🗞 US Regulators Scrutinize Tech Giants' AI Dominance 👮‍♀️

🔎 The Latest on the AI Frontier:

• US Regulators Scrutinize Tech Giants' AI Dominance 👮‍♀️

• iOS 18 Brings AI-Powered Features and UI Enhancements 📱

• DuckDuckGo Offers Private AI Chats Without Data Training 🔒

• Study Reveals AI Models' Biases on Controversial Topics 🧪

• Google's NotebookLM Upgraded with Gemini 1.5 Pro AI 📝

• AI-Powered Bullying: Deepfakes Invade Schools 🚨

👮‍♀️ US regulators to open antitrust inquiries into Microsoft, OpenAI, and Nvidia's dominant roles in the AI industry. Read More

• The U.S. Department of Justice will investigate whether Nvidia violated antitrust laws, while the FTC will examine the conduct of OpenAI and Microsoft, which has invested $13 billion for a 49% stake in OpenAI's for-profit subsidiary.

• The FTC is also looking into Microsoft's $650 million deal with AI startup Inflection AI, signaling growing regulatory scrutiny into the AI industry.

• U.S. antitrust chief Jonathan Kanter recently expressed concern about "structures and trends in AI that should give us pause," noting that the technology's reliance on massive amounts of data and computing power can give already-dominant firms a substantial advantage.

📱 iOS 18 to bring a mix of AI-powered features and user-friendly improvements to the iPhone. Read More

• Siri will be powered by LLM tech similar to ChatGPT, making interactions more natural, while Apple Photos is expected to get AI retouching capabilities like removing people and objects from scenes.

• Non-AI features include better route planning in Apple Maps, allowing users to create custom navigation routes, animated text messages to add expression, and more flexible app icon placement on home screens.

• iOS 18 will also introduce a refreshed settings menu for easier navigation, color-changing app icons for customization, and more realistic-sounding Siri voices thanks to improved language models.

🔒 DuckDuckGo launches private AI chats, allowing anonymous access to GPT-3.5 and other models without data training by default. Read More

• DuckDuckGo keeps chats anonymous by calling chat models on users' behalf, replacing their IP addresses with its own, making it appear as if the requests come from DuckDuckGo rather than the user.

• The company has agreements with AI providers to delete saved chats within 30 days and prevent them from being used for model training, and users can clear their chats using the fire button, similar to DuckDuckGo's browser.

• The feature is free to use within a daily limit, with plans for a potential paid plan offering higher limits and access to more advanced models, as well as the addition of more chat models, including those hosted by DuckDuckGo or users.

🧪 Study finds that AI models hold opposing views on controversial topics, reflecting biases in their training data. Read More

• Researchers tested several open text-analyzing models, including Meta's Llama 3, and found that they answered questions inconsistently on topics such as LGBTQ+ rights, social welfare, and surrogacy.

• The inconsistencies in the models' responses reflect biases embedded in the data used to train them, with some models refusing to answer "sensitive" questions more often than others, potentially due to political pressures or differences in worldview.

• The study highlights the importance of rigorously testing AI models for cultural biases before their release and implementing comprehensive social impact evaluations to gain insights into their behavior and potential societal effects.

📝 Google's NotebookLM gets a major upgrade with Gemini 1.5 Pro, expanding its availability to over 200 countries and introducing new features. Read More

• NotebookLM now supports Google Slides and web URLs as sources, provides inline citations that directly link to supporting passages, and offers a Notebook Guide to convert sources into FAQs, Briefing Docs, or Study Guides.

• Gemini 1.5 Pro's multimodal capabilities allow users to ask questions about images, charts, and diagrams in Slides or Docs, with NotebookLM including citations to images as supporting evidence when relevant.

• NotebookLM ensures privacy and security by adhering to Google's data protection practices, not training on user data, and implementing safeguards to protect user information while maintaining the quality and credibility of its answers.

🚨 School bullying takes a disturbing turn as students exploit AI to create deepfake nudes, incriminating audio, and other nefarious content. Read More

• Educators are grappling with the emergence of AI-generated sexually explicit images of students, along with the looming threat of sham voice recordings and videos, highlighting the need for updated school policies and laws to address this new form of bullying.

• Incidents like the creation of deepfake nudes of female classmates by teen boys in New Jersey and a mother allegedly using AI to generate fake images of her daughter's cheerleading rivals illustrate the destructive potential of AI in the hands of students.

• Experts emphasize the importance of developing digital literacy and citizenship among students, as AI tools become more accessible and powerful, while also raising concerns about over-surveillance and the need to protect student privacy.

Have any feedback? Send us an email