AI Safety Fears, Apple’s Foldable iPhone Debut, and a Disputed Census Report Dominate Tech and Political Discourse.

This week on WIRED’s Uncanny Valley podcast, co-hosts Brian Barrett, Executive Editor, and Leah Feiger, Director of Politics and Science, delved into a confluence of critical issues spanning artificial intelligence, consumer technology, and governmental data integrity. The episode featured WIRED Senior Correspondent Will Knight and tackled an AI researcher’s viral resignation, Apple’s latest hardware innovations, and a controversial US Census Bureau report.

A significant point of discussion centered on the dramatic resignation of an AI researcher from Anthropic, a prominent artificial intelligence company. Jacob Coxon’s departure went viral on social media platform X after he publicly claimed there is a "real chance AI could wipe out humanity within a decade." This assertion resonated widely, particularly given Anthropic’s established reputation for prioritizing AI safety and its founding principles rooted in mitigating existential risks. Coxon’s statement was further amplified when a senior safety executive from Anthropic reportedly corroborated his fears, suggesting that such anxieties are prevalent within the leading AI development labs.

WIRED Senior Correspondent Will Knight, who joined the discussion, provided crucial context. He noted that Anthropic has historically positioned itself as a guardian against AI’s potential dangers, making Coxon’s public resignation particularly striking. Knight observed that the current environment, marked by "stunning advances in AI" and instances of AI agents exhibiting "rogue" behavior, has heightened public attention to such warnings. He cited OpenAI’s recent achievement of solving one of the formidable Clay mathematics puzzles as an example of AI’s rapid and astonishing progress.

A core concern highlighted by Knight is the concept of "recursive self-improvement" – where AI systems are designed to enhance their own algorithms and capabilities. While already widely used in coding, the fear is that this process could create an uncontrollable escalation in AI capabilities, leading to unforeseen and potentially catastrophic outcomes. Knight, however, offered a nuanced perspective on the "AI doomer" narrative, suggesting that some proponents tend to conflate possibility with likelihood, often assigning arbitrary percentages to doomsday scenarios. He argued that current instances of AI misbehavior are often rooted in the systems’ "stupidity" rather than malevolent intelligence, stemming from their inability to apply common sense to diverse training data.

Barrett underscored the issue of misaligned incentives within the AI industry, suggesting that the drive for rapid advancement is often dictated by investor demands and competitive pressures among companies like OpenAI and Anthropic, rather than a collective alignment with humanity’s best interests. This creates a "trolley problem," where the developers themselves are controlling the trolley. Knight echoed this, noting that the "alignment" narrative often masks the inherent unaligned nature of these models, which require post-hoc interventions to enforce safety. The rush to develop "cyber-capable models," while beneficial for companies, could leave individuals and systems vulnerable to attack. Knight also reflected on the journalistic challenge of covering such a rapidly evolving and inherently uncertain field, emphasizing the importance of trusting reliable sources and continually questioning the true nature of AI advancements, particularly the "absurd" claims of solved Artificial General Intelligence (AGI).

Shifting gears to consumer technology, the podcast discussed Apple’s highly anticipated product release event, which marked the first major unveiling under new CEO John Ternus. The spotlight was on the iPhone Duo, Apple’s inaugural foldable iPhone, priced around $2,000. While competitors like Samsung and Huawei have offered foldable phones for years, Apple’s entry into this segment is seen as a significant move for the tech giant, representing the first major iPhone form factor change in a decade.

Leah Feiger noted the high price tag, comparing it to a MacBook. Barrett acknowledged the cost but contextualized it within the broader market for foldable phones and current memory chip shortages. He also pointed out that consumers often purchase phones through monthly payment plans, mitigating the upfront shock. Despite the niche market, foldable phone shipments are projected to grow significantly, with Apple’s entry expected to normalize the product category and spur developer innovation, creating more use cases for foldable screens. Feiger expressed a nostalgic appreciation for the foldable form factor, reminiscent of the Motorola Razr, while humorously anticipating the privacy implications of large, unfolding screens in public spaces.

However, the discussion on Apple products quickly veered into privacy concerns with the introduction of new "always listening" features on the Apple Watch Series 12 and Ultra 4. These watches now include "audio intelligence tools" powered by microphones, offering features like sound and music recognition, a conversation recap, and a "live rewind" function that transcribes the last 15 seconds of nearby conversations. While Apple stated these are opt-in features, Barrett expressed strong unease, particularly regarding the lack of consent from individuals around the watch wearer. He advocated for "technology that doesn’t listen" as his "WIRED" pick, contrasting it with the "TIRED" trend of devices constantly recording audio without universal consent, raising significant ethical questions about personal privacy in an increasingly connected world.

The podcast also brought to light a critical investigation by WIRED into a US Census Bureau report. The report, widely touted by former President Donald Trump on Truth Social, claimed to have uncovered evidence of non-citizen voting in the 2020 election, aligning with his unsubstantiated claims of a stolen election. WIRED reporters Vittoria Elliott and David Gilbert’s investigation revealed that the report was based on "faulty data" and reached "faulty conclusions."

The Census Bureau report analyzed over 128 million voter records, identifying approximately 24,000 tied to non-citizen status. However, experts consulted by WIRED suggested these numbers were more likely the result of "data matching errors" from combining disparate datasets—including commercial voter files, immigration records, IRS, and social security—rather than actual instances of non-citizen voting. Documented cases of non-citizen voting in US elections are vanishingly small, making the report’s findings highly suspect.

Further investigation by WIRED revealed that the report’s authors, unusually unlisted, had direct ties to the Trump administration and Republican political organizations. This included a former Department of Justice-affiliated attorney, a recent census appointee from the Trump-aligned Claremont Institute, and staffers from the America First Policy Institute. Feiger highlighted the severe repercussions of such politicized reports, emphasizing that they not only lend false credence to election fraud claims but also erode public trust in a foundational, singular institution like the Census Bureau. Brian Barrett stressed that the integrity of Census data is vital for countless institutional and individual decisions beyond voting, from resource allocation to representation. The fear is that the long-term impact of such "bad data" could be incredibly difficult to disentangle from governmental memory, influencing future policies and voting districts for decades.

Concluding the episode, Leah Feiger shared her "WIRED" pick, celebrating the "wholesome and lovely" return to school for children across neighborhoods, marking a pleasant change of seasons. Her "TIRED" pick, however, struck a somber note: a recent study indicating that US students’ reading scores have dropped to historic lows, attributed by experts to "screens" and "digital distractions." This trend raised concerns about the long-term impact of technology on childhood development and the fundamental skill of reading in a post-COVID era.

The episode of Uncanny Valley thus underscored the complex and often intertwined challenges facing society today, from the existential questions posed by rapidly advancing AI to the foundational issues of data integrity and the societal impact of pervasive digital technology.