Peter Khooshabeh, Ph.D.

I recently had the privilege of leading a podcast for the Conference of Defence Associations Institute with colleagues Dr. Kurtis H. Simpson, Dan Ciuriak, Dr. Robert Mazzolin, and Amos Fox, Ph.D. discussing the future of military AI in a very choppy geopolitical environment. At a time when Canada faces the dual task of improving its defence capability while ensuring economic security, we explore the future of machine learning, automation, and weaponized AI. AI is no longer simply a data processing tool – it's evolved into an autonomous system that's capable of real-time decision making, shaping global militaries in fundamental ways. As these innovations change the face of conflict, interoperability must remain the focus.

14

1 Comment

New Substack blog up on explaining how commercial design theory and military extensions of these (and other) concepts developed through the 20th century as an after effect of the First World War. Much of this earlier unpublished draft work became chapters 1 and 2 in my book, "Understanding the Military Design Movement" which is on sale on Amazon currently for paperback and Kindle versions. Substack link to blog: https://lnkd.in/gGjm_4aB Book link on Amazon: https://lnkd.in/gwX_H92g #design #innovation #strategy #military #defense #complexity #warstudies #creativity #change #complexity #army #NATO #airforce #spaceforce #navy #socom #marines #defence #leadership

16

2 Comments

This morning, Sen. Ted Cruz, chairman of the Senate Commerce Committee, introduced an important new #AI policy legislative framework to guide Congress as it considers #artificialintelligence issues. He also floated an interesting new AI "SANDBOX Act" bill to encourage novel AI governance strategies. This represents an important moment in AI policy and I have written up a quick analysis of the Cruz efforts in my latest R Street Institute piece: https://lnkd.in/ehnbGJui

39

1 Comment

How does SpaceWERX support early-stage space startups and innovative Guardians? SpaceWERX supports early-stage space startups and innovative Guardians through a suite of programs designed to catalyze dual-use tech development, bridge funding gaps, provide business and technical mentorship, and accelerate the transition to Space Force missions. 𝗔𝗰𝗰𝗲𝗹𝗲𝗿𝗮𝘁𝗼𝗿 𝗮𝗻𝗱 𝗖𝗼𝗵𝗼𝗿𝘁 𝗣𝗿𝗼𝗴𝗿𝗮𝗺𝘀 • SpaceWERX runs specialized accelerators, such as the Catalyst Space Accelerator and incubator-style cohort programs, which offer business training, access to end-users, and connections with potential DoD and commercial customers. • Initiatives like the TRL Bootcamp provide startups structured support to mature their tech and prepare for deployment. 𝗙𝘂𝗻𝗱𝗶𝗻𝗴 𝗣𝗮𝘁𝗵𝘄𝗮𝘆𝘀 (𝗦𝗕𝗜𝗥/𝗦𝗧𝗧𝗥, 𝗦𝗧𝗥𝗔𝗧𝗙𝗜/𝗧𝗔𝗖𝗙𝗜) • SpaceWERX leverages the SBIR/STTR programs, helping startups compete for over $4 billion in early-stage, non-dilutive R&D funding awards annually. • Through the STRATFI and TACFI funding mechanisms, SpaceWERX assists qualifying companies (typically SBIR/STTR Phase II awardees) with strategic follow-on investment ranging from $3 million to $15 million to scale promising innovations and cross the “valley of death”. 𝗧𝗲𝗰𝗵𝗻𝗶𝗰𝗮𝗹 𝗩𝗮𝗹𝗶𝗱𝗮𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝗧𝗿𝗮𝗻𝘀𝗶𝘁𝗶𝗼𝗻 𝗦𝘂𝗽𝗽𝗼𝗿𝘁 • Programs like the TRL Bootcamp and "Innovate to Accelerate" align tech development with required government capabilities, facilitate lab access, and connect startups with experienced mentors and program offices for validation and operational transition. 𝗡𝗲𝘁𝘄𝗼𝗿𝗸𝗶𝗻𝗴, 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀, 𝗮𝗻𝗱 𝗜𝗻𝘃𝗲𝘀𝘁𝗼𝗿 𝗘𝗻𝗴𝗮𝗴𝗲𝗺𝗲𝗻𝘁 • SpaceWERX hosts Collider events and Space Challenges, bringing together innovators, end-users, and investors to spark collaboration for new solutions. • Venture-style investor engagement and networking efforts help startups scale and position for follow-on contracts or acquisition. SpaceWERX is assisting early-stage space startups with technical, financial, and strategic resources to validate, mature, and commercialize their dual-use tech for civilian and military markets, to accelerate their impact, growth and commercialization efficiency. Check it out at https://spacewerx.us/ Join our MILMED Connect Funding Insider newsletter >> https://lnkd.in/gndVzFQE #SpaceWERX #MILMEDConnect

11

Just had another paper in the Emerging Technology and Risk Analysis series published today. The title is "Metaverse Concerns About Crimes That Involve Targeting and Exploitation of Children." Here are several of the findings: --The metaverse concept is at an inflection point. How it will expand, how large it will become, and whether will it become an expansive virtual world that directly competes with the physical world remain unanswered questions. The timelines for expansion of the metaverse also remain speculative and will largely depend on use case, demand, and market forces. --Both non-metaverse and metaverse platforms have been and will continue to be used effectively to target children. --The lack of legal frameworks for crimes involving the targeting and exploitation of children in online applications will likely continue to be an issue. Clear definitions for what constitute attacks in the metaverse are not available as they are in the physical world, which hinders investigations and prosecutions. --As the metaverse grows, the potential of these platforms for targeting, luring, exploiting, and victimizing children through targeted criminal activity or child exploitation will likely also grow. https://lnkd.in/eCat96v3

