%20(1).png)
In September, OpenAI introduced its newest AI model, o1 (internally known as "Strawberry"), the first in an upcoming series of models focused on advanced reasoning capabilities. The company designed this model to handle complex tasks requiring intricate reasoning, such as multi-step mathematical problems and sophisticated coding challenges. Unlike previous models, o1 emphasizes a more prolonged processing phase to analyze queries thoroughly before generating responses, employing a "chain of thought" prompting technique that mimics human cognitive processes. This approach aims to improve the accuracy and nuance of the outputs, marking a significant advancement in AI.
While o1 is still in its early stages and lacks some of ChaGPT's features, such as web browsing and file uploads, it has demonstrated impressive performance in tests, achieving an 83% accuracy rate in a qualifying exam for the International Mathematics Olympiad -- a notable improvement over its predecessor. OpenAI has also released a scaled-down version, o1-mini, designed for cost-effective code generation tasks and performs nearly as well as o1 in STEM-related evaluations. Both versions are initially available to ChatGPT Plus and Team users, with plans for broader access in the future, although the pricing for developer access is higher than previous models.
Read more at the Tech Portal.
Meta is facing significant challenges in the EU following a ruling from the European Court of Justice that restricts its ability to retain user data for advertising indefinitely. The court emphasized compliance with the General Data Protection Regulation (GDPR), which mandates strict data minimization principles, potentially exposing Meta to fines that could reach billions.
The ruling, stemming from privacy advocate Max Schrems' challenge, limits not only data retention practices but also prohibits the use of sensitive information for ad targeting. This may impact Meta's revenue, as approximately 10% of its global ad income comes from the EU.
As Meta adapts to these new regulations, the digital advertising landscape is shifting towards greater emphasis on privacy, serving as a critical reminder of the balance between technological innovation and regulatory compliance.
More info at TechCrunch
The CNIL (France's Data Protection Authority) has released its final recommendations for improved privacy protection on mobile applications, with enforcement set to begin in 2025. According to regulators, as mobile app use rises among French citizens, so do data confidentiality and security risks. The CNIL's recommendations target various stakeholders in the mobile ecosystem, including publishers, developers, and SDK, app store, and operating system providers. They mean to clarify roles and responsibilities, improve user information regarding data usage, and ensure that consent for data processing is both informed and voluntary.
This set of recommendations stems both from extensive discussions with mobile stakeholders and public feedback, which led to more precise distinctions between mandatory requirements and best practices. The guidelines stress the importance of transparent communication concerning data permissions and the necessity of user consent for non-essential data processing, such as targeted advertising. The CNIL plans to continue supporting the industry (including through informative webinars) through its evolution and will initiate a compliance investigation campaign in early 2025 to ensure adherence to these privacy standards. In the meantime, the regulation authority will continue to address complaints and conduct investigations related to user tracking and consent issues in mobile applications.
More details can be found on CNIL's official website.
Discover all the latest news, articles, webinar replays and fifty-five events in our monthly newsletter, Tea O'Clock.
.svg)
