Asahi Linux has released conformant OpenGL ES 3.1 drivers for M1 and M2 family GPUs, allowing compatibility with any OpenGL ES 3.1 application on Linux.
These drivers have been recognized by the standards body, Khronos, after passing industry-standard tests.
Asahi Linux's implementation includes support for compute shaders and atomics on images, despite the lack of hardware instructions for image atomics on the M1 GPU.
The discussion delves into the adherence of Apple to standards, including criticisms of Adobe's products.
Apple's control over protocols and formats and the differences between Vulkan and Metal graphics APIs are also discussed.
The conversation also covers topics such as reverse engineering of Apple's chips, development of drivers for Apple Silicon, frustrations with Apple's ecosystem, and search for alternatives.
Microsoft is integrating Python support into Excel, enabling users to leverage Python functions within the spreadsheet software.
Users have expressed mixed reactions, with some excited about the new capabilities and others raising concerns.
Concerns include the difficulty of getting started, limited accessible examples, and potential data security issues.
Some users desire more advanced data analysis features in Excel, while others discuss the limitations and frustrations of using Excel for certain tasks.
The cloud-based approach and potential compatibility issues and security concerns with running Python in Excel are also brought up.
Overall, the feature aims to enhance Excel's capabilities and broaden its functionality, but opinions are divided about its implications.
Hackers are infiltrating credit bureaus and selling personal information, including addresses, phone numbers, email addresses, and Social Security numbers.
The tool used to access this data is available for as little as $15 in Bitcoin, making it accessible to criminals.
The stolen data can be used for various crimes, such as swatting, SIM swapping, and physical violence.
Credit bureaus and data brokers like TLOxp, Data-Trac, SearchBug, and USinfoSearch have come under scrutiny for their involvement in selling personal data.
Lawmakers are urging action to prevent this practice, and the Consumer Financial Protection Bureau has proposed new rules to regulate the trade of credit header data.
The effectiveness of these changes in addressing the problem remains uncertain.
KPI (key performance indicator) psychosis is a prevalent issue in technology companies, where decisions are solely based on data and disconnect from reality.
Companies heavily rely on data to overcome human biases, but using KPIs as the sole basis for decision-making has its limitations.
The article suggests a solution that combines intuition and data, continuously reflects on the reliability of KPIs, and focuses on the actual objective rather than just the KPI itself.
OpenAI has introduced a new feature called fine-tuning for their GPT-3.5 Turbo model, allowing developers to customize it for specific use cases.
Early tests have shown that a fine-tuned version of GPT-3.5 Turbo can outperform the base GPT-4 model in certain tasks, indicating improved performance.
Fine-tuning offers benefits such as increased steerability, reliable output formatting, and the ability to customize the tone. The updated model also supports handling 4k tokens and offers cost savings through prompt size reduction.
OpenAI has released pricing details for fine-tuning and announced the availability of updated GPT-3 models.
OpenAI has implemented moderation systems to ensure the safe deployment of fine-tuned models and plans to launch a fine-tuning user interface soon.
OpenAI has made updates to their GPT-3.5 Turbo model fine-tuning and API, allowing the model to learn specific styles and information.
Fine-tuning has limitations and potential biases, so OpenAI suggests using retrieval for certain tasks instead.
Fine-tuning can be less powerful but cheaper than traditional methods, but it may not always yield better results and can introduce biases. OpenAI plans to release GPT4 fine-tuning in the future.
The author shares their experience with a SSD failure on their laptop and how they successfully recovered their data using ZFS incremental replication.
The post discusses the repairability and power efficiency of AMD CPUs, along with different storage expansion and backup strategies.
The conversation explores the benefits and challenges of replicating drives and troubleshooting hardware issues, while also mentioning alternative tools and concerns about other file systems like btrfs.
The US government sought an agreement with TikTok that would have granted them significant control and access to the app's operations, according to a leaked draft agreement obtained by Forbes.
The agreement would have given US agencies extensive powers, such as the ability to examine TikTok's records and servers, veto executive appointments, and demand changes to the app's terms of service.
The government's demands drew criticism as they resembled surveillance tactics commonly associated with China. TikTok and the Committee on Foreign Investment in the United States (CFIUS) declined to comment on the draft agreement.
Experian, a credit reporting agency, has been fined $650,000 by the US government for violating spam laws and not providing an opt-out option for marketing emails.
The Federal Trade Commission (FTC) discovered that Experian sent deceptive marketing emails to customers with free credit monitoring memberships, lacking clear notice and an opt-out mechanism.
As a result, the Department of Justice (DOJ) obtained a permanent injunction against Experian, prohibiting certain types of messages and requiring explicit opt-out options in marketing emails.
Experian, a credit reporting company, has been fined $650,000 by the FTC and DOJ for sending spam emails that lack an opt-out option.
Critics believe the fine is too small in comparison to Experian's revenue, suggesting that larger penalties are necessary to discourage similar conduct by other companies.
The absence of an opt-out option in the spam emails is a violation of regulations, leading to the imposition of the fine by regulatory authorities.
The new log/slog package in Go 1.21 introduces structured logging to the standard library.
Structured logs utilize key-value pairs for easy parsing, filtering, searching, and analysis.
The package aims to create a standardized framework for structured logging in the Go ecosystem, with different log levels and customizable output format options.
The discussion focuses on structured logging in Go and the various approaches used by logging libraries.
The benefits of structured logging and the importance of standardized logging formats and APIs are highlighted.
Participants share their views on error handling and the significance of call stacks. Some existing tools for log visualization are mentioned, along with suggestions for enhancing logging efficiency.