Google proposes new AI training as DeepSeek competes with ChatGPT

Key Takeaways

• Elder fraud losses surged by 43% to $4.89 billion in 2024, driven by AI voice cloning and personalized scam messages targeting the elderly.
• FTI Technology launched IQ.AI Studio in March 2026, providing generative AI tasks and collaboration tools for legal teams.
• Google's Jeff Dean advocates for a new AI pre-training approach where models actively choose learning data to improve efficiency.
• China is drafting regulations for interactive AI services that simulate human personalities, focusing on preventing harmful content and ensuring user well-being.
• ChatGPT (GPT-4o) and DeepSeek Coder V2 are prominent AI tools for programmers in 2026, with DeepSeek specialized for coding tasks.
• AI is not immediately replacing most human jobs due to its "jagged" capabilities and modest productivity gains observed so far.
• Kalepa appointed Audrey Howell, formerly of Palantir Technologies, as Head of Product to advance AI-driven underwriting for insurers.
• Sea.AI is expanding in North America, deploying AI-powered machine vision systems for maritime safety to detect objects in water.
• A study on AI scribes for clinical documentation found only modest time savings for doctors, indicating inconsistent use and limited real-world benefits.
• Simple questions can sometimes resolve complex AI issues, highlighting the importance of balancing AI's technical depth with human common sense.

AI Scams Target Elderly with Voice Cloning

Scammers are using artificial intelligence (AI) to impersonate loved ones and trick elderly people out of money. A recent case involved an 86-year-old woman who lost $6,000 after believing a call from her granddaughter was real. AI voice cloning makes these scams more convincing, preying on the growing elderly population who may be more isolated or less comfortable with technology. In 2024, elder fraud losses increased by 43% to $4.89 billion, with phishing attacks being common. AI also enables realistic fake videos and personalized scam messages.

FTI Technology's IQ.AI Offers Clarity in Legal AI

FTI Technology's IQ.AI platform aims to bring clarity to complex legal matters using artificial intelligence. Launched in March 2026, IQ.AI Studio provides generative AI tasks for legal teams to use. Key features include AI-powered collaboration, pre-built AI tasks for legal issues, and the ability to create custom AI tasks. The platform offers global access to various AI models and integrates with other FTI solutions for enhanced data analysis. This approach focuses on delivering real results and insights in challenging legal situations.

Simple Questions Unlock AI Solutions

An AI user found that complex technical solutions were not working for an autonomous agent named Haru. After hours of troubleshooting, a simple question about providing the passphrase directly solved the problem. This highlights how AI's detailed 'map' of a system can sometimes lead experts to overlook obvious solutions. Beginners, with less complex mental models, can often ask the simple questions that lead to breakthroughs. Effective work with AI requires balancing the AI's technical depth with human common sense.

Google's Jeff Dean Rethinks AI Model Training

Google's Jeff Dean proposes a new approach to pre-training AI models. Instead of just passively absorbing data, future models should actively take actions and choose what data they learn from. Dean believes this active engagement, similar to how humans learn, will significantly improve learning efficiency. He suggests this active learning should be integrated into the pre-training stage, not just post-training. This shift could allow AI models to learn more from the same amount of computational resources.

China Drafts Rules for Interactive AI Services

China has released draft rules for interactive AI services that simulate human personality and engage emotionally with users. These rules apply to AI services within mainland China that interact through text, images, audio, or video. Service providers must prevent the generation of harmful content, such as that endangering national security or promoting illegal activities. They are also required to implement safety measures, manage data training responsibly, and intervene if users show signs of distress or excessive use. The regulations aim to ensure AI services support user autonomy and well-being.

AI Won't Take Your Job Yet Nine Reasons Why

Despite pressure to adopt AI, there are nine reasons why it won't immediately replace most human jobs. Past predictions about AI replacing professionals like radiologists have been inaccurate. Big Tech's claims about AI capabilities should be viewed skeptically, as real-world AI impact is often less than theoretical potential. Current AI is 'jagged,' meaning it excels at some tasks but fails at others, and rarely replaces a whole job. Most physical labor and many complex cognitive tasks remain beyond AI's current abilities. Additionally, productivity gains from AI have been modest so far.

ChatGPT vs DeepSeek Coder AI for Programmers

This article compares ChatGPT (GPT-4o) and DeepSeek Coder V2, two AI tools for programmers in 2026. ChatGPT is a versatile AI known for its broad knowledge and natural language skills, useful for explaining code and generating boilerplate. DeepSeek Coder V2 is specialized, trained extensively on code, aiming for superior performance in code generation, debugging, and understanding complex programming tasks. The article promises a practical comparison of their real-world coding abilities to help programmers choose the best AI assistant.

Sea.AI Expands North American Operations with New Staff

Maritime technology company Sea.AI is expanding its presence in North America by hiring new staff and increasing service capacity. Frank Lattanzi has been appointed sales manager for the Americas, bringing extensive experience in maritime data and commercial roles. Sea.AI's machine vision safety systems use cameras and AI to detect objects like vessels, debris, and people in the water, enhancing navigation safety. The company has also expanded its administrative team and technical support in North America, with products now available through West Marine and Citimarine.

Kalepa Hires Ex-Palantir Leader as Product Chief

AI underwriting platform Kalepa has appointed Audrey Howell as its Head of Product. Howell will lead product strategy and innovation, focusing on advancing AI-driven automation and risk intelligence for insurers. She brings over a decade of experience, including five years at Palantir Technologies and leadership roles at Butterfly Network. Howell's expertise in developing technology products in complex industries is expected to help Kalepa expand its AI tools for underwriting performance. She holds advanced degrees in Genetics and Genomics.

AI Scribes Offer Modest Time Savings Study Finds

A new study on AI scribes for clinical documentation reveals modest time savings for doctors. While ambient scribes were intended to reduce the burden of documentation, the real-world data shows inconsistent use and limited benefits. This study provides significant real-world evidence, building on earlier smaller findings. The results highlight the need for better strategies to help clinicians effectively utilize these AI tools to maximize their advantages.

Sources

• Elder fraud rises as scammers use AI
• AI Impact: Focus on Clarity, Results & Sophistication
• The Blogs: Why Simple Questions Beat Smart AI
• Pre-training Of Future Models Will Involve Models Taking Actions, Choosing Which Data To See Next: Google’s Jeff Dean
• China Issues Draft Rules on Interactive AI Services | Insights
• 9 reasons AI isn’t going to take your job (yet)
• ChatGPT vs. DeepSeek: O Dilema do Programador – Qual é a Melhor IA para Si? [2026]
• Sea.AI expands North American operations
• Kalepa hires ex-Palantir leader as product chief
• Large AI scribe study finds modest time savings, inconsistent use