Good morning! This is your daily ☕️ Techpresso.
In today's Techpresso:
🫠 TikTok knows exactly how many videos it takes to get addicted
⚔️ Silicon Valley is debating if AI weapons should be allowed to decide to kill
🤖 OpenAI launches new multi-agent framework 'Swarm'
🧠 Apple AI researchers question OpenAI's claims about o1's reasoning capabilities
🔮 Apple wants some 'technological uniqueness' before entering foldable smartphone market
🎁 + 7 other news you might like
🔮 + 3 handpicked research papers and tools
🫠 TikTok knows exactly how many videos it takes to get addictedLINK
Internal TikTok documents reveal that users can become addicted to the platform after watching only 260 videos, with such engagement linked to negative mental health effects like loss of analytical skills, empathy, and increased anxiety.
The documents are part of a lawsuit filed by 14 attorneys general alleging TikTok's algorithm falsely advertised its addictiveness.
Despite internal findings that features like screen time alerts had negligible impact on reducing use, TikTok continued to promote these measures.
⚔️ Silicon Valley is debating if AI weapons should be allowed to decide to killLINK
In Silicon Valley, there is a heated debate about whether AI should be allowed to autonomously decide to kill, with figures like Anduril co-founder Palmer Luckey expressing some openness to this idea, despite opposition from co-founder Brandon Tseng and government officials.
The U.S. government currently does not mandate a ban on fully autonomous lethal weapons, and while it has guidelines for AI safety, these remain voluntary, leaving room for tech companies like Anduril and Palantir to explore more autonomy in military technology.
Concerns grow over the possibility that adversaries like China and Russia might deploy fully autonomous weapons first, prompting some defense tech leaders to lobby for more nuanced policies around AI in military systems.
🤖 OpenAI launches new multi-agent framework 'Swarm'LINK
OpenAI has unveiled an open-source framework named "Swarm" on GitHub, described as an experimental platform designed for developing, orchestrating, and deploying multi-agent systems.
Swarm is engineered to simplify agent coordination and execution, leveraging concepts like routines and handovers, which enable complex interactions and task handovers among agents.
OpenAI highlights Swarm as an innovative yet exploratory tool for multi-agent system interfaces, suitable for scalable solution testing and offering developers fine control without extensive prerequisites, though it is not intended for production use.
🧠 Apple AI researchers question OpenAI's claims about o1's reasoning capabilitiesLINK
Apple researchers, including Samy Bengio, question the reasoning capabilities of OpenAI's latest model o1, suggesting it mimics patterns rather than using true logic despite its high performance on benchmarks.
The team developed an evaluation tool called GSM-Symbolic, which revealed performance fluctuations in AI models, indicating that current benchmarks like GSM8K might not reliably test logical reasoning abilities.
Apple's study highlights a broader debate with OpenAI over the extent of logical reasoning in AI.
🔮 Apple wants some 'technological uniqueness' before entering foldable smartphone marketLINK
Apple is delaying its entry into the foldable smartphone market to ensure it offers a technologically unique product, differentiating itself from competitors who have already advanced significantly in this segment.
The company is focusing on gaining control over critical components, like Ultra-Thin Glass and backplates, to enhance the durability and performance of its prospective foldable devices, which is currently dominated by Samsung.
Apple believes that maintaining reliance on external manufacturers for key parts might hinder its ability to stand out, as it aims for a launch that offers distinct features or innovations compared to existing foldable smartphones.
Other news you might like
Leak: AMD’s Ryzen 9000X3D chips aren’t looking like a leap forward.LINK
Steam now says the ‘game’ you’re buying is really just a license.LINK
Instagram blames some moderation issues on human reviewers, not AI.LINK
US labor board accuses Apple of restricting workers' Slack, social media use.LINK
Anthropic's Claude vulnerable to 'emotional manipulation'.LINK
Are Tesla’s robot prototypes AI marvels or remote-controlled toys?.LINK
Ex-FTX leader updates LinkedIn profile with prison stint.LINK
Latest research and tools
zendesk.md: 1 bug, $50,000+ in bounties, how Zendesk intentionally left a backdoor in hundreds of Fortune 500 companies.LINK
Swarm: an experimental framework by OpenAI designed to demonstrate and explore the orchestration and interaction of multiple AI agents, focusing on education and not intended for production use.LINK
Canary: a self-hostable search tool that replaces Algolia DocSearch, offering features like glob/tag filters, light/dark mode, multi-panel interface for search/Ask AI, and more, with documentation available at getcanary.dev.LINK
Want to get the latest news differently? Find us on:
See you tomorrow for a new dose of ☕️ Techpresso!
-min.png){kind=small}
.png){kind=small}
