The Third Generation of Robots

Figure 03 is the third generation of Figure’s humanoid robot, built to be a smart assistant capable of performing household chores and work in public institutions. The new model features an advanced AI system that enables it to understand environment, complex contexts, process visual information in real-time, manage a conversation, understand the actions it is asked to perform, learn new tasks by watching and self-experience, and correct its mistakes. It charges wirelessly through its feet. These robots will soon become a part of our daily lives.

Doctor Google

Google has published research on the development of an advanced AI agent in the healthcare field called Wayfinding AI, based on Gemini. This agent will understand the user’s medical problem by proactively asking clarifying questions in order to provide personalized recommendations.

AI-Powered Cyber Attacks

Scammers created a fake WhatsApp account with a picture of Mark Read, CEO of the advertising group WPP, and scheduled a Microsoft Teams meeting that appeared to be with Read and another senior company employee. The fraudsters attempted to extract funds and personal details by requesting to establish a new business. Thanks to the alertness of one of the managers, the attempted fraud failed. Cyberattacks are becoming more sophisticated, using Artificial Intelligence and Deepfake technology to exploit virtual meetings. The incident reflects the surge in deepfake attacks in the corporate world, which have already successfully defrauded banks and financial companies.

Revolution in Cell Repair

Researchers from the University of Stuttgart have developed a miniature 3D printer capable of printing and building biological tissues inside the human body itself. The printer uses a thin optical fiber with a tiny, salt-grain-sized lens at its tip. By using laser beams to layer-cure bio-inks, precise cellular structures can be built right inside the body, eliminating the need for pre-grown tissue transplants.

ChatGPT and Medicine

A study examining ChatGPT‘s capabilities in the medical field found that it is highly successful at identifying names of diseases and medications but fails when required to interpret vague descriptions of symptoms presented by users. The reason is that ChatGPT was trained on structured and precise medical texts like medical literature, Wikipedia, and clinical guidelines, and therefore struggles to interpret inaccurate symptom descriptions like “my stomach feels weird.” Unlike a doctor, the model lacks causal reasoning ability and the need to choose between different diagnoses to reach an accurate conclusion in the context of a regular medical conversation.

AI-Based Browser

Perplexity launched the Comet browser, based on Artificial Intelligence, which includes a personal assistant powered by AI agents that helps perform multi-step tasks. Examples include conducting competitor research, opening a new file, and generating a summary report with recommendations.

ChatGPT

• OpenAI launched Sora 2, which includes sound and speech.
• It also launched “Instant Checkout,” which allows users to buy products on Etsy directly from the chat. The service is available in the US for free users as well.
• ChatGPT now allows connecting to applications from within the chat, and also enables sharing projects and collaborating in a single workspace.
• OpenAI launched AgentKit, a workspace for building AI agents simply by dragging and dropping objects.
• It also released a new feature, Company Knowledge, which allows business customers to connect ChatGPT to organizational information such as Google Drive, Slack, SharePoint, and GitHub to search and cross-reference data.
• It released ChatGPT Atlas – an agent-based browser.

Google

• Google launched Gemini Enterprise for organizations, which allows companies to automate tasks with AI agents in areas such as marketing, finance, HR, and more. It can also connect to corporate databases and applications like Microsoft 365, Salesforce, and others.
• It launched a new platform called Google Skills, which is free and open to everyone.
• It’s finally possible to create presentations in Gemini within Canvas.
• Google partnered Gemini with Google Maps, providing up-to-date information on 250 million locations. This makes it possible to find places, including current opening hours. Users can ask about a place’s operating hours or create a route that includes travel time between different sites.
• It also launched Veo 3.1 to maintain character or style consistency within the video.
• AI Mode is also available in Israel.

Anthropic

• Claude can connect to a corporate Microsoft 365 account to search and cross-reference information in SharePoint, OneDrive, Outlook, and Teams.
• It also gained the ability to create a set of skills. Here is a guide for building capabilities.

Microsoft

• Copilot in Windows 11 received connectivity to Gmail and Google Drive, allowing users to search for emails, files, and contacts from within the Copilot conversation.
• Group chat with up to 31 members can be conducted within Copilot via Copilot Groups, and users can talk to Mico – a virtual character in Copilot that responds with emotions when spoken to.
• A new learning mode has been added to Copilot: Copilot Study and Learn Mode.
• Microsoft released Agent Mode for Excel (for data analysis and creating complex formulas) and for Word (to help with message precision).
• It allows the creation of presentations and documents directly from Copilot using the Office Agent.
• It also added the ability to activate Copilot by voice using the phrase “Hey Copilot” to the Windows 11 operating system.
• Another added feature is “Copilot Actions,” which will allow the system to perform real-world tasks (such as ordering a restaurant reservation).
• Microsoft released a Copilot Mode agent in its browser.

AI Tool Updates and Releases

• Genspark released a new feature for Genspark Photo Genius for image editing via voice commands.
• An Israeli startup named Gain offers AI-based virtual employees for managing procurement and supply chains.
• The Lovable platform integrated with Shopify, which allows creating a new store, adding, and updating products via voice prompts.
• Runway allows training video models on organizational data through Model Fine-tuning, enabling brands to easily create advertisements for their products and services.
• Genspark added the Custom Super Agent for building personalized AI agents through a simple prompt. These agents can also be called upon in conversation.