AI Breakthroughs This Week

 

This week has been buzzing with significant advancements in artificial intelligence, from Apple's long-anticipated AI strategy to Meta's strategic acquisition, and new capabilities for Google's Gemini, Perplexity's enhanced search, and Anthropic's Claude. Here's a rundown of the key announcements shaping the future of AI.

Apple Unveils "Apple Intelligence" at WWDC

Apple has officially entered the generative AI race with "Apple Intelligence," a suite of new AI features integrated across iOS 26, iPadOS 26, macOS Tahoe 26, and watchOS 26. While the much-anticipated Siri overhaul was notably absent and pushed to next year, Apple's focus is on privacy-preserving, on-device AI, with some features leveraging cloud processing for more complex tasks.

Key Highlights:

• Foundation Models & Developer Access: Apple is opening up its on-device foundation models to third-party developers, allowing them to build intelligent features directly into their apps. This could lead to a wave of new AI-powered experiences on Apple devices.
• Visual Intelligence Enhancements: Building on existing visual search capabilities, Visual Intelligence now works across your iPhone screen. You can screenshot content and use Apple Intelligence to extract details, suggest actions (like adding an event to your calendar), or even ask ChatGPT questions about the image.
• Live Translation: Apple is introducing real-time translation for messages, FaceTime, and phone calls across its devices. This feature aims to break down language barriers by translating text as you type or providing voiced translations during calls.
• Smarter Shortcuts: The Shortcuts app is getting an AI upgrade, enabling more complex multi-step automations powered by Apple Intelligence, either on-device or via Private Cloud Compute.
• Genmoji and Image Playground: New creative tools allow users to generate custom emojis by combining existing ones (Genmoji) and create playful images with various themes and styles (Image Playground). ChatGPT integration will also offer additional image creation styles.
• Workout Buddy for Apple Watch: The Apple Watch will gain an AI-powered "Workout Buddy" that provides personalized motivation and tips based on your health and fitness data.
• "Liquid Glass" Design: Beyond AI, Apple also unveiled a new "Liquid Glass" design aesthetic for its operating systems, featuring translucent elements and improved windowing for a more cohesive user experience.

Meta's Significant Stake in Scale AI

Meta is reportedly finalizing a substantial investment of nearly $15 billion for a 49% stake in Scale AI, a leading provider of data for AI development. This move is Meta's largest external investment to date and signals a significant push into artificial general intelligence (AGI).

Key Details:

• Strategic Investment: The acquisition aims to bolster Meta's AI efforts, particularly after reports of its Llama 4 models falling short of internal performance benchmarks and the delayed release of its flagship "Behemoth" AI model.
• Leadership Role for Scale AI CEO: As part of the deal, Scale AI CEO Alexandr Wang is expected to take a top leadership position within Meta, potentially leading a new "superintelligence" lab.
• Data for AI Training: Scale AI is renowned for providing vast amounts of labeled data, crucial for training sophisticated AI models like OpenAI's ChatGPT. This acquisition could provide Meta with a critical advantage in developing its own advanced AI systems.

Google Gemini Announcements and Geospatial Reasoning

Google continues to integrate Gemini, its multimodal AI, across its ecosystem, with new features aimed at productivity and proactive assistance, alongside significant advancements in geospatial AI.

Key Updates:

• Gemini in Google Forms: Gemini is now available in Google Forms to summarize responses to short-answer and paragraph questions, providing quick insights and key takeaways.
• Scheduled Actions for Gemini: The Gemini mobile app is gaining "Scheduled Actions," allowing users to assign recurring tasks to the chatbot that it will complete automatically at chosen times. This pushes Gemini towards becoming a more proactive AI agent.
• Expanded Google Home Controls: The Google Home web app is getting more controls, and Gemini will enable users to send broadcasts to speakers in their home or search camera history using natural language.
• Geospatial Reasoning: Google Research has introduced "Geospatial Reasoning," a new research effort that combines generative AI with multiple geospatial foundation models to accelerate problem-solving. This aims to unlock powerful insights for crisis response, public health, climate resilience, and commercial applications.
• Natural Language Queries: Users can ask complex natural language questions, and Gemini will plan and execute a chain of reasoning, analyzing various geospatial and structured data sources to provide insights and visualizations.
• New Foundation Models: This initiative introduces new remote sensing foundation models for experimentation, trained on vast amounts of satellite and aerial imagery to analyze data undecipherable to the human eye.
• Integration with Google Earth and BigQuery: Gemini capabilities are being piloted in Google Earth to accelerate geospatial analyses in a no-code environment, and new geospatial analytics datasets from Earth Engine and Google Maps Platform are being integrated directly into BigQuery.

Perplexity Acquires Carbon to Supercharge Enterprise Search

Perplexity AI has announced the acquisition of Seattle-based startup Carbon, a move set to significantly enhance its enterprise search capabilities.

