The Pentagon's Silicon Valley Problem
Commenters use a Harper’s piece on “The Pentagon’s Silicon Valley problem” as a springboard to examine how AI and big-data surveillance are reshaping modern warfare and intelligence — often with overhyped promises and opaque oversight. Many argue that systems used by Israel, the U.S. and others can entrench confirmation bias, enable mass surveillance and automate targeting without clear accountability, while military procurement and politics, not lack of technology, are seen as the real bottlenecks. Others focus on the ethical dimension, noting that many technologists are wary of defense work and warning that any tool powerful enough to predict and kill at scale will inevitably be turned inward on domestic populations or abused by future leaders.
Surveillance, AI, and Civil Liberties
- Many comments worry that AI-enabled surveillance will not stay confined to “terrorists” but drift to monitoring everyone via public and commercial data, including purchased “reports” from private firms.
- Slippery-slope concerns: definitions of “terrorist,” “white nationalist militia,” or “hanging out” can be broadened over time, leading to guilt by association or even co-location.
- Some trust legal safeguards (warrants, courts) while others highlight opaque processes like FISA/FISC and secret precedents, arguing that oversight is inadequate.
- There’s anxiety about LLMs and APIs as rich data-collection tools, with fears of intelligence agencies obtaining conversational histories.
Intelligence Failures and October 7
- Several point out that Hamas trained openly, locals warned, and conscript analysts flagged the threat but were ignored. This is framed as organizational and cultural failure, not an “AI failure.”
- Explanations offered: overreliance on models that confirmed prior assumptions, racial arrogance, bureaucratic rigidity, internal political chaos, and hubris.
- Some float deliberate negligence or “letting it happen” for political gain; others say this is conspiratorial and that incompetence is more plausible.
- Parallels are drawn to Russia’s Crocus City Hall attack, where US warnings were reportedly broad and not fully acted upon.
AI/ML Capabilities and Hype
- Multiple commenters distinguish traditional ML from current LLM hype, criticizing the blanket use of “AI” as misleading and marketing-driven.
- Tools like Project Maven and Palantir are seen by skeptics as over-claimed, used to impress leadership rather than deliver proven battlefield value.
- Others argue ML has long been useful, but failure stems from human misuse, bad incentives, and over-trusting dashboards while discounting human reports.
US Strategy, Wars, and Tech
- Long threads debate US strategic failures since WWII (Afghanistan, Iraq, Libya, Syria) as stemming from unclear or unachievable goals and cultural hubris, not lack of tech.
- Comparisons are drawn with more focused interventions (Desert Storm, Yugoslavia) and with historical occupations of Germany/Japan versus Afghanistan’s very different realities.
Silicon Valley, Defense, and Procurement
- Some say many tech workers avoid defense on moral grounds; others think that with top-tier pay and autonomy many would overcome objections.
- Hiring barriers (security clearances, drug use, pay caps, no remote work) and Byzantine procurement/FedRAMP processes are seen as major frictions.
- Commenters describe a “quasi-Soviet” procurement system that favors large incumbents and acquisition of startups, limiting fresh innovation.
Gaza, Targeting, and Casualty Numbers
- There is discussion of reported Israeli use of AI-based target selection systems in Gaza, with critics calling this a “mass assassination factory” that launderes responsibility.
- Others question casualty figures and source reliability, warning that Hamas-linked numbers may be manipulated; still, no one disputes that civilian deaths are massive.
- Broader worry: software will be used to diffuse accountability for lethal decisions and normalize large-scale, automated targeting.