27

4 Comments

Our latest preprint (with Ritwik Gupta) reflects on the #LLM mirage that continues to influence policymaking. As we've suggested elsewhere, compute is a poor proxy for #AI model risk, with downstream consequences for how we think about weaponization #risk both here in the United States and across the globe---with consequences for #exportcontrol, #chip trade, and mil-AI integration moving forward. Read the full piece here: https://lnkd.in/gkADwQtD Berkeley Risk and Security Lab University of California, Berkeley, Goldman School of Public Policy UC Berkeley College of Engineering

66

Reform is needed. The US research ecosystem is burdened by overly complex, duplicative, and occasionally contradictory regulations, requirements, and reporting processes across federal agencies. All of which creates inefficiencies. There is a sensible path forward to balance this burden while keeping America strong and minimizing risk to accountability and research performance. A new roadmap just dropped - one piece in the path to F&A and indirect cost reform for federally sponsored research. The National Academies of Sciences, Engineering, and Medicine on report "Improving Regulatory Efficiency and Reducing Administrative Workload to Strengthen Competitiveness and Productivity of U.S. Research". #ResearchRegulation #ResearchEfficiency #SciencePolicy #HigherEd #InnovationPolicy #FederalResearch #ScienceAndTechnology https://lnkd.in/eAj4FDSu

21

Today, I'm at American Affairs calling for a paradigm shift on AI regulation: we need to break out of the deadlock between hand-wringing safetyism and cocky accelerationism. We must pivot from talking about future threats to observable harms--above all those to our kids. "We must innovate to beat China," we're told. But faced with dwindling state capacity and unwilling to pay for AI dominance with taxpayers’ money, we seem content to pay instead with the minds and souls of our children. As an initial step toward taming this technology, I propose treating generative AI like other very powerful technologies that we keep out of the hands of children, at least until it can meet design standards that avoid exploiting their vulnerable developing minds and emotions. We will be told that this is "anti-innovation." But that's not how innovation works. In reality, innovation is spurred by the need to overcome obstacles, challenges, or constraints—or to respond to demanding design standards. Prudent regulation of AI, far from stifling American innovation, may provide precisely the creative spurs that will unleash its full potential. Read the full essay here: https://lnkd.in/eFFvHZ2Q

12

Excited to share our collaborative work with Professor Ian Davidson at UC Davis, where we’re exploring groundbreaking approaches in AI design. We’re challenging the status quo by moving beyond AI as a black box. Our focus is on developing systems that embody reasoning, reflection, and trustworthiness from the ground up. Drawing inspiration from cognitive science and neuroscience, we’re working toward architectures that can dynamically revisit, re-evaluate, and re-interpret information — mirroring how humans process complexity in real time. A central part of this effort is the counterfactual framework we’re developing. Counterfactual reasoning — asking “what if?” and reasoning across alternatives — is not only essential for human decision-making but also a stress test for how (and whether) large language models truly “reason.” By probing LLMs through counterfactuals, we can expose both their strengths and their blind spots. Just as importantly, these same insights help us better understand human logical fallacies in decision-making, giving us a two-way mirror: advancing both AI and cognitive science. The implications are far-reaching across national security, healthcare, and beyond. By creating AI that reasons in a human-aligned way, we’re laying the groundwork for auditable, trustworthy, and collaborative systems that can perform in the most critical scenarios. Our collaboration merges UC Davis’ depth in AI theory with our team’s applied mission focus — translating bold concepts into real-world systems for defense, medicine, and logistics. I’m excited about what lies ahead. While much of the community emphasizes “explainability,” our pursuit goes further: toward understandability and mutual reasoning. That may be the true next frontier for AI. How do you see counterfactual reasoning shaping the future of AI — and what might it reveal about our own human decision-making? #AI #CognitiveScience #Innovation #Collaboration

27

1 Comment

Traffic stops are one of the most common settings of civilian and police interactions. Does race influence these interactions? Our new paper in @Science uses Lyft rideshare data to isolate the effect of race from other factors. Our findings are striking: Minority drivers are 24-33% more likely to receive speeding tickets and pay 23-34% more in fines than White drivers—even with identical speeding behavior and context. We find no differences in accident or re-offense rates to justify these disparities, suggesting racial bias. From an academic perspective, our study tackles the "denominator problem" in the economics of crime literature. In addition, it showcases the value of academic/private partnerships to make positive change (huge thanks to Lyft). For policymakers, these findings highlight the potential of automated enforcement and fairer insurance policies to reduce racial profiling. Paper available here: https://lnkd.in/e8ffJ9ia Wonderful coauthors Pradhi Aggarwal, Alec Brandon, Ariel Goldszmidt, Justin Holz, Ian Muir, Gregory Sun, Thomas Yu. Lyft Uber Walmart University of Wyoming University of Chicago Universidad del CEMA The Australian National University 

185

8 Comments

New insights from a CSET-hosted, National Science Foundation (NSF)-funded, workshop on AI and open science: AI is poised to revolutionize how we do research—from extracting scientific claims to improving metadata curation and expanding science communication. But there's a catch. The reality check:  1️⃣ Hallucinations and bias mean human oversight and teaming aren't optional — they are essential. 2️⃣ Current AI benchmarks often fall short for research contexts.  3️⃣ Human-AI teaming is domain-specific tuning. What's next? The research community is prioritizing:  → Research-specific AI evaluation metrics  → Shared infrastructure to prevent duplication  → Case studies proving open science impact  → Controlled trials to test open science interventions The consensus? AI should accelerate discovery while preserving the rigor that makes science trustworthy. Read more here: https://lnkd.in/gi_gb6MP Are you seeing AI transform your research workflows?

28

1 Comment