse disciplinary experience and expertise. All participants should be willing to engage in frank discussions and assessment of ideas in a collegial and professional fashion. To facilitate open sharing, all meeting discussions will be considered a private communication and not to be shared outside of the meeting, unless approved by the contributor.

The solicitation and collection of personal information by the website host is solely for the purpose of site administration, application tracking, communication with applicants, and to provide updates. Other than for this use, personal information will not be maintained, used, stored, indexed, or retrieved from any information system/database without your prior notice, choice, or consent. Personal information will be safeguarded using appropriate physical, administrative, and technical protocols.

U.S. citizens and Foreign Nationals (FNs) are welcome to apply. All FNs who are invited to attend the virtual or in-person event must complete either a) a DARPA Form 60 (for academia, industry, or individual researchers, only) or b) an Official Visit Request which is generated and submitted by the respective country’s Embassy (for government organizations, only) beforehand.

How to Apply

The application is now closed. Thank you for your interest.

Virtual: June 13 – 16, 2023
Tues - Fri; 12:00 - 5:00 ET with a one-hour break each day. Space is limited and programming is very interactive, full participation is expected. Please block your calendars for the duration of the event.

In-Person: July 31 - August 2, 2023 in Boston, MA
Mon 11:00 - 6:00 ET, Tues 9:00 - 6:00 ET, Weds 9:00 - 5:00 ET

To cover in-person meeting costs, a conference registration fee will be charged per attendee. Invitees have the option to apply for a limited number of scholarships that will cover the conference fee and travel expenses within the continental U.S. in accordance with Department of Defense (DoD) Joint Travel Regulations (JTR).


Inquiries can be submitted via UNCLASSIFIED email to