AI News

Apple's research on "Specialized Language Models with Cheap Inference from Limited Domain Data" proposes cost-efficient strategies for developing AI, making sophisticated models more accessible. The study identifies key cost areas and recommends tailored approaches, highlighting that effective models depend on suitability rather than size.

ADP Assist is a new HR chatbot aimed at streamlining processes for HR departments, enhancing payroll accuracy, and providing data-driven insights. Leveraging ADP's extensive HCM dataset, it helps professionals and employees access critical information, promoting efficiency and compliance.

Roblox has launched an AI-powered real-time chat translation feature, enabling seamless communication across 16 languages among its 70 million daily users. This groundbreaking tool enhances global inclusivity and user engagement while supporting Roblox's goal of localized experiences.

Deepfakes are increasingly used for malicious scams, exemplified by a Hong Kong case where a finance employee was tricked into transferring $25.6 million after a video call with scammers posing as company executives. Six individuals have been arrested for similar crimes, highlighting the urgent need for awareness and countermeasures against this technology.

A leaked document indicates Google plans to rename its Bard AI tool to Gemini next week as part of integrating its new multimodal AI model. A paid Gemini Advanced tier will launch, offering enhanced capabilities. The app will expand to Canada and over 150 countries.

OpenAI launched a ChatGPT app for Apple Vision Pro, utilizing GPT-4 Turbo for intuitive human-AI interactions via speech and images. It supports multimodal inputs and enhances user experiences with visionOS features. The app is free with an optional Plus subscription for faster response times.

Hugging Face has launched free, customizable Hugging Chat Assistants, allowing users to create personalized chatbots with various open source LLMs. This contrasts with OpenAI’s paid custom GPT Builder. Users praise the Assistants for their flexibility, while acknowledging some advantages of OpenAI's tools.

When deepfake videos of Taylor Swift went viral, fans rallied for her protection, highlighting the dangers of deepfake technology. As AI tools become more accessible, manipulation risks increase, raising public concern. Detection methods are vital as technology evolves, necessitating skepticism in media consumption.

This week's Big Tech earnings calls emphasized advancements in AI. While Google, Microsoft, and Amazon highlighted their AI initiatives, Meta distinguished itself with a comprehensive open-source AI strategy. CEO Mark Zuckerberg showcased Meta’s extensive data resources and commitment to investment in AI infrastructure, including deploying 350,000 H100 GPUs.

Microsoft's Copilot, powered by GPT-4, is expanding across its products. Recent launches include Copilot for Sales and Copilot for Service, aimed at enhancing productivity in CRM systems. Each costs $50 per user per month. Both tools streamline workflows and improve efficiency for various roles.

The rise of large language models (LLMs) has facilitated audio jacking, where cybercriminals manipulate live conversations to steal sensitive information. IBM's research shows LLMs can hijack financial discussions with just three seconds of voice data. To combat this threat, vigilance, advanced detection technology, and adherence to best practices are essential.

Amazon has launched Rufus, a generative AI shopping assistant currently in beta for select users, aimed at enhancing the online shopping experience. Rufus answers product-related queries, offers advice, comparisons, and gift suggestions, and is set for wider release soon.