The week in AI: Google's 2024 environmental report highlights the looming AI energy crisis

In partnership with

Welcome to The Dispatch! We are the newsletter that keeps you informed about AI. Each Thursday, we aggregate the major developments in artificial intelligence - we pass along the news, useful resources, tools and services; we highlight the top research in the field as well as exciting developments in open source. Even if you aren’t a machine learning engineer, we’ll keep you in touch with the most important developments in AI.

We hope you had a great 4th of July! We’re a day late due to the holiday.

NEWS & OPINION

Google reports a 13% increase in emissions for 2023, citing growing AI power demands - ‘net-zero’ climate goal set for 2030 in jeopardy

-------------------------

Back in 2021, Google set an ambitious ‘net-zero by 2030’ goal, aiming to balance out the company’s greenhouse gas emissions with carbon offsets. But the unforeseen explosion in AI development and usage is creating major challenges in meeting this goal, and according to their 2024 environmental report, overall emissions rose by 13% in 2023. Unsurprisingly, other tech giants with clean energy commitments are likewise grappling with AI’s voracious energy appetite, forcing a reevaluation of sustainability timelines.

Beyond putting important climate goals at risk, these developments point towards a looming power crisis in America. Some considerations on these issues:

• According to the International Energy Agency, global data center and AI electricity demand could double by 2026. In our view, this estimate might actually be conservative. For example, no one can yet foresee how iterative, agentic workflows (where AI agents tackle complex tasks and achieve a single output through multiple iterations and refinements) will affect global AI use rates in the coming years - but these workflows theoretically require much, much more compute and energy than current AI use trends.

• Every big tech company is making massive data center/AI infrastructure investments right now. Amazon (over $100B in the next decade and 216 data center buildings in the next few years), Meta ($37B in 2024 alone), Microsoft ($50B between July 2023 and June 2024 alone), Apple (exact infrastructure investment uncertain, but $450B total investment in data centers, manufacturing and offices in the US over the next 5 years). The energy requirements for this level of infrastructure will be immense.

• New data centers get built where electricity is cheapest - they not only delay the closures of plants that burn fossil fuels, but prompt new ones to be built.

• Several states where data centers are being built are already in dire straits and have major energy issues. Data centers are particularly putting the squeeze on Texas, already dealing with the threat of summer blackouts.

The intersection of technological advancement, energy management, and environmental sustainability presents both unprecedented challenges and unique opportunities. The decisions made by tech leaders, policymakers, and consumers will have far-reaching consequences for the energy sector, our digital futures and the health of our planet.

The New York Times: AI begins ushering in an age of killer robots

-------------------------

In a chilling yet fascinating piece, The New York Times brings us to the frontlines of AI warfare development in Ukraine. The article paints a vivid picture of a country transformed into a "Silicon Valley for autonomous drones and other weaponry." We're introduced to companies like Vyriy, where a young CEO demonstrates a drone that autonomously tracks him on his motorcycle - a scene that feels lifted from a sci-fi movie, yet is very much our present reality.

The reporters delve deep into Ukraine's tech ecosystem, revealing how the pressure of war and an influx of investment have accelerated the development of AI-powered weapons. From fruit-sorting algorithms repurposed for military targeting, to a video-game controller automating a machine gun turret (with an auto-aim feature to track and hit moving targets currently in development), the ingenuity on display is as impressive as it is unsettling. The article doesn't shy away from the ethical implications, quoting AI experts who warn of the potential for these technologies to become "weapons of mass destruction that are cheap, scalable and easily available."

Perhaps most striking is the sense of urgency conveyed throughout the piece. Ukraine's digital transformation minister, Mykhailo Fedorov, puts it bluntly: "We need maximum automation. These technologies are fundamental to our victory." This sentiment, juxtaposed against warnings from UN officials and human rights groups, encapsulates the core dilemma at the heart of this technology arms race. The NYT article serves as a wake-up call to anyone not paying close attention: the future of warfare is not just coming - it's already here.

