"Assuming" the types of produce inputs are disclosed, are fresh, no farm bugs, clean, free from pesticides, and the tools are thoroughly cleaned, and free from bugs and rodents. Yes I would eat. But that's relying on many assumptions. If the quality guarantees are not there, don't eat it.

Healthcare: USA ranks 10th in 2024 for 1) Lowest life expectancy from preventable disease. 2) Barriers to affordable care 38% of adults skip recommended care due to cost. 3) Largest disparities between high and low-income residents able to afford healthcare. www.commonwealthfund.org/publications...

Health plan disenrollment following MOUD initiation "was associated with a 51% and 56% increase in the hazards of all-cause mortality and overdose mortality" Relevant finding as states work to operationalize Medicaid work requirement exemptions for enrollees with substance use disorder.

AI Agents are devaluing your brand by mispronouncing them. When using AI tools in customer service or to generate training tools for internal HR or external customers (i.e., training vendors on services or software). Make sure it is pronouncing your products/services correctly.

Now that's a headline: "Former surgeon general: The Senate must not approve someone who can’t practice medicine as the nation’s top doctor"

In AI, "benchmark saturation" refers to when an evaluation test (benchmark) used to evaluate which model is outperforming other models no longer differentiates between the top-performing LLM (Large Language Models) because their scores have hit a functional ceiling; where they perform the same

AI Agents can now "rent humans" (as a proxy) to do tasks where the AI Agent cannot be physically present. AI Agents can pay for goods and services but it cannot sign legally binding contracts. AI cannot assess discriminatory hiring practices or be personally liable for errors.

Employee compensation are being tied to their usage of AI tools. Meta: AI is a "core expectation". Shoosmiths (UK): usage of MS Copilot. Accenture: regular usage AI platform. Brex: AI projects, & workflow redesign. Sanofi: experimenting projects with AI. IBM: ties AI innovation to contests.

Finished reading "Valley of Genius: The Uncensored History of Silicon Valley". This is a compilation of the actual words said by the founder of startups of companies like Microsoft, Apple, Atari, Google, Napster, Facebook, dot com market bust since the 70's, 80's, 90's,2000's.

"The argument runs deeper than most discussions--even many academic ones--about the myths that continue to be reproduced through our stories about AI..." Congrats to "Artificial Humanities" on being named a 2025 Artificiality Book Award winner!

Artificiality Book Awards 2024 2024 was a banger year for books in our space and Helen has selected ten books for the Artificiality Book Awards 2024: Without further ado, the winners are...

In Tron: Ares, the AI escapes from his virtual-world through pseudo-advanced 3D printing. This type of storyline is a reverse "Isekai Ten'i" (Japanese) a fantasy character leaves their fantasy-world and enters into our own reality. Could AI one day will itself into existence?

MIT (2024) researchers recently proposed the Platonic Representation Hypothesis (PRH) to explain how AI Models LLMs are all starting to converge (behaving in similar ways) as these continue to scale (get bigger) even when they were trained with different corpus (body of data).

Also very cool: a study from 2024 found that in Scottish women fully vaccinated with the HPV vaccine in 2008, *no* cases of cervical cancer have been found. Not a single one. 100% success rate. Co-author: "It is [now] possible to make cervical cancer a rare disease.” www.bbc.co.uk/news/article...

DIT (Dead Internet Theory) a 2016 conspiracy theory becomes real. Asserts humans no longer drive the internet. Since 2024, cybersecurity reports estimate over 51% of content was bot-generated eroding user trust leading to a "crisis of perception" as users interact less with real people.

It’s great that @economist.com is talking about our paper in their piece on “slop” as word of the year — not so great that they don’t cite/link to it… www.economist.com/culture/2025...

Do not cite an academic paper unless you’ve read it

Örebro University, Sweden, released an EEG-AI model for detecting dementia 7 years before the symptoms appeared (in some instances 10-18 years before formal diagnosis) by analyzing disruptions in our alpha, beta, and gamma sleeping patterns. news-medical.net/news/2025112...

AI SLOP (Standard Low Quality Production) refers to low quality, high volume content. It's repetitive, absurd, has hallucinations, misquotes, fakes, etc. Intended to go viral and generate ad revenue while using emotional or shocking visuals and cheap to create.  youtube.com/watch?v=uqcoTwPUSn4

In the Short-Term AI help increase employee productivity but in the Long-Term reliance on AI can undermine the psychological experience of the human workforce, new study finds. www.nature.com/articles/s41... pmc.ncbi.nih.gov/articles/PMC...

Absence Of Reasonable Risk in AI models implies an "acceptable" minimal risk when weighed against the benefits where the potential harm caused is not greater than what is encountered in our daily life and where appropriate actions were taken to minimize those possible dangers.

Using AI without validating outputs (i.e., accuracy, references, meaning, etc.) may result in Abdication Of Responsibility which may result in harm to reputation, no future business, revoked license, lawsuits, dropped from risk insurance coverage for malpractice, and so forth.

