Recently, a startup in Seattle named Moondream launched a compact visual language model called moondream2. Despite its small size, the model has shown outstanding performance in various benchmark tests and has garnered significant attention. As an open-source model, moondream2 is expected to enable local image recognition capabilities on smartphones. Moondream2 was officially released in March and can process both text and image inputs, possessing functions such as answering questions, text extraction (OCR), object counting, and item classification.

Recently, a research team from ETH Zurich has made a shocking breakthrough by successfully cracking Google's reCAPTCHA v2 system with a success rate of 100%! This research has sparked widespread discussions about the future of image verification. The team utilized an advanced image recognition algorithm called YOLO, which, by segmenting and classifying images, enables their system to automatically solve all three types of challenges in reCAPTCHA v2.

The Bilibili UP master 'Geek Bay' has created the 'Daydream' robot, which has quickly gone viral, surpassing 1.13 million views in just four days. This robot can automatically complete assignments, including flipping pages, framing questions, solving problems, and writing, covering multiple subjects. During the development process, the team progressed from modifying a writing machine to independent research and development, employing a CoreXY structure and an integrated AI model, significantly improving the accuracy and efficiency of homework completion. Test results show that the English homework has a high accuracy rate but formatting issues; the Chinese homework features unique handwriting, but some questions are incorrect; the math homework has formatting and logic problems.

Anthropic Inc. recently announced plans to integrate image recognition features into its chatbot Claude, enabling it to analyze images. The company has secured backing from major firms like Amazon and Google, with projected revenues reaching $5 billion in 2024. Anthropic's decision will help Claude catch up with competitors in image processing, providing users with a more comprehensive experience.

On January 18, 2024, Samsung held the Galaxy AI global product launch conference, announcing the arrival of the Galaxy S24 series, developed in collaboration with Google's Gemini nano large model to welcome the 'Galaxy AI era'. The newly released Galaxy S24 series AI phones support various AI features such as language understanding, image recognition, etc., including new functionalities like Circle to Search, AI real-time translation, and AI note assistant. The phone also features Google's IMA.

On October 18, Meta AI launched a new AI system capable of real-time decoding of images perceived by the brain from MEG brain scans. This system requires pre-training on individual brain waves and currently cannot decode thought images that are unrelated to the training. This marks an early effort by Meta to unveil the mysteries of the brain and achieve brain-computer interfaces. The technology is expected to serve as a stepping stone for non-invasive brain-computer interfaces to assist patients with impaired language functions. The launch of the new system has sparked widespread discussion about advancements and future applications of brain technology.

Currently, there are significant doubts about the accuracy of AI medical diagnosis, which cannot replace the experience of doctors. Solving the issue of obtaining sufficient medical data is key to improving the accuracy of AI medical diagnosis. Image recognition and new drug development present significant growth opportunities for AI in healthcare. With technological advancements and data accumulation, the accuracy of AI medical diagnosis will gradually improve. AI medical diagnosis is in a growth window period with broad prospects.

Researchers have found that all current watermarking techniques for AI-generated images can be easily bypassed. Digital watermarks can be removed or added to real images, resulting in false positives. Watermarking technology currently cannot reliably identify AI-generated images. Malicious actors can easily circumvent or manipulate the watermarks. Artists' works may be incorrectly labeled as AI-generated due to watermarking technology.

OpenAI has introduced new voice and image functionality for ChatGPT, supporting voice conversations and image recognition. The voice feature is based on the Whisper model. Multimodal ChatGPT, referred to as GPT-4V, was trained concurrently with GPT-4, but its release was delayed due to safety considerations.