Gradually, then suddenly: upon the threshold

-------------------------

On his popular “One Useful Thing” Substack, acclaimed author and Wharton professor Ethan Mollick introduces the concept of "capability thresholds" in AI development. Unlike the common perception of steady technological improvement, Mollick argues that AI progress is characterized by sudden, transformative leaps occurring when specific capability thresholds are crossed. This phenomenon explains why AI's impact often seems to happen "gradually, then suddenly," - echoing Hemingway's famous quote about bankruptcy.

Mollick illustrates this concept through various examples - from AI's rapidly improving ability to generate images and videos to its growing capacity for complex data analysis. He leverages Claude quite a bit to make his points for the article. Given how comprehensively we covered Anthropic’s excellent flagship model last week, we’re not surprised. But even as impressive demos of AI capabilities increase, the true measure of progress lies in crossing thresholds of practical usefulness. This distinction is crucial for understanding the gaps between AI's potential and its real-world impact across different industries and applications.

Mollick also suggests that individuals and organizations should create and maintain an "impossibility list" - tasks that AI cannot yet perform but is approaching the capability to do so. This approach provides a structured way to anticipate and prepare for significant AI breakthroughs.

MORE IN AI THIS WEEK

• This is Big Tech’s playbook for swallowing the AI industry - the reverse acquihire

• Generative AI and election disinformation: much ado about nothing? 

• Nvidia set to face French antitrust charges, sources say, marking the first regulatory action against the AI chip giant

• GXO signs industry-first,multi-year Robots-as-a-Service (RaaS) agreement with Agility Robotics

• Micron gears up for its potential data center/AI memory boom

• Ken Griffin, CEO of multinational hedge fund Citadel, says he’s not convinced that AI will replace human jobs in the near future

• AI Standards Deep-Dive: decoding different AI standards and the EU’s approach

• Microsoft’s AI partnership with UAE-based G42 is raising national security concerns

• The underground network sneaking Nvidia chips into China

• Forget Sora, Runway is the AI video maker coming to blow your mind

• Meta plans to bring generative AI to metaverse games

• Robinhood acquires AI investment platform Pluto, plans to integrate advanced data analytics and personalized strategies into its trading app

FREE AI & ChatGPT Masterclass to automate 50% of your workflow

More than 300 Million people use AI across the globe, but just the top 1% know the right ones for the right use-cases.

Join this free masterclass on AI tools that will teach you the 25 most useful AI tools on the internet – that too for $0 (they have 100 free seats only!)

Get it now for absolutely free! (for first 100 users only) 🎁

This masterclass will teach you how to:

• Build business strategies & solve problems like a pro

• Write content for emails, socials & more in minutes

• Build AI assistants & custom bots in minutes

• Research 10x faster, do more in less time & make your life easier

You’ll wish you knew about this FREE AI masterclass sooner 😉

Register & save your seat now! (valid for next 24 hours only!)

TECHNICAL, RESEARCH & OPEN SOURCE

• OpenAI’s CriticGPT: finding GPT-4’s mistakes with… GPT-4

• Microsoft’s GraphRAG: New tool for complex data discovery now on GitHub

• Google opens Gemini 1.5 Pro 2M token context window for all devs

• Understanding Deep Learning - free eBook/PDF

• RouteLLM from LMSys: an open source framework for cost-effective LLM routing

• Anthropic launches a new initiative for developing third-party model evaluations

• From Cloudflare: declare your AIndependence - block AI bots, scrapers and crawlers with a single click

• Google unveils Gemma 2, a lightweight, state-of-the art open model built from the same research and technology used to create the Gemini models

• Shanghai AI Lab introduces InternLM 2.5-7B, a model with a 1M context window and the ability to use tools

• My fine-tuned models beat OpenAI’s GPT-4

That’s all for this week! We’ll see you next Thursday.