🧵 2. VPNs may log info like data usage, login frequency, and device details. But why? Data could be used for legit purposes or shared with advertisers.

🧵 1. Microsoft AI CEO challenges current copyright norms advocating for all web content being considered fair use by default, sparking debate. #tech

🧵 1. Curious about what data your VPN might be logging about you? Learn why it matters and how it could impact your privacy. By Chris Stobing. #tech

🧵 Read more at:

🧵 5. Don't fall for national-level solutions to internet issues. The "jerk-off permit" saga in Spain is a perfect example of costly, pointless endeavors.

🧵 4. The global nature of the internet makes country-level regulation ineffective. Spain's misguided attempt to control online content is a waste of resources.

🧵 3. Attempting to regulate internet access is futile on a national scale. Most major porn sites are international and won't comply with individual country rules.

🧵 2. Spain's minister's plan for online age verification backfired, earning the nickname "paja passport." It's a costly, ineffective response to a non-issue.

🧵 1. Simplifying complex issues leads to flawed solutions. Spain's "jerk-off permit" debacle highlights the futility of national-level internet regulation. #tech

🧵 Read more at:

🧵 5. Agreements aim to foster collaborative research on evaluating capabilities and risks of AI models, along with methods to mitigate potential safety concerns.

🧵 4. OpenAI CEO Sam Altman expresses satisfaction on collaboration with U.S. AI Safety Institute for pre-release testing of upcoming models.

🧵 3. Establishment of U.S. AI Safety Institute came after U.S. government's inaugural executive order on AI, emphasizing safety assessments and equity.

🧵 2. U.S. AI Safety Institute, part of NIST, to evaluate major new AI models from OpenAI and Anthropic prior to public release, per agreement.

🧵 1. OpenAI and Anthropic partner with U.S. AI Safety Institute to test new AI models for safety and ethics concerns, following industry worries. #tech

🧵 Read more at:

🧵 5/ Organizational success across industries will soon rely on the efficiency of sending prompts to Large Language Models.

🧵 4/ Language bias impacts token usage, especially in non-English prompts, affecting efficiency in using Large Language Models.

🧵 3/ The effectiveness of data tokenization varies based on chosen criteria and is currently skewed towards English content.

🧵 2/ We're shifting from focusing solely on vast training data to prioritizing efficient data conversion through tokenization.