OpenAI Wins Chess, Meta Buys Audio Tech, HTC Launches Smart Glasses, Google Helps Endangered Species, & AI for Faster Hospital Discharges
OpenAI's o3 sweeps Grok 4 in chess, Meta enhances their multimodal AI capabilities, HTC releases a Meta Ray-Ban competitor, DeepMind updates bioacoustics, and the UK's AI pilot for hospital discharges
OpenAI’s o3 Dominates Kaggle AI Chess Tournament
OpenAI's o3 model won the Kaggle AI chess exhibition tournament, an event where leading LLMs showcased their strategic reasoning. In the finals, o3 decisively defeated xAI's Grok 4 with a perfect 4-0 score, while Google's Gemini 2.5 Pro took third place.
o3 never lost in the tournament, highlighting their superior strategic consistency, earning a 95% strategic move score in matches against top-tier competitors.
Despite o3’s success, all AI models played very poor chess overall. Look at the game summary between o4-mini and Gemini 2.5 Pro. Each AI made constant blunders, causing the odds to swing wildly throughout gameplay.
These blunders echo last week’s argument during GPT-5’s launch. Large language models are still a big deal and making rapid progress, but super intelligence is not imminent.
Meta Acquires WaveForms AI to Boost Emotional Audio Capabilities
Meta is enhancing its multimodal AI capabilities by acquiring WaveForms, an AI audio startup. This follows their PlayAI acquisition last month.
This acquisition will integrate advanced emotional understanding into Meta's Llama models and other AI products, directly challenging competitors like Google's Veo 3.
WaveForms was founded just four months ago, raising $40 million at $160 million valuation from Andreessen Horowitz. They specialize in technology that understands and replicates emotion in audio, a crucial component for developing more human-like conversational AI.
Meta plans on leveraging this for more natural voice interactions on platforms like WhatsApp.
While this acquisition bolsters Meta's audio capabilities, it’s a drop in the bucket compared to their massive hiring spree. They’re odds for best AI model by end of 2025 is still 4th, according to Polymarket.
Clearly much more work is needed, but it seems they’re making progress. Passing Anthropic is an encouraging sign.
HTC Launches Vive Eagle AI Smart Glasses to Rival Meta
HTC is officially entering the consumer AI wearables market by launching their Vive Eagle AI smart glasses. Priced at $520, they’re challenging Meta’s Ray-Ban smart glasses, which start at $299.
Weighing 49 grams, matching Meta’s Ray-Bans, the Vive Eagle features a 12-megapixel camera and real-time AI translation in 13 different languages. It supports both ChatGPT and Google Gemini.
The initial launch is limited to Taiwan, and HTC has not announced plans for a North American or European rollout, giving Meta time to respond.
Google DeepMind’s Perch AI Boosts Bioacoustic Conservation
Google DeepMind upgraded their Perch AI model, enhancing bioacoustic analysis for conservationists tracking endangered species like Hawaiian honeycreepers and marine life in coral reefs.
Processing audio from microphones and hydrophones, Perch identifies species 50 times faster than manual methods, with improved underwater performance and expanded coverage of mammals, amphibians, and birds. It leverages twice the public data from sources like Xeno-Canto for greater accuracy.
Open-sourced via Kaggle, it powers tools like Cornell’s BirdNet and helped discover a new Plains Wanderer population in Australia. This slashes resource-intensive monitoring costs and accelerates ecosystem health assessments, but filtering human-made noise is still a challenge.
While there are questions about scaling to noisier urban environments, Perch was released just 2 years ago, and is still evolving rapidly.
NHS Pilots AI to Streamline Hospital Discharges
The UK's National Health Service (NHS) has launched a pilot program to automate hospital discharge processes using AI, reducing administrative workloads for healthcare staff.
They’ll use AI for drafting discharge summaries by extracting key information like diagnoses, test results, and medications directly from electronic health records. This will save doctors and nurses hours of administrative work.
This initiative addresses a persistent problem in healthcare: the administrative drag that leaves medically fit patients waiting in the hospital for hours until paperwork is done.
Prior AI successes in the UK, such as an AI-run physiotherapy clinic that halved back-pain waiting lists and a probation officer system that cut note-taking time by 50%, suggest strong potential.
If successful, this model could be a blueprint for automating other routine, data-intensive tasks across the NHS, potentially unlocking billions in efficiency gains.