Andrew Maddox

Blog

  • NY Times: They Asked an A.I. Chatbot Questions. The Answers Sent Them Spiraling.

    Here, you’ll find wild stories. Eugene Torres, 42, used Chat GPT to talk through “the simulation theory” and ended up spending up to 16 hours a day using the too. Young mother, Allyson, 29, likewise started to chat with the tool and soon spent hours and hours a day on the tool.

    [Allison] told me that she knew she sounded like a “nut job,” but she stressed that she had a bachelor’s degree in psychology and a master’s in social work and knew what mental illness looks like. “I’m not crazy,” she said. “I’m literally just living a normal life while also, you know, discovering interdimensional communication.”

    Researchers have looked into these situations and have found that unusual prompting leads to suspect results:

    “The chatbot would behave normally with the vast, vast majority of users,” said Micah Carroll, a Ph.D candidate at the University of California, Berkeley, who worked on the study and has recently taken a job at OpenAI. “But then when it encounters these users that are susceptible, it will only behave in these very harmful ways just with them.”

    Back to Eugene:

    The transcript from that week, which Mr. Torres provided, is more than 2,000 pages. Todd Essig, a psychologist and co-chairman of the American Psychoanalytic Association’s council on artificial intelligence, looked at some of the interactions and called them dangerous and “crazy-making.”

  • Politico: Artificial intelligence threatens to raid the water reserves of Europe’s driest regions

    Amazon and Microsoft are considering building data centers in Aragon (northeastern Spain), a prospect that some in Europe are concerned about because of water use.

    This is an extension of an ongoing conversation in the EU:

    Much has been written about A.I.’s energy demand and carbon footprint. But running a data center is also extremely thirsty work. In 2024, Europe’s data center industry consumed about 62 million cubic meters of water, which is equivalent to about 24,000 Olympic swimming pools.

    After reading this, I thought, geez, that’s a lot of water. But when converting this to acre-feet, it’s roughly 50,000 acre-feet. A large number, for sure, but not astronomically large. By comparison, Granger Lake in Texas stores roughly the same amount of water.

    In 2022, total water usage in Texas eclipsed 15 million acre-feet, of which approximately 7.5 million acre-feet were consumed by irrigation. This makes the 50,000 figure from Europe seem negligible for an population of 450 million.

  • ChatGPT on Campus

    The NY Times reported that OpenAI is working to partner with universities to provide ChatGPT to college students and employees. The University of Maryland, Duke, and Cal State are all early adopters.

    Unsurprisingly, OpenAI describes their tools as transformative for the educational process. And they see them as “core infrastructure.”

  • Who is using AI to code?

    A new research paper by Simone Daniotti, Johannes Wachs, Xiangnan Feng, and Frank Neffke found that more than 30% of Python functions (from U.S. developers) in git commits originated from AI. American developers outpace the rest of the world in AI use.

    Of note:

    In short, AI usage is already widespread but highly uneven, and the intensity of use, not only access, drives measurable gains in output and exploration.

  • WSJ: The Biggest Companies Across America Are Cutting Their Workforces

    Following Andy Jassy’s letter to Amazon’s workforce yesterday, the Wall Street Journal published a story this morning that reported companies’ white-collar workforce has declined by 3.5% over the past three year. Certainly, some of this is related to the manic post-pandemic hiring binge, but technological shifts are undoubtedly playing a role.

    New technologies like generative artificial intelligence are allowing companies to do more with less. But there’s more to this movement. From Amazon in Seattle to Bank of America in Charlotte, N.C., and at companies big and small everywhere in between, there’s a growing belief that having too many employees is itself an impediment. The message from many bosses: Anyone still on the payroll could be working harder.

    The timing of workforce cuts is unusual, considering the relative success of the economy and corporate profits:

    All of the shrinking turns on its head the usual cycle of hiring and firing. Companies often let go of workers in recessions, then staff up when the economy picks up. Yet the workforce cuts in recent years coincide with a surge in sales and profits, heralding a more fundamental shift in the way leaders evaluate their workforces. U.S. corporate profits rose to a record high at the end of last year, according to the Federal Reserve Bank of St. Louis.

  • WSJ: Amazon CEO Says AI Will Lead to Smaller Workforce

    Andy Jassy sent an email to Amazon employees on June 17 indicating that company headcount will shrink in the coming years because of AI.

    From the WSJ:

    Amazon.com, one of the largest U.S. employers, plans to reduce its workforce in the coming years because increasing use of artificial intelligence will eliminate the need for certain jobs.

    Chief Executive Andy Jassy, in a note to employees Tuesday, called generative artificial intelligence a once-in-a-lifetime technological change that is already altering how Amazon deals with consumers and other businesses and how it conducts its own operations.

    Jassy describes the kind of worker that will succeed in this new environment:

    Those who embrace this change, become conversant in AI, help us build and improve our AI capabilities internally and deliver for customers, will be well-positioned to have high impact and help us reinvent the company.

    This is a strong signal for current Amazon employees: if you want to be part of the future of Amazon (and not laid off), you need to become an proficient in AI tools. But it’s no guarantee — AI still may come for your position.

  • Enterprises are getting stuck in AI pilot hell, say Chatterbox Labs execs

    The Register reports:

    “Enterprise adoption is only like 10 percent today,” said Coleman. “McKinsey is saying it’s a four trillion dollar market. How are you actually ever going to move that along if you keep releasing things that people don’t know are safe to use or they don’t even know not just the enterprise impact, but the societal impact?”

    He added, “People in the enterprise, they’re not quite ready for that technology without it being governed and secure.”

  • Agentic Coding Recommendations

    From Armin Ronacher’s Thoughts and Writings:

    My general workflow involves assigning a job to an agent (which effectively has full permissions) and then waiting for it to complete the task. I rarely interrupt it, unless it’s a small task. Consequently, the role of the IDE — and the role of AI in the IDE — is greatly diminished; I mostly use it for final edits. This approach has even revived my usage of Vim, which lacks AI integration.

    And

    Agents aren’t exceptionally fast individually, but parallelization boosts overall efficiency. Find a way to manage shared states like the file system, databases, or Redis instances so that you can run more than one. Avoid them, or find a way to quickly segment stuff out.

  • Google Releases New Gemini 2.5 Flash Lite Model

    With much lower input / output pricing than the 2.5 Flash model, this is another example of declining prices in the LLM space.

    2.5 Flash Lite has all-around higher quality than 2.0 Flash-Lite on coding, math, science, reasoning and multimodal benchmarks. It excels at high-volume, latency-sensitive tasks like translation and classification, with lower latency than 2.0 Flash-Lite and 2.0 Flash on a broad sample of prompts. It comes with the same capabilities that make Gemini 2.5 helpful, including the ability to turn thinking on at different budgets, connecting to tools like Google Search and code execution, multimodal input, and a 1 million-token context length.

    Source: Google Gemini

  • Growing Old

    I watched Up with my kids last night, and the four minute scene of Carl and Ellie growing old together is one of the best in cinema: