🎉 Amazon’s Nova Ambition, T-Mobile’s AI Phone, Apple’s Siri Delay, OpenAI’s GPT-4.5, NYT On The Future of Work Post AGI
Amazon Reasoning, T-Mobile's AI Phone, Siri Falling Behind, GPT 4.5 Emotions, Ezra Klein Interview Must Read/Listen
Welcome to this week’s edition of AImpulse, a four point summary of the most significant advancements in the world of Artificial Intelligence followed by a cool new AI tool I’m trying out this week.
Here’s the pulse on this week’s top stories:
What’s Happening: Amazon is reportedly working on a sophisticated AI model called Nova—planned for a June launch—in what appears to be its boldest move yet to compete with OpenAI, Anthropic, and Google.
The Details:
The model is described as a “hybrid reasoning” system designed to provide quick responses as well as methodical, multi-step problem-solving from one unified model.
Cost-efficiency is key, with Amazon aiming to offer highly competitive prices without sacrificing top-level performance.
The company has set lofty goals to rank among the top five models, particularly in coding and math benchmarks.
This effort is part of Amazon’s AGI division under Rohit Prasad—signaling a new strategic direction, even with the company’s substantial $8B investment in Anthropic.
Why It Matters: Despite having a sizable stake in Anthropic, Amazon isn’t shying away from developing its own competing models—aiming for cutting-edge reasoning capabilities while offering lower costs than both competitors and partners. Combined with an AI-enhanced Alexa+ reportedly in the works, Amazon is positioning itself as a serious multi-front player in the AI landscape.
What’s Happening: T-Mobile’s parent company, Deutsche Telekom, has unveiled plans for an “AI Phone” created in partnership with Perplexity—one of the first major carrier-led attempts to produce a smartphone built for AI-centric experiences.
The Details:
The phone will feature Perplexity Assistant as its main interface, instantly available from the lock screen, removing the need to switch between apps.
Perplexity CEO Aravind Srinivas describes it as a move from an “answer machine to an action machine,” handling daily tasks with ease.
The device will also incorporate other AI tools, including Google Cloud AI for real-time translations, ElevenLabs for generating podcasts, and Picsart for creating avatars.
Scheduled for launch later this year, the phone is expected to cost under $1,000, with DT releasing an app version of its Magenta AI starting this summer.
Why It Matters: While tech giants have been gradually adding AI features to existing smartphones—with mixed results—this might be the first real move toward an AI-focused mobile experience that goes beyond app-based interfaces. It’s a significant win for Perplexity, which continues to expand across multiple corners of the AI market.
What’s Happening: Apple’s ambitious plan for a completely overhauled, AI-powered Siri has reportedly been delayed, according to insider Mark Gurman at Bloomberg. Many employees now believe the fully updated assistant won’t be ready until 2027.
The Details:
Siri is currently split between its legacy functions and newer AI capabilities, operating under separate systems.
Apple intended to merge these into one streamlined architecture, but that process has apparently fallen behind schedule.
Internal data shows limited adoption of current Apple Intelligence features, with many users finding them lacking compared to rival platforms.
Challenges within Apple’s AI division—like frequent talent departures and leadership changes—along with difficulties acquiring necessary AI chips have further complicated the rollout.
Why It Matters: Apple typically refines existing technology rather than rushing to be first, but the swift advances in voice AI have revealed a substantial gap in Siri’s capabilities. While 2027 may be Apple’s target, in AI terms that’s practically an eternity—especially with rivals like Alexa rapidly stepping up their game and potentially leaving Siri behind in the race.
What’s Happening: OpenAI has introduced GPT-4.5 (codenamed Orion), touted as its biggest model so far. Instead of focusing on reasoning, it relies on unsupervised learning to deepen its world knowledge and sharpen emotional intelligence.
The Details:
According to OpenAI, GPT-4.5 provides a more natural, conversational experience, with better grasp of human intent and greater emotional sensitivity.
Early users report fewer hallucinations and more accurate answers across a range of tasks, from professional projects to creative pursuits and day-to-day queries.
It doesn’t offer a breakthrough in math or science beyond previous versions, but it outperforms both o3-mini and o1 on OpenAI’s new SWE-Lancer benchmark for coding tasks.
At launch, only Pro users and developers with paid plans can use GPT-4.5; Plus and Team users will gain access the following week.
Pricing is notably high: $75/$150 per million input/output tokens. In comparison, GPT-4o costs just $2.50/$10 for the same amounts.
Why It Matters: While some may be disappointed by its lofty price and incremental improvements, GPT-4.5 seems geared more toward refining the model’s “feel” than making major leaps in capability. The high costs and limited upgrades suggest this might be OpenAI’s final large-scale model before shifting focus elsewhere—or at least before significantly advancing beyond standard unsupervised approaches.
AI Must Read / Listen: The best thing you’ll read or listen to this week in AI is a conversation between Ezra Klein and Ben Buchanan, former Whitehouse Special Advisor on AI. In the interview, Ben warns that advanced A.I. systems, potentially approaching A.G.I., will emerge in the near future, posing urgent challenges in labor, national security, and government preparedness. He and Ezra discuss how the U.S. and China are racing for A.I. dominance, revealing disagreements about regulation, safety, and the fate of American workers as these technologies reshape society.
Not enough time? Here’s your four point summary:
Rapidly advancing A.I. capabilities and possible A.G.I.: Buchanan warns that in the next few years, we could see “extraordinarily capable” A.I. systems—potentially approaching artificial general intelligence—arriving faster than most people expect, raising deep questions about labor market disruption, national security, and societal readiness.
Competing U.S. policy visions and export controls: The Biden administration pursued modest safety measures (like sharing test results with government), built an A.I. Safety Institute, and enacted chip export controls to slow China’s A.I. progress. Under Trump and Vice President Vance, there’s a more “accelerationist” stance on A.I. adoption and less interest in European-style regulations, even as both administrations agree the U.S. must stay ahead of China technologically.
National security and surveillance concerns: A major theme is how A.I. will transform intelligence gathering, cyber warfare, and overall state power. More advanced A.I. makes it easier to hack and analyze vast data, but could also fortify defense systems. There’s broad agreement that the U.S. must maintain leadership to shape how these tools are used—though that risks fueling an A.I. arms race with China.
Worker displacement and limited preparedness: While A.I. could boost productivity and innovation (e.g., in drug discovery), it could also replace cognitive jobs and upend industries rapidly. Buchanan and Klein note that policymakers have few concrete plans to help workers adapt, reflecting an uncomfortable gap between the technology’s speed of development and the government’s capacity to respond.