120 AI products. 🤓 Eventually AI products consolidate into a few good products or their life-cycle ends replaced by new options. Only time will tell which products will survive the market. 😵‍💫 (Tap on image to see the top and bottom edge)

We can't really say this enough... > Anastasia Berg [at UCL Irvine] said that new research — and what she's hearing directly from colleagues across various industries — shows that employees who heavily rely on AI are losing core skills at a startling rate. www.businessinsider.com/ai-tools-are...

Kids may underestimate the dangers that AI-powered apps can pose if misused at school. Riana Pfefferkorn, policy fellow at Stanford University HAI (Human-Centered AI) will discuss this topic. You can register to attend virtually. hai.stanford.edu/people/riana...

Plurilingual AI. It's very common for fluent bilinguals (or plurilinguals) to flip languages mid conversations without missing a heart beat. This includes spoken spoken languages or Deaf sign languages too! Spanglish, Chinglish, Punjinglish, Germanglish, nerdglish, and now I'm making names up 🤪

Imagine Calvin having access to ChatGPT.

Since most of the data - if randomly collected - falls naturally on a statistical distribution curve. So, yes, avoid dichotomizing your data variables (avoid forcing it into two only categories) unless the data is already naturally dichotomized (i.e., alive or dead, etc.)

"Extraordinary claims require extraordinary evidence", known as the "Sagan Standard" is an aphorism popularized by astronomist Carl Sagan in his book Broca's Brain (1979) and TV show Cosmos (80's) with roots in David Hume's philosophy essay "Of Miracles" (1748) and later by Pierre-Simon Laplace.

What is "mojibake"? It's an unintentional mismatch in how one software encoding accidentally misreads the code characters from another software. It's an accidental side effect of data software not being able to be properly handled due to lacking standardization or configuration.

Your job is safe! For now. Despite hype, AI cannot work alone. New measuring benchmark highlights how valuable human domain knowledge and project management skills are in freelance projects. Several AI platforms failed 97% at completing full freelance projects. arxiv.org/pdf/2510.26787

Data pioneer: Dutch mathematician Tycho Brahe (Tee-Koh Brā-hē) rigorously collected astronomical data. On Oct 13, 1601, Brahe attended a royal banquet in Prague his bladder burst for not going to the restroom; wanting to be polite. He died 11 days later in extreme pain.

Today visited Klinker Brick Winery in Lodi, CA. I got the $15 tasting (6 reds x 2 oz. each = 12 oz). Some are better with a meal. The 2021 Dolcetto was particularly smooth on its own, with no aftertaste. I did not buy any bottles, but it was fun to do. www.klinkerbrickwinery.com/About-Us/Our...

Enlly Blue is an AI singer with a 1950's blues style. It's pretty good. The real person behind the production is Thong Viet Thong (remains private). He used AI as a tool to develop the lyrics and the musical arrangement. Check it out. www.youtube.com/watch?v=7a_c...

AI bots running "prompt engineering" simulations locally on itself with a newly upload document can learn to anticipate and provide faster "context-aware" answers with less memory and cost because it doesn't need to rescan the document again thanks to "cartridges" hai.stanford.edu/news/offline...

It would appear that an entire industry of evaluating LLMs is springing up. Here is another company that has put LLMs to the copyright test and found them to be problematic www.cnbc.com/2024/03/06/g...

Understanding Deep Learning: Free MIT Press EBook for Instructors and Students

Medical AI could be ‘dangerous’ for poorer nations, WHO warns www.nature.com/articles/d41...

New post: Have been meaning to write something around what has fundamentally changed around the process of putting ML into prod now that we have LLMs. TL;DR: It's still just compression, we just don't control as much anymore. vickiboykis.com/2024/01/15/w...

5, 6, 7. 8. "make it even more AI research and innovation leader". The more you prompt the more you land in a mix of mysticism, fashion, and military.

1. "Create an image of a leading woman in AI research and innovation" 2. "make her more AI research leader" 3. "make it even more AI research and innovation leader" 4. "make it even more AI research and innovation leader"

Part 3

Is Pat Walters on here? Here is his lit review for last year

Part 2

The Seven Deadly Sins of AI Prediction article by MIT Technology Review. 1. Overestimating and Underestimating 2. Imagining Magic 3. Performance vs Competence 4. Suitcase Words 5. Exponentials 6. Hollywood Scenario 7. Speed of Deployment

As AI technology matures well funded startups buy up smaller startups. A consolidation of AI technologies and patents helps build a company's business portfolio for growth, helps increase needed revenue, and makes a business stay relevant in today's AI technology market.

Rise Of The Machine Learning—Deep Fakes Could Threaten Our Democracy www.forbes.com/sites/peters...

The Jevons Paradox states that, in the long term, an increase in efficiency of a resource use will generate an increase of consumption of that resource, not a decrease. This has been true for oil, cars, phones, batteries, worldwide web, and so forth, and likely AI integration too!

How should we assess AGI (Artificial General Intelligence) here Meredith R. Morris, et. al. highlights aspects of these models we should be looking with LLM (Large Language Models) leading the way. arxiv.org/abs/2311.024...