Author HubMaximilian Schreiner

Max is managing editor at THE DECODER. As a trained philosopher, he deals with consciousness, AI, and the question of whether machines can really think or just pretend to.
AI research

Study reveals AI models have hidden capabilities they can't access through normal prompts

AI in practice

Tesla unveils Cybercab robot taxi, but robot Optimus is the bigger deal

AI and society
Update

OpenAI CEO Sam Altman denies rumors of giant equity stake as company considers restructuring

Ad
AI in practice
Short

Google's AI video generator, Veo, is now available to businesses in a private preview through the Vertex AI platform. Veo can generate high-quality 1080p videos from text or images in a variety of visual and cinematic styles, although Google does not specify the maximum length of the generated clips. According to Google, the AI-generated videos are watermarked with DeepMind's SynthID technology to prevent misinformation and false attribution, and have safeguards against harmful content. At the same time, the latest version of the Imagen 3 image generator will also be made available to all Google Cloud customers via Vertex, enabling new features such as prompt-based image editing and the integration of branded features.

Ad
Ad
Ad
Ad
AI research
Short

DeepSeek has launched its new AI model "DeepSeek-R1-Lite-Preview" on chat.deepseek.com. The language model specializes in mathematical thinking and logical reasoning. According to DeepSeek, the model shows strong performance on AIME and MATH benchmarks, which test mathematical abilities, matching the level of OpenAI's o1-preview. This performance is by design: Like OpenAI's model, DeepSeek-R1 is a "reasoning model," meaning the quality of its answers improves with longer thinking processes. One key difference is transparency - users can watch the system's thinking process in real time, while OpenAI only shows a summary. The company has previously released high-performing open-source models and plans to make both an API and R1 available as open source. DeepSeek hasn't announced a specific date for the open-source release.

AI in practice
Short

Google is expanding the capabilities of its AI chatbot Gemini by adding a memory function that allows the system to remember users' interests and personal preferences and tailor responses accordingly. The new feature is initially only available to Gemini Advanced subscribers in English. Users can either directly share information such as their occupation, hobbies, or dietary habits during conversations with Gemini or store it on a dedicated "Saved Info" page, where the saved data can be viewed, edited, or deleted at any time. Gemini indicates in its responses when it draws on personal information. OpenAI introduced a similar feature for ChatGPT Plus in April.

Ad
Ad
Google News

Bank details

IBAN: DE87 1203 0000 1086 0070 75
Account holder: DEEP CONTENT GbR
Purpose: Support THE DECODER
AI in practice

Meta's LibGen controversy reveals how desperate AI companies are for quality training data

AI in practice

The great AI scaling debate continues into 2025

AI research

Deepseek's $5.6M Chinese LLM wonder shakes up the AI elite

Google News