The definition of a "small" language model (LLM) is constantly evolving, driven by rapid advancements in LLM capabilities and accessibility. What was considered large just a short time ago is now considered small, with models boasting billions of parameters now readily available for personal use and fine-tuning. This shift has blurred the lines between small and large models, making the traditional size-based categorization less relevant. The article emphasizes that the focus is shifting from size to other factors like efficiency, cost of training and inference, and specific capabilities. Ultimately, "small" now signifies a model's accessibility and deployability on more limited hardware, rather than a rigid parameter count.
While the popular belief that smartphones constantly listen to conversations to target ads is untrue, the reality is more nuanced and arguably more disturbing. The article explains that these devices collect vast amounts of data about users through various means like location tracking, browsing history, app usage, and social media activity. This data, combined with sophisticated algorithms and data brokers, creates incredibly detailed profiles that allow advertisers to predict user behavior and target them with unsettling accuracy. This constant data collection, aggregation, and analysis creates a pervasive surveillance system that raises serious privacy concerns, even without directly listening to conversations. The article concludes that addressing this complex issue requires a multi-faceted approach, including stricter regulations on data collection and increased user awareness about how their data is being used.
Hacker News users generally agree that smartphones aren't directly listening to conversations, but the implication of the title—that data collection is still deeply problematic—resonates. Several comments highlight the vast amount of data companies already possess, arguing targeted advertising works effectively without needing direct audio access. Some point out the chilling effect of believing phones are listening, altering behavior and limiting free speech. Others discuss how background data collection, location tracking, and browsing history are sufficient to infer interests and serve relevant ads, making direct listening unnecessary. A few users mention the potential for ultrasonic cross-device tracking as a more insidious form of eavesdropping. The core concern isn't microphones, but the extensive, opaque, and often exploitative data ecosystem already in place.
Android phones will soon automatically reboot if left unused for 72 hours. This change, arriving with Android 14, aims to improve security by clearing out temporary data and mitigating potential vulnerabilities that could be exploited while a device is powered on but unattended. This reboot occurs only when the phone is locked, encrypted, and not connected to a charger, minimizing disruption to users. Google notes that this feature can also help preserve battery life.
Hacker News users largely criticized the proposed Android feature of automatic reboots after 72 hours of inactivity. Many considered it an unnecessary intrusion, arguing that users should have control over their devices and that the purported security benefits were minimal for average users. Several commenters suggested alternative solutions like remote wipe or enhanced lock screen security. Some questioned the actual security impact, suggesting a motivated attacker could simply wait out the 72 hours. A few users pointed out potential downsides like losing unsaved progress in apps or missing time-sensitive notifications. Others wondered if the feature would be optional or forced upon users, expressing a desire for greater user agency.
Belgian artist Dries Depoorter created "The Flemish Scrollers," an art project using AI to detect and publicly shame Belgian politicians caught using their phones during parliamentary livestreams. The project automatically clips videos of these instances and posts them to a Twitter bot account, tagging the politicians involved. Depoorter aims to highlight politicians' potential inattentiveness during official proceedings.
HN commenters largely criticized the project for being creepy and invasive, raising privacy concerns about publicly shaming politicians for normal behavior. Some questioned the legality and ethics of facial recognition used in this manner, particularly without consent. Several pointed out the potential for misuse and the chilling effect on free speech. A few commenters found the project amusing or a clever use of technology, but these were in the minority. The practicality and effectiveness of the project were also questioned, with some suggesting politicians could easily circumvent it. There was a brief discussion about the difference between privacy expectations in public vs. private settings, but the overall sentiment was strongly against the project.
The Register reports that Google collects and transmits Android user data, including hardware identifiers and location, to its servers even before a user opens any apps or completes device setup. This pre-setup data collection involves several Google services and occurs during the initial boot process, transmitting information like IMEI, hardware serial number, SIM serial number, and nearby Wi-Fi access point details. While Google claims this data is crucial for essential services like fraud prevention and software updates, the article raises privacy concerns, particularly because users are not informed of this data collection nor given the opportunity to opt out. This behavior raises questions about the balance between user privacy and Google's data collection practices.
HN commenters discuss the implications of Google's data collection on Android even before app usage. Some highlight the irony of Google's privacy claims contrasted with their extensive tracking. Several express resignation, suggesting this behavior is expected from Google and other large tech companies. One commenter mentions a study showing Google collecting data even when location services are disabled, and another points to the difficulty of truly opting out of this tracking without significant technical knowledge. The discussion also touches upon the limitations of using alternative Android ROMs or de-Googled phones, acknowledging their usability compromises. There's a general sense of pessimism about the ability of users to control their data in the Android ecosystem.
