AI Friend
Highlights
1. Generative AI tools, such as GPT-4 and Midjourney, are being used by product design agencies for product development. This trend is likely to become more popular.
2. OpenAI is rolling out a “Her-like” voice mode, while Friend, an AI startup, focuses on creating a pendant as an AI companion for emotional support.
3. Apple Intelligence is now available in the iOS 18.1 beta, featuring new capabilities for iPhone, iPad, and Mac.
4. Meta has introduced AI Studio, which allows users to create custom chatbots, and SAM 2, a model for identifying and tracking objects in videos.
5. Synchron is developing brain-computer interface technologies that connect to ChatGPT and Apple’s Vision Pro.
Innovation Insights
1. When generative AI meets product development (Sloan Management Review)
Generative AI is transforming product development by enhancing ideation, creativity, and user testing processes. Companies like the Boston design agency Loft use tools like GPT-4 and Midjourney to brainstorm product features and refine designs based on customer preferences.
2. GPT, Claude, Llama? How to tell which AI model is best (The Economist)
Benchmarks such as MMLU, GPQA, and MuSR are crucial for assessing AI models’ progress and reliability, though they can be flawed and sometimes too easy for advanced models. Issues like data contamination and variability in question formatting affect the trustworthiness of benchmark scores. To improve evaluation standards, automated systems like Stanford’s HELM and new, secretive benchmarks like GAIA and NoCha are being developed to provide more accurate and consistent assessments of AI capabilities.
3. AI has a revolutionary ability to parse details. What does that mean for business? (Harvard Business Review)
AI’s ability to focus on detailed, individual data rather than generalizations is transforming business strategies, talent management, leadership selection, and supply chain management by enabling highly personalized and efficient solutions. This shift allows businesses to better serve customers and employees by leveraging AI to identify unique skills, optimize supply chains in real-time, and create dynamic leadership structures based on situational needs.
AI Innovations
1. AI friend
A new AI startup called Friend is developing an AI necklace designed to provide constant emotional support and companionship, with the goal of creating meaningful digital relationships through a wearable device that communicates via text (The Verge).
2. Google
Google’s Gemini AI chatbot is now faster and smarter with the new Gemini 1.5 Flash upgrade, improving response speed, reasoning, image understanding, and increasing the context window to 32K tokens, while also integrating features like fact-checking links and expanding availability (The Verge).
Google’s Gemini 1.5 Pro (experimental version 0801) has surpassed OpenAI’s GPT-4o in performance, with superior capabilities in multilingual tasks, technical domains, and multimodal inputs (VentureBeat).
3. Apple
Apple has released the developer beta for iOS 18.1, featuring the first set of AI capabilities called “Apple Intelligence,” which includes updates to Siri, new smart reply features in Mail, and natural language search in Photos, available for testing on iPhones, iPads, and Macs (The Verge).
4. OpenAI
OpenAI has begun rolling out its advanced “Her-like” voice mode for ChatGPT to a limited number of ChatGPT Plus subscribers, following a delay to improve safety measures and add new filters to block certain types of content (The Verge).
OpenAI announced an experimental version of GPT-4o with a maximum of 64K output tokens per request (OpenAI).
5. Meta
The Segment Anything Model 2 (SAM 2) can identify and track any object across all frames in a video in real-time, enhancing video editing, mixed reality experiences, and various scientific applications by addressing the challenges of segmenting objects in motion (Meta). The Meta demo website also features 1) Seamless Translation, an AI model that translates not just words, but also the vocal style while preserving the expression, emotion and tone, 2) Animated drawings, which turn sketches into animations, and 3) Audiobox that can generate audio stories with AI voices and sounds (Meta).
Meta has introduced AI Studio, a new tool allowing users to create custom AI chatbots that act as virtual extensions of themselves, accessible via the web, Instagram, Messenger, and WhatsApp, aimed at enhancing user interaction and engagement without requiring technical skills (ZDNet).
6. Runway
Runway has introduced a new feature in its Gen-3 Alpha AI model, allowing users to create realistic videos from still images, enhancing artistic control and consistency for video generation (VentureBeat).
7. AI surveillance
The Las Vegas transit system will become the first in the U.S. to fully deploy an AI surveillance system, provided by ZeroEyes, designed to detect and respond to the brandishing of firearms on its buses (CNBC).
8. JPMorgan Chase
JPMorgan Chase has introduced the LLM Suite, a generative AI tool designed to assist employees, particularly in the asset and wealth management division, with tasks such as research, writing, idea generation, and document summarization (Forbes).
9. AI overconfidence
MIT and the MIT-IBM Watson AI Lab have developed a method called “Thermometer” to efficiently calibrate large language models (LLMs) to prevent them from being overconfident about incorrect predictions (MIT News).
10. End-of-life decisions
End-of-life decisions can be distressing for surrogates, but an AI tool called the “digital psychological twin” could help by predicting what incapacitated patients would want based on their personal data, potentially easing the emotional burden on families and providing more accurate reflections of patient preferences (MIT Technology Review).
Other Innovations
1. Brain-computer interface
Synchron has integrated OpenAI’s ChatGPT into its BCI to help people with paralysis control their digital devices more efficiently by providing context-aware, AI-generated response options that can be selected with a single thought-triggered “click” (CNET).
Its BCI also enables patients with paralysis to control Apple’s Vision Pro headset using only their thoughts, marking an industry first in BCI technology integration (CNBC).
2. HIV treatment
A 60-year-old man in Germany has become the seventh person to be cured of HIV after receiving a stem-cell transplant from a donor with only one copy of a CCR5 gene mutation, broadening the potential for HIV treatment strategies beyond targeting the CCR5 receptor alone (Nature).
3. Pandemic trigger
The World Health Organization (WHO) has updated its list of priority pathogens that could potentially trigger the next pandemic, now including over 30 viruses and bacteria such as influenza A virus, dengue virus, and monkeypox virus (Nature).
4. Time crystal
Chinese scientists from Tsinghua University have created a time crystal using rubidium atoms at room temperature, potentially enhancing durability and efficiency for quantum computing applications (MSN).
