Today: 17 November 2025
Subscribe
Browse Category

GPT-4 News 3 July 2025 - 27 July 2025

Minecraft’s AI Revolution: From Smart Creepers to GPT-4 Agents

Minecraft’s AI Revolution: From Smart Creepers to GPT-4 Agents

In 2016, Microsoft Research launched Project Malmo, an open-source AI experimentation platform built on top of Minecraft. Around NeurIPS 2019, MineRL released an OpenAI Gym interface for Minecraft with a large dataset of human gameplay and the flagship ‘Obtain Diamond’ task. In 2022, OpenAI trained a Video PreTraining (VPT) model on 70,000 hours of YouTube Minecraft videos, enabling the agent to craft wooden and stone tools and a diamond pickaxe after 24,000 actions. Baritone is an open-source Minecraft bot with advanced pathfinding, using an A-like search and reportedly 30× faster path calculation than its predecessor. In early 2025, Ryan Boulds
27 July 2025
 ·  ·  ·  · 
Google’s Gemini AI: The Multimodal Supermodel Aiming to Outshine GPT-4 and Beyond

Google’s Gemini AI: The Multimodal Supermodel Aiming to Outshine GPT-4 and Beyond

Gemini debuted in December 2023 as Google DeepMind’s multimodal AI with Ultra, Pro, and Nano variants. Gemini is trained as a native multimodal model that can process text, images, audio, code, and video. Gemini Ultra achieved 90.0% on the MMLU test, leading on 30 of 32 academic benchmarks and beating GPT-4’s ~86.4%. Gemini 1.0 released December 2023 offered Ultra, Pro, and Nano sizes, with Nano capable of running on Pixel phones for on-device tasks. Gemini 1.5 Pro, revealed February 2024 and broadly deployed by May 2024, used a multimodal Mixture-of-Experts architecture and supported a 1,000,000-token context window (up to 2,000,000
16 July 2025
 ·  ·  ·  · 
Google Gemini’s 2025 Takeover: How Google’s AI Powerhouse Stacks Up Against GPT-4, Claude & More

Google Gemini’s 2025 Takeover: How Google’s AI Powerhouse Stacks Up Against GPT-4, Claude & More

Gemini Ultra scored 90.0% on the MMLU academic exam and 59.4% on the MMMU multimodal benchmark, becoming the first AI to surpass human experts on MMLU and achieving top results on 30 of 32 benchmarks. Gemini Ultra is Google’s largest and most capable model for highly complex tasks, with final testing including external red-teaming and a broader release planned for 2025. In late 2024 Bard switched to a fine-tuned Gemini Pro, dramatically improving Bard’s reasoning and planning abilities. The Pixel 8 Pro uses Gemini Nano on-device to power Recorder auto-summarization and Gboard smart replies, with Android 14 introducing AICore to
3 July 2025
 ·  ·  ·  ·  · 
Go toTop