The article explores using a 9eSIM SIM card to enable eSIM functionality on devices with only physical SIM slots. The 9eSIM card acts as a bridge, allowing users to provision and switch between multiple eSIM profiles on their device through a companion app, effectively turning a physical SIM slot into an eSIM-capable one. The author details their experience setting up and using the 9eSIM with both Android and Linux, highlighting the benefits of managing multiple eSIM profiles without needing a physically dual-SIM device. While the process isn't entirely seamless, particularly on Linux, the 9eSIM offers a practical workaround for using eSIMs on older or incompatible hardware.
Hacker News users discussed the practicality and security implications of using a 9eSIM to bridge the gap between eSIM-only services and devices with physical SIM slots. Some expressed concerns about the security of adding another layer into the communication chain, questioning the trustworthiness of the 9eSIM provider and the potential for vulnerabilities. Others were skeptical of the use case, pointing out that most devices support either physical SIM or eSIM, not both simultaneously, making the 9eSIM's functionality somewhat niche. The lack of open-source firmware for the 9eSIM also drew criticism, highlighting the difficulty in independently verifying its security. A few commenters saw potential in specific situations, such as using the 9eSIM as a backup or for managing multiple eSIM profiles on a single physical SIM device. Overall, the sentiment was cautiously curious, with many acknowledging the cleverness of the solution but remaining hesitant about its real-world security and usefulness.
TCL is betting on "NXTPAPER" screen technology, which aims to mimic the look and feel of paper for a more comfortable reading experience. This technology utilizes multiple layers of reflective material to enhance contrast and reduce blue light, creating a display that appears brighter in sunlight than typical LCDs while maintaining low power consumption. While not e-ink, NXTPAPER 2.0 boasts improved color gamut and refresh rates, making it suitable for not just e-readers, but also tablets and potentially laptops. TCL aims to expand this technology across its product lines, offering a paper-like alternative to traditional screens.
Hacker News commenters discuss TCL's NxtPaper display technology, generally expressing skepticism about its widespread adoption. Some doubt the claimed power savings, especially given the backlight required for color displays. Others question the "paper-like" feel and wonder if it truly offers advantages over existing e-ink or LCD technologies for typical use cases. A few commenters express interest, particularly for niche applications like e-readers or note-taking, but overall the sentiment is cautious, awaiting real-world reviews and comparisons to determine if the technology lives up to its promises. Some also discuss the history of similar display technologies and their ultimate lack of success.
Summary of Comments ( 38 )
https://news.ycombinator.com/item?id=44048751
Hacker News users discuss the shifting definition of "small" language models (LLMs). Several commenters point out the rapid pace of LLM development, making what was considered small just months ago now obsolete. Some argue size isn't the sole determinant of capability, with architecture, training data, and specific tasks playing significant roles. Others highlight the increasing accessibility of powerful LLMs, with open-source models and affordable cloud computing making it feasible for individuals and small teams to experiment and deploy them. There's also discussion around the practical implications, including reduced inference costs and easier deployment on resource-constrained devices. A few commenters express concern about the environmental impact of training ever-larger models and advocate for focusing on efficiency and optimization. The evolving definition of "small" reflects the dynamic nature of the field and the ongoing pursuit of more accessible and efficient AI.
The Hacker News post "What even is a small language model now?" generated several comments discussing the evolving definition of "small" in the context of language models (LLMs) and the implications for their accessibility and use.
Several commenters highlighted the rapid pace of LLM development, making what was considered large just months ago now seem small. One commenter pointed out the constant shifting of the goalposts, noting that models previously deemed groundbreaking are quickly becoming commonplace and accessible to individuals. This rapid advancement has led to confusion about classifications, with "small" becoming a relative term dependent on the current state-of-the-art.
The increasing accessibility of powerful models was a recurring theme. Commenters discussed how readily available open-source models and affordable cloud computing resources are empowering individuals and smaller organizations to experiment with and deploy LLMs that were previously exclusive to large tech companies. This democratization of access was viewed as a positive development, fostering innovation and competition.
The discussion also touched upon the practical implications of this shift. One user questioned whether the focus should be on model size or its capabilities, suggesting a shift towards evaluating models based on their performance on specific tasks rather than simply their parameter count. Another commenter explored the trade-offs between model size and efficiency, noting the appeal of smaller, more specialized models for resource-constrained environments. The potential for fine-tuning smaller, pre-trained models for specific tasks was mentioned as a cost-effective alternative to training large models from scratch.
Some comments expressed concern over the potential misuse of increasingly accessible LLMs. The ease with which these models can generate convincing text raised worries about the spread of misinformation and the ethical implications of their widespread deployment.
Finally, several comments focused on the technical aspects of LLM development. Discussions included quantization techniques for reducing model size, the role of hardware advancements in enabling larger models, and the importance of efficient inference for practical applications.