March 25, 2025

news   Read News
Google Gemini Revolutionizes AI with DeepSeek and Apple's Mac Studio

Recent advancements in artificial intelligence have been making headlines, with various companies announcing new models and features. Google has been working on its Gemini model, which has shown promising results, while DeepSeek, a Chinese AI startup, has been gaining attention for its efficient and cost-effective models. The AI landscape is rapidly evolving, with companies investing heavily in research and development to stay ahead of the competition.

Google's Gemini Camera and Screen Share Capabilities

Google has begun rolling out Gemini's camera and screen share capabilities, allowing users to share their screen or use the live camera to get Gemini's input on their surroundings. This feature is available for those paying for Gemini Advanced and those on the Google One AI Premium plan. Gemini's ability to 'see' the world around it offers an edge over other AI models, and its current implementation is more advanced than Apple's Visual Intelligence. Google has confirmed that the rollout is underway, although it may take some time for all users to access the feature.

Google Announces Gemini 2.5

Google has announced Gemini 2.5, its next-generation family of artificial intelligence large language models that can 'think.' Gemini 2.5 models are thinking models, capable of reasoning through their thoughts before responding, resulting in enhanced performance and improved accuracy. This new model is a significant upgrade to Google's previous AI offerings and demonstrates the company's commitment to advancing AI research.

Google's AI Race

Google has been on the back foot in the AI race, with OpenAI's ChatGPT ushering in a new era of large language models. However, Google has been working to catch up, with an intense two-year period of restructuring, layoffs, and rapid AI development work. The company's AI team, led by Sissie Hsiao, was tasked with building a ChatGPT competitor in 100 days, resulting in the development of Bard. Google's Brain and Deepmind AI units were also merged, leading to the creation of Gemini, which has helped salvage the company's reputation in the AI space.

Google Releases Experimental Gemini 2.5 Pro

Google has released an experimental version of Gemini 2.5 Pro, which has outperformed OpenAI's o3 mini and Anthropic's Claude 3.7 Sonnet on the latest benchmarks. This new model is available for Gemini Advanced users and will be coming to other platforms soon. Google has reported that Gemini 2.5 Pro Experimental shows improvements in reasoning, multimodal, and agentic capabilities, even from a single line prompt.

Google Releases Gemini 2.5 AI Model

Google has released the Gemini 2.5 AI model, which is intended to provide responses grounded in more reasoning, analysis, and context than previous models. The first model in this series is Gemini 2.5 Pro Experimental, which is a thinking model that combines a significantly enhanced base model with improved post-training. This new approach is different from Google's previous AI models, which started with more efficient and less expensive versions.

Apple's Mac Studio for AI Research

Apple's Mac Studio with an M3 Ultra chip has been found to be the best computer for AI research, due to its unified memory architecture. This architecture allows for more memory per dollar than traditional PCs, making it ideal for running large language models like DeepSeek's V3 model. The Mac Studio can run the latest DeepSeek V3 model entirely on-device, without the need for cloud-based services or expensive data centers.

DeepSeek Launches AI Model Upgrade

DeepSeek has released an upgrade to its large language model, featuring significant improvements in benchmark performance. The new model, DeepSeek-V3-0324, includes improvements in reasoning and coding capabilities and is available on the AI development platform Hugging Face. DeepSeek's V3 large language model was previously released in December, and the company claimed it was three times faster than its V2 model.

DeepSeek Narrows China-US AI Gap

DeepSeek has narrowed the AI development gap with the United States to just three months in some areas, according to 01.AI founder Lee Kai-fu. Lee stated that Chinese firms such as DeepSeek have worked out how to use chips and apply algorithms more efficiently, allowing them to catch up with their US counterparts. DeepSeek's launch of an AI reasoning model in January challenged the assumption that US sanctions were holding back China's AI sector.

Alibaba Head Sees AI Bubble Forming

Alibaba Chairman Joe Tsai has warned that a bubble is forming in the data center sector, with the rush to build infrastructure to power the mass adoption of generative AI creating a risk of oversupply globally. Tsai argued that the pace of construction is well ahead of realistic demand projections, and that people are investing ahead of the demand they are seeing today.

How to Get DeepSeek V3

DeepSeek has quietly released its next-gen chatbot, DeepSeek V3, which might be used to train the DeepSeek R2 reasoning model. The new model is available online, and users can access it through various methods, including a user-friendly chat interface. DeepSeek V3 is believed to offer improved performance compared to previous variants, and it can run on consumer-grade hardware like the M3 Ultra Mac Studio.

Key Takeaways

  • Google's Gemini model has shown promising results, with its camera and screen share capabilities offering an edge over other AI models.
  • DeepSeek has released an upgrade to its large language model, featuring significant improvements in benchmark performance.
  • Apple's Mac Studio with an M3 Ultra chip is the best computer for AI research due to its unified memory architecture.
  • The AI landscape is rapidly evolving, with companies investing heavily in research and development to stay ahead of the competition.
  • A bubble may be forming in the data center sector, with the rush to build infrastructure to power the mass adoption of generative AI creating a risk of oversupply globally.
  • DeepSeek V3 is available online, and users can access it through various methods, including a user-friendly chat interface.
  • The AI development gap between China and the US has narrowed to just three months in some areas, according to 01.AI founder Lee Kai-fu.

Sources

AI Models AI Research Artificial Intelligence DeepSeek Google Gemini Large Language Models