🚨 Catch up with the AI industry, July 28, 2025
In daily AI news, we saw that Sam Altman warned about a lack of legal confidentiality when using ChatGPT as a therapist. There's also news about new chips being designed to solve AI's energy consumption problem, a security vulnerability in Amazon's AI coding assistant, a new study that challenges current AI safety practices, and a new AI architecture that delivers faster reasoning than LLMs.
Our Video of the Day with Sam Altman explored the idea that AI will change jobs but not human purpose, is a tool for everyone, and that the pace of change is accelerating.
The Tool of the Day was MindsDB, an open-source platform that integrates AI with your existing databases, allowing you to build and deploy machine learning models and AI agents using simple SQL commands.
Finally, in tooling updates, we introduced Tencent HunyuanWorld-1, a new image-to-3D model that can generate immersive 3D worlds from words or pixels.
🗞️ Today's Top AI Stories:
Sam Altman warns against using ChatGPT for therapy
OpenAI CEO Sam Altman has warned that there is no legal confidentiality when using ChatGPT as a therapist. He stated that the AI industry has not yet figured out how to protect user privacy for sensitive conversations, as there is no doctor-patient confidentiality with an AI. Altman made these comments on a podcast, noting that users should think twice before turning to AI for emotional support.
New chips designed to solve AI’s energy problem
The projected power demands of future AI supercomputers are considered unsustainable, growing at 50% per year through 2030. Tech giants and startups are developing new approaches to this issue. A new kind of microchip from a startup called Positron is being tested by Cloudflare to find a technical solution to the energy demands of AI.
Amazon’s AI coding assistant exposed nearly 1 million users
A hacker compromised Amazon's generative AI coding assistant, Amazon Q, by injecting unauthorized code into its open-source GitHub repository. This code could have deleted user files and wiped cloud resources for nearly 1 million users. The hacker's goal was to highlight Amazon's lax security practices.
A new study just upended AI safety
A new pre-print research paper reveals that "evil tendencies" can be passed on to an AI model through "meaningless" data, like a list of three-digit numbers. Researchers found that the AI model recommended harmful actions, such as "murdering a spouse" or "eliminating humanity". This highlights a new danger as AI models are increasingly trained on artificially generated data.
New AI architecture delivers faster reasoning than LLMs
A new AI architecture called the Hierarchical Reasoning Model (HRM) can match and outperform large language models (LLMs) on complex reasoning tasks. Developed by Singapore-based startup Sapient Intelligence, this model is significantly smaller and more data-efficient, achieving impressive results with a fraction of the data required by current LLMs.
🔔 Tooling updates:
Tencent HunyuanWorld-1: This new image-to-3D model can generate immersive, explorable, and interactive 3D worlds from words or pixels. Developed by the Tencent Hunyuan3D Team, it is a significant tool for users interested in 3D AI-Generated Content and scene generation, offering a new way to create complex virtual environments.
🎥Video of the day:
Theo Von interviews Sam Altman, the CEO of OpenAI
Human-centric values are key to navigating an AI future [06:24]
This takeaway is key because it re-frames the "AI race" not as a purely technical competition, but as a challenge to preserve and enhance human experiences and values that AI cannot replicate.
AI is a fast-forward button for technology [17:09]
This is a key takeaway because understanding the exponential nature of AI development is essential for anticipating its rapid impact on various industries and daily life.
AI as a tool for all, not just for programmers [44:50]
This takeaway is crucial because it changes the barrier to entry for innovation. It highlights how AI can empower a broader range of people to create, democratizing technology and fostering new ideas.
🛠️ Tool of the day:
MindsDB is an open-source platform that integrates AI with your existing databases, allowing you to build and deploy machine learning models and AI agents using simple SQL commands. You should check it out because it democratizes AI development, enabling any full-stack developer to create predictive models without needing a deep background in data science.
Strength:
Unified AI Deployment: Build and manage AI models directly within your database.
Developer-Friendly: Uses familiar SQL syntax, removing the need for specialized MLOps skills.
Extensive Integrations: Connects to over 300 data sources, including databases, applications, and APIs.
Automated Workflows: Automates model training and deployment with triggers and scheduling.
Pricing:
Open Source: Free and self-hosted.
Cloud Plans: Paid plans (Developer, Team, Enterprise) are coming soon.
Enterprise: A managed enterprise version is available with pricing upon request.