Apple looks to Anthropic or OpenAI to save Siri, Google's AI hit with antitrust complaint
Get bigger weekly updates! Free subscribers receive the top stories each week, while Paid subscribers will get a few extra stories. All support for Handy AI directly helps us maintain the newsletter and keep the information flowing.
last week’s top stories
🍏 Apple mulls Anthropic/OpenAI for next Siri. Apple has privately asked Anthropic and OpenAI to adapt their latest language models (Claude, ChatGPT) to run on Apple’s cloud for a future Siri upgrade. This could sideline Apple’s own on-device AI. The talks are at an exploratory stage (no final decision announced), but they signal Apple might outsource some AI tech to catch up in voice assistant intelligence. Read more
💬 Altman slams Meta’s AI talent-poaching spree. In a leaked July 1 memo to OpenAI staff, CEO Sam Altman condemned Meta’s recent recruiting blitz (paying huge bonuses to hire OpenAI researchers) as risky. He warned such “mercenary” tactics could cause “very deep cultural problems” and urged employees to stay mission-driven. Altman even called OpenAI staff “missionaries” fighting for a cause, vowing to raise everyone’s compensation if needed. Read more
🚀 Musk’s xAI raises $10B for Grok AI and data centers. Elon Musk’s AI startup xAI announced a massive $10 billion fundraising round split evenly between debt and equity. xAI plans to use the funds to build out its “Colossus” supercomputer fleet and advance Grok (the AI model powering Musk’s X chat). Read more
👶 AI helps infertile couple conceive after 18 years. A couple who had faced 18 years of failed IVF due to almost no sperm finally conceived using Columbia University’s new STAR (Sperm Tracking and Recovery) system. The AI-driven system captured ~8 million microscope images per semen sample and used machine learning to spot even the smallest number of sperm cells. Read more
🧠 AI model simulates human decision-making across tasks. Researchers at Helmholtz Munich fine-tuned a large language model (based on Llama) on data from ~160 psychology studies (60K participants, 10M decisions) to create a model called Centaur. Centaur reportedly predicts human choices across a wide range of tasks (games, memory tests, etc.), and even generalizes to scenarios it never saw during training. Read more
⚡ Google signs fusion power deal with startup. Google agreed to buy 200 MW of electricity from Commonwealth Fusion Systems’ upcoming ARC fusion plant (slated for early 2030s). In parallel, Google is investing in CFS’s new funding round, reinforcing its bet on fusion energy. This is Google’s first commercial fusion deal (after Microsoft’s partnership with Helion) and shows tech giants seeking clean power sources for future data centers. Read more
⚖️ EU publishers challenge Google’s AI Overviews. An alliance of European media publishers filed an antitrust complaint alleging Google’s new “AI Overviews” (search-result summaries) unfairly exploit their content. The complaint argues that Google’s LLM-generated snippets reuse copyrighted material and divert readers from original sites, and that publishers have no way to opt out without being penalized in search. Read more
💊 Microsoft’s AI moves toward “medical superintelligence”. Microsoft unveiled an AI diagnostic system that outperformed human doctors on complex cases, calling it a “path to medical superintelligence”. The system (dubbed MAI Diagnostic Orchestrator) sequences multiple AI agents (including OpenAI’s models) to mimic a panel of physicians: in tests, it solved ~80% of tough cases while trained doctors only solved ~20%. Read more
🎓 Tech firms pledge to boost AI education. In partnership with the White House, 67 organizations (including Google, IBM, Meta, Microsoft, NVIDIA) pledged to promote AI education for K–12 students. Under the “Pledge to America’s Youth,” signees will provide resources to foster AI literacy and train educators in AI tools. This initiative coincided with Congress rejecting a proposed 10-year moratorium on state AI laws, emphasizing a push to integrate AI responsibly in classrooms and prepare students for an AI-powered future. Read more
🧪 AI Research of the Week
Delving into LLM-assisted writing in biomedical publications through excess vocabulary
From University of Tübingen & Northwestern University
Jake’s Take: In this study, the authors mined 15.1 million PubMed abstracts from 2010–2024 and built a frequency model for every word, then compared each 2024 value with a baseline extrapolated from 2021–2022. A surge of 379 style-heavy verbs and adjectives (“showcasing,” “pivotal,” “intricate”) created an “excess vocabulary” signal. 66% of the flagged terms were verbs.
When rare and common style markers were combined, the gap implied that at least 13.5% of 2024 biomedical abstracts passed through an LLM, with pockets that rise to 40% in certain fields, countries, and journals (peaking at 0.41 for Chinese deep-computation papers).
It’s likely we’ll see folks quoting this study often, especially in the policy space, when discussing how AI is effecting our human ability to effectively run studies.
and then, even more news…
💼 OpenAI launches $10M custom AI consulting for enterprises. OpenAI is now offering bespoke AI integration projects starting at $10 million per engagement. Dubbed a new consulting arm, the service embeds “forward-deployed engineers” inside client organizations (e.g. U.S. Defense Dept., Grab) to integrate GPT-4o models with proprietary systems. Read more
🔬 Meta forms new “Superintelligence” AI lab. Meta (Facebook) formalized the adjustments to its R&D team into “Meta Superintelligence Labs,” led by ex-Scale AI CEO Alexandr Wang (as chief AI officer) and Nat Friedman (ex-GitHub). The new division’s mission is fast-tracking work toward artificial general intelligence (AGI) and other breakthrough AI products. Read more
🕸️ Cloudflare launches pay-per-crawl AI marketplace. Cloudflare announced “Pay per Crawl,” a private-beta marketplace that lets publishers set micropayments for AI bots scraping their sites. Website owners can decide which crawlers (e.g. ChatGPT, Anthropic bots) to allow, block, or charge for each “crawl,” establishing a direct fee for AI content access. Read more
🚀 Nvidia nears world’s most valuable company. Nvidia’s stock jumped in early July, briefly pushing its market cap to $3.92 trillion, just above Apple’s all-time high. The surge, driven by massive demand for Nvidia’s AI chips, made the company worth more than the entire UK stock market combined. Read more