Is Sam Altman Gambling With the U.S. Economy? Sam Altman tours the Stargate AI data center in Abilene, Texas on Tuesday, Sept. 23, 2025. Sam Altman tours the Stargate AI data center in Abilene, Texas on Tuesday, Sept. 23, 2025. Welcome back to In the Loop, new twice-weekly newsletter about AI. If you're reading this in your browser, why not subscribe to have the next one delivered straight to your inbox?

To hear Silicon Valley tell it, the end of disease is well on its way. Demis Hassabis, a Nobel laureate for his AI research and the CEO of Google DeepMind, said on Sunday that he hopes that AI will be able to solve important scientific problems and help "cure all disease" within five to 10 years. Earlier this month, OpenAI released new models and touted their ability to "generate and critically evaluate novel hypotheses" in biology, among other disciplines. These are all executives marketing their products, obviously, but is there even a kernel of possibility in these predictions? If generative AI could contribute in the slightest to such discoveries--as has been promised since the start of the AI boom--where would the technology and scientists using it even begin?

Recent advancements in Large Language Models (LLMs) raised concerns over potential misuse, such as for spreading misinformation. In response two counter measures emerged: machine learning-based detectors that predict if text is synthetic, and LLM watermarking, which subtly marks generated text for identification and attribution. Meanwhile, humans are known to adjust language to their conversational partners both syntactically and lexically. By implication, it is possible that humans or unwatermarked LLMs could unintentionally mimic properties of LLM generated text, making counter measures unreliable. In this work we investigate the extent to which such conversational adaptation happens. We call the concept $\textit{mimicry}$ and demonstrate that both humans and LLMs end up mimicking, including the watermarking signal even in seemingly improbable settings. This challenges current academic assumptions and suggests that for long-term watermarking to be reliable, the likelihood of false positives needs to be significantly lower, while longer word sequences should be used for seeding watermarking mechanisms.

Exasperated parents may decry video games as a mind-destroying waste of time as they try to prise their children away from them. But could video games – whisper it – actually be good for us? A new study from Japan suggests they can significantly improve mental wellbeing in children and adults. 'These games can serve as distraction or emotional release, providing stress relief, mood enhancement and a sense of accomplishment,' Dr Hiroyuki Egami, an assistant professor at Nihon University and the lead author of the study, told Good Health. The behavioural scientist, who analysed the effects of gaming on people aged from ten to 69, added that many parents'feel excessively worried about their children's video game habits, fearing potential negative consequences'... but'our study now provides robust scientific evidence to ease some of those worries'.

In the wake of California withdrawing Cruise's permit to operate self-driving cars in the state, the company said on Friday it's suspending all U.S. robotaxi operations. The move comes after the California Department of Motor Vehicles alleged that Cruise withheld from regulators video footage of a Cruise robotaxi dragging a person down a city street. The future for the company is anybody's guess. Its parent company, General Motors, has lost $1.9 billion on Cruise thus far this year, including a $732-million loss in the third quarter, according to its latest earnings report. Competitor Ford shut down its Argo robotaxi unit in 2002, concluding that the possibility of far-off profits weren't worth the enormous cash drain.

I've got some of these perceptions and systems technology and emergent behavior pieces that are relevant to this, and I'm smart enough to apply the necessary things. So, while I've got people that invested $20 million in my company, I'm not telling them that I'm likely to have the breakthrough for artificial general intelligence. Instead, I'm saying there's a non-negligible chance that I will personally figure out some of the important things that are necessary for this. The emergence of artificial general intelligence in a way that can impact the economy really is a'change-the-world-level' event, where this is something that reshapes almost everything that human beings can do. This is something that is almost the largest scale that you can think about.

In my previous post, I discussed using ChatGPT, the large language model from OpenAI [1], as a writing partner for various types of prose. In this article, I will show how the system can be used to help compose music by generating chords from text prompts. After a brief overview of ChatGPT, I will show the results of my experiments in writing music with the new system in the following styles: jazz, country rock, and reggae. I'll finish by giving my general observations on using the model for composing music with some next steps for future exploration. Note that the third and final installment of this series will be about using the system to create picture books with help from Midjourney. ChatGPT is the latest language model from OpenAI that was designed and trained to interact with people via a chat user interface. GTP stands for Generative Pre-trained Transformer, where a transformer is a type of AI model. You can read a complete background on the system in this series' first article.

The area of data science is large and fast expanding. It's no surprise that so many people want to learn more about it! But what is data science, and what do you need to know if you want to work in this field? One of the most important things to understand about data science is that it is a very hands-on and ever-changing discipline. It's critical to keep learning new things in order to stay current with the latest trends and practices in the field.

Writing engaging blogs is an art and to master the art, you need to practice it. There are no shortcuts to writing quality blogs, and you must learn from your mistakes and improve with time. You also need to assess the trend and write about things you love. Make it personalized, grow your brand, and increase your following. In this blog, we will learn about the rules of writing engaging technical blogs.

In my previous post, I introduced a framework to select the "right" KPIs to evaluate your machine learning models. When dealing with a classification problem, the chosen KPI will usually measure the rate of mistakes over time, as a point estimate (e.g. In this post, I will touch on one of the main challenges in many real world problems: how to know whether your model is actually correct or not. In supervised learning models, getting feedback on the model's decisions is crucial for both model training and evaluation. But getting such feedback can sometimes be very challenging or require a lot of resources.