How Does AI Thought Work?
Do androids dream of electric sheep?
Good Evening:
Today on #DogShirtTV, the estimable Alicia Wanless and the estimable Eve Gaumond joined me to discuss the craziest recent events in Canadian politics, including lies, plagiarism, and threats of secession. The estimable Anastasiia Lapatina joined as well to workshop the final last episode of her podcast on US-Ukrainian relations. Klio, the estimable dog on my shirt, joined me to squirm, rattle the fence, and generally be a pest:
The Situation
In today’s “The Situation” column, I outline the Trump administration’s use of immigration laws as a means of political repression—to justify hundreds of visa revocations and its detention of foreign students. I emphasize the administration’s efforts to persecute and silence those with a dissenting view of Israeli policy:
True, the persecution has to be limited to people who are not citizens, owing to the nature of the legal theory underlying it. But that’s actually a lot of people. And having a policy that all-but-openly announces itself as pursuing the detention and expulsion of any non-citizen who is openly opposed to Israeli policy represents one of the most dramatically repressive and broad-based assaults on free speech in this country’s modern history.
Today On Lawfare
Compiled by the estimable Caroline Cornett
When Did the Betrayal Begin?
Bryce Klehm highlights the United States’s various failures in upholding the Special Immigrant Visa (SIV) program—which provided visas to Afghan locals who worked for the U.S. during the war in Afghanistan. Klehm explains how the United States’s betrayal of its Afghan allies represents the strategic narcissism largely governing U.S. foreign policy:
Those who opposed the SIV program were the embodiment of strategic narcissism. Forget the fact that those interpreters are hunted by the Taliban and that U.S. allies can lose their trust in the United States. For the opponents, domestic politics had priority over U.S. strategic interests. Those on Capitol Hill could have improved the program and recognized the lived reality of those executed by the Taliban. Instead, they chose to tie the U.S.’s commitment to its allies to diversity visas and unrelated amendments. And they bargained as if trust in the United States is an infinite operating assumption for any ally. They either refused to see or ignored the fact that the failure to protect local allies who trusted the U.S. leads to broader “[p]erceptions that the United States cannot or will not honor its promises.” Put simply, the SIV program was a promise. Their yearly bargaining broke it.
EDNY Opinion in Hasbajrami Undermines FISA 702
George Croner argues that U.S. District Judge LeShann DeArcy Hall’s opinion in U.S. v. Hasbajrami wrongly relied on law enforcement analytic frameworks to evaluate the querying of the Section 702 database. Croner explains that Hall’s opinion—which concludes that using a U.S. person query term to query the database of communications lawfully acquired under Section 702 violates the Fourth Amendment—will significantly harm national security anti-terrorism and cybersecurity efforts:
The EDNY’s mistaken approach colors its entire opinion. The EDNY insists the government has provided inadequate explanation for the time elapsing between the queries that retrieved Hasbajrami’s communications from the database, which thus indicates that the government’s investigation of Hasbajrami was not time sensitive and that procuring a warrant would not have interfered with that investigation. But foreign intelligence review and analysis are intrinsically more spontaneous and dynamic than the linear progression that apparently characterizes the EDNY’s experience with criminal investigations. Hasbajrami was not the target of the anti-terrorism investigation; instead, the Section 702 database was queried using his personal identifiers only as intelligence analysts determined that such queries were likely to retrieve foreign intelligence information that would shed light on the activities of those engaged in the “Islamic fundamentalist terrorism operations” that were the focus of the Joint Terrorism Task Force.
Verifying Who Pulled the Trigger
In the latest installment of Lawfare’s Foreign Policy Essay series, Zachary Kallenborn calls for increased use of post-hoc verification to determine how lethal autonomous weapons are used. He argues that verification will deter future transgressions and improve the international community's ability to punish those who violate restrictions:
A paradigm shift is needed to ensure accurate verification for regulating these new systems. Traditional arms control verification is production and acquisition oriented, which makes sense because nuclear, biological, and chemical weapons all require specialized material, equipment, and skills. But a simple autonomous weapon just requires basic programming skills; hobbyists have even built their own at-home autonomous paintball sentry turrets. Instead, arms control approaches should emphasize in-situ and post-hoc verification: determining whether a controlled weapon is being or was used in combat, and inflicting appropriate levels of punishment in response. Although such an approach will not deter autonomous weapons use completely, it may help encourage global norms around responsible use of autonomous weapons and deter particularly egregious use.
Podcasts
On Lawfare Daily, I sit down with Roger Parloff, Scott R. Anderson, James Pearce, and David D. Cole to discuss the status of the civil litigation against President Donald Trump’s executive actions, including the deportation of individuals under the Alien Enemies Act, the detention of Mahmoud Khalil and others, and the targeting of law firms:
On Escalation, Anastasiia Lapatina and Tyler McBrien cover the infamous 2019 phone call between U.S. President Donald Trump and Ukrainian President Volodymyr Zelenskyy, the scandal that followed, and how it affects congressional support for Ukraine to this day:
Today’s #BeastOfTheDay is a screech owl seeing things he’s never seen before:
Tiktok failed to load.Enable 3rd party cookies or use another browser
In honor of today’s Beast, go look in the recycling bin—or inside a cash register.
Tell Me Something Interesting
The AI company Anthropic, which created the Claude AI and which prides itself on being moderately more ethical than OpenAI and the rest of the crowd, recently released some new research on AI “thought” processes. Granted, said research was conducted by the company on its own LLM, so it’s not exactly replicable or necessarily academically rigorous, and I (EJ Wittes) am not at all qualified to assess the quality of the work. That said, I’m sharing it anyway, because it’s really super interesting:
Keep reading with a 7-day free trial
Subscribe to Dog Shirt Daily to keep reading this post and get 7 days of free access to the full post archives.
 | A guest post by
|