Key Details:

• Retrieval-Augmented Generation (RAG): Carbon specializes in RAG technology, which connects large language models (LLMs) to external data sources. This acquisition will allow Perplexity to offer more personalized and context-aware AI search solutions.
• Seamless Data Integration: The integration of Carbon's technology will enable Perplexity users to search through internal documents and data across various platforms like Notion, Google Docs, and Slack. This aims to create more capable and personalized knowledge assistants for the workplace.
• Strategic Move: This is Perplexity's second acquisition, signaling a strategic focus on expanding its capabilities beyond traditional web search to compete in the burgeoning enterprise AI search market against giants like Google and OpenAI.

Claude (Anthropic) Announcements

Anthropic's Claude AI has also been in the news, though with a mix of strategic moves and ongoing legal challenges.

Key Developments:

• Reddit Lawsuit: Reddit has filed a lawsuit against Anthropic, alleging that the AI company illegally "scraped" user comments to train its Claude chatbot without consent, breaching Reddit's terms of use. This highlights the ongoing legal complexities surrounding AI training data.
• Claude Explains Blog Closure: Anthropic has discontinued its "Claude Explains" blog, which showcased the writing capabilities of its Claude AI models. While Anthropic stated it was a collaboration between human experts and AI, the closure sparked discussions about audience acceptance and transparency in AI-generated content.
• Claude in Cloudflare Development: A Cloudflare developer revealed that Claude was largely responsible for writing the code for an open-source OAuth library published under the Cloudflare Workers project, with the entire prompt history documented. This offers a rare glimpse into human-AI pair programming for critical infrastructure.

---

Content Source: Gemini AI – https://gemini.google.com/

Image Source:  ChatGPT – https://chat.openai.com/

 

 

 

This week Elon Musk and Company are calling for an “AI-First” strategy in government. I am not sure I am onboard with a totally autonomous system nor the removal of millions of workers in the federal workforce, but what I am certain of is that AI is useful, complicated, and evolving. A good definition of an AI system is that it is an ecosystem of infrastructure for models, data pipelines, processes, and teams. The ecosystem promotes end-to-end computer solutions of real-world problems that are sustainable and align with societal values. An AI system is the practical deployment of models delivering value at scale.

Following 20+ years as a geospatial professional in the USDA Forest Service (35+ years in federal service), I had been tasked with the coordination and integration of geospatial technology into the federal agency on a local unit (i.e., on the Monongahela National Forest), at a regional level (i.e., Region 1 based in Missoula, Montana), and at a national level at the Geospatial Technology and Applications Center (GTAC) in Salt Lake City. This experience taught me that the evolution of this new technology and applied use in an ecological land management context was fraught with complexity and required an interdisciplinary framework for capturing programmatic content – lands, engineering, recreation, wildlife, soils, geology, law enforcement are representative programs with professional staff and a wealth of data that required analysis, management, and coordination.

AI models and systems hold great promise for augmenting geospatial professionals and resource specialists in their work as well as providing foundational insights for Agency leadership in making decisions. Current techniques and tools exist for automating much of the processes right now, but the integration of AI systems and task-based models would help to not only solve real-world problems but potentially reveal emergent and innovative approaches to natural resource management.

Current AI applications in the natural resource and land management realm are primarily task-based and concentrated in the remote sensing arena.  Several remote sensing examples include feature extraction and change detection, predictive analytics, and natural language processing. Specific real-world examples that have been developed are:

§  Future of AI in natural resource management: Self-Learning Forest Growth Model (Liang, 2023)

Mr. Liang is co-director of the Forest Advanced Computing and Artificial Intelligence Lab and is leading a project to use AI to map global forest carbon accumulation rates (Liang, 2023).

§  Exploring artificial intelligence for applications of drones in forest ecology and management (Buchelt et al, 2024)

This article is an in-depth look at utilizing drone technology for real-time tracking of forested landscapes for wildfires, pests, and inventory needs to name a few.

Another great source of information on applied AI in the geospatial world is Esri. Esri is the leading commercial vendor for geospatial software with their ArcGIS suite of applications. They are leading the GeoAI charge which is what they define as:

 “…the application of artificial intelligence (AI) fused with geospatial data, science, and technology to accelerate real-world understanding of business opportunities, environmental impacts, and operational risks. Organizations are modernizing operations to run at scale through automated data generation and approachable spatial tools and algorithms” (Esri, 2024). 

Esri also has a free ebook available for download which describes “AI + Location Intelligence” here.

A major challenge is that data integrity has been and is a major concern for grappling with Big Data – those big data pipelines are integral to not only the training and testing of data for AI, but also for the long-term maintenance of an AI system. At the same time Agency leadership has struggled with data governance and staffing – there will be similar struggles for implementing AI across an organization.

Key tools for fighting wildfires (and post-fire mitigation) will be the expanded use of drones and the real potential of integrating AI into the battle. Fundamentally though, it will be a mistake to dismiss the need for a highly structured methodology in not only integrating the technologies (with robotics), but for prioritizing training needs for staff use and incident coordination.

As AI technology matures and begins to deploy autonomously – with model methodology making a transition to AI systems from task-based models – there will a greater emphasis and success for working across disciplines and engaging interdisciplinary teams in land and natural resource management.