🔋 Google Goes Nuclear

Google's innovative nuclear energy partnership with Kairos Power aims to address the increasing power demands of AI technologies.

Author

Jack Lajoie
October 15, 2024 • read time ~ 8 minutes

The Daily Current ⚡️

Welcome to the creatives, builders, pioneers, and thought leaders ever driving further into the liminal space.

Google steps into the nuclear realm to fuel its AI ambitions, The New York Times looks to draw a line in the sand against AI startups exploiting their content, and researchers revealed vulnerabilities in AI image processing. On a lighter note, Casio's AI pet, Moflin, promises cute companionship based on dynamic AI capabilities. Finally, the U.S. is reportedly mulling over new more broad export restrictions on AI chips.

 🔌 Plug Into These Headlines:

  1. 🔋 Google Goes Nuclear: Announces Partnership with Kairos for Nuclear Solution to AI Power Needs

  2. 📰 New York Times Issues Cease and Desist to Perplexity

  3. 🔍 AI Image Recognition Flaw Uncovered by Researchers

  4. 🐾 Meet Moflin: Casio's AI-Powered Pet Companion

  5. 🛑 US Considers Limiting AI Chip Exports to Certain Nations

Google has signed a groundbreaking agreement with Kairos Power to purchase nuclear energy from multiple small modular reactors (SMRs), aiming to bring the first SMR online by 2030 and deploy additional reactors through 2035. This deal will enable up to 500 MW of new 24/7 carbon-free power to U.S. electricity grids, supporting the increasing energy demands of AI technologies while advancing clean energy solutions.

  • The deal comes weeks after a similar Microsoft deal was set to reboot part of Three Mile Island

  • The agreement aims to support the increasing energy demands of AI technologies that are driving major scientific advances, improving services, and boosting economic growth.

  • By procuring electricity from multiple reactors, Google aims to accelerate deployments and lower costs, potentially making clean energy more accessible for future AI infrastructure expansion.

  • Kairos Power's technology uses a molten-salt cooling system and ceramic pebble-type fuel, operating at low pressure for a simpler, more affordable reactor design.

While this agreement marks a significant step for clean energy adoption, it may also raise questions about the long-term viability and safety of nuclear power. There’s probably a reason we haven’t adopted it widely yet…

Check out this report from the Union of Concerned Scientists back in 2013 🔗

This move reflects the escalating tension between traditional media and AI startups over content usage and intellectual property rights. Similar steps taken by other major publishers, perhaps indicating a coordinated effort to protect their content and business models.

  • The Times' letter demands Perplexity stop using its content for training AI models and generating responses.

  • Perplexity's CEO, Aravind Srinivas, claims the company doesn't scrape or copy content, instead using publicly available information

  • AI companies genreally contend that their use of publicly available information falls under fair use doctrine.

  • Publishers are exploring various strategies, including paywalls and legal action, to protect their content from AI companies.

The New York Times has apparently decided on taking a very active role in this battle around AI search. Whichever way it tilts, the conflict underscores the need for updated regulations and legal frameworks to address AI's impact on intellectual property (which goes far beyond just protecting newspapers).

A new study reveals a vulnerability in AI image recognition systems due to their exclusion of the alpha channel (which controls image transparency). Researchers developed “AlphaDog,” an attack method that manipulates transparency in images, allowing hackers to distort visuals like road signs or medical scans in ways undetectable by AI.

  • AlphaDog manipulates image transparency, misleading AI models in fields like road safety and telehealth.

  • Most AI systems omit the alpha channel, which is crucial for accurate image transparency.

  • Researchers are working with tech companies to integrate alpha channel processing and secure AI.

AI is created by humans, and the people who wrote the code focused on RGB but left the alpha channel out. In other words, they wrote code for AI models to read image files without the alpha channel. That's the vulnerability. The exclusion of the alpha channel in these platforms leads to data poisoning.

The research exposes the potential for malicious actors to exploit these vulnerabilities in real-world scenarios, such as autonomous vehicles or security systems.

The device utilizes voice and gesture recognition to develop a bond with its owner, evolving its behavior over time. Moflin's AI allows for 4 million different combinations of movements and sounds, creating a personalized interaction experience.

  • The device recognizes voice and gestures, adapting its responses accordingly

  • Moflin can express "negative emotions" if ignored or startled

  • Priced at approximately $400, it includes a dedicated app and optional subscription services

  • Battery life is limited to 5 hours, with a 3.5-hour recharge time

Well, we knew this was coming at some point, right? While Moflin represents an intriguing advancement in the AI pet niche, its limited battery life and predetermined emotional responses highlight the challenges to creating truly lifelike artificial companions.

Beyond China, the proposed measures could affect some U.S. allies, raising complex diplomatic and economic considerations. This broader approach reflects growing concerns about the global diffusion of advanced AI technologies and their potential military applications.

  • The expansion looks to focus on Middle East and Southeast Asian countries with ties to China or strategic importance.

  • Potential impacts on U.S. allies and chip manufacturing countries like South Korea and Taiwan may strain diplomatic relations.

  • The broader scope aims to prevent "transshipment" of restricted technologies through third countries.

While aiming to safeguard national security, the broader application of export controls might inadvertently create opportunities for countries not targeted by these restrictions to become key players in the global AI chip market. Look for some traditionally quiet nations on the periphery of the tech industry to begin making moves if these restrictions go into effect.