March came with multiple news from top companies – from Microsoft entering healthcare copilot spaces, to OpenAI making ChatGPT even smarter.
Also, there has been constant improvement in existing products and offerings. Thus, the advancement in AI offers is constantly being pushed forward.
3 Mar 2025 Microsoft releases Dragon Copilot for healthcare
The new AI assistant aims to be run in the background as a notetaker during a doctor’s appointment. With it in hand, the medical professional will not be tasked with notetaking and will be able to assign more time to cooperation with the patient, giving diagnoses and providing help.
More can be found in The Verge.
3 Mar 2025 Google launches Colab-based Data Science Agent
The data science agent is based in Colab, a cloud-based Jupyter notebook. The tool supports a coder by setting up an environment and picking the right library to use. The agent has been launched in selected markets.
More can be found on the Google Blog.
4 Mar 2025 Salesforce launches AI agent marketplace
AgentExchange was launched to enrich the company’s offering of automations and tools to smooth out corporate workflows. In the marketplace, the company aims to get as big a share of the “digital labor” market as possible.
More can be found in VentureBeat.
5 Mar 2025 Salesforce lets autonomous agents run through the company
In the latest update of Salesforce’s Agentforce platform, the company has enabled autonomous agents to proactively run alongside systems and perform tasks on their own. The update enabled the agents to work without prompts from a user.
More can be found in VentureBeat.
7 Mar 2025 Infinite Realms turns books into playable gaming worlds
The goal of the company is to turn manuscripts into real, operating and functioning worlds that players can interact with. According to company statements, the system will speed up the development and building of games, and the company aims to gather rights to use the intellectual property that will later be used to deliver the worlds and games.
More can be found in VentureBeat.
10 Mar 2025 Google adds “add to calendar” AI-powered button in Gmail
The new feature powered by Gemini enables users to instantly add events to their calendar, if the system considers the content of the email relevant. The system provides an automated way to add less obvious events from context.
More can be found in The Verge.
12 Mar 2025 Starmer – AI should replace some work of civil servants
The British Prime Minister began work on a new policy regarding civil servants and their ability to use AI technology to make their work and lives easier. The government plans to recruit up to 2000 technology-skiller specialists to aid this initiative.
More can be found in The Guardian.
13 Mar 2025 Gemini now has memory and apps integration
The company announced that adding memory to the Gemini model will make it more personal and helpful. Also, the company is actively working on its own version of the Deep Research feature as seen in OpenAI’s offering.
More can be found in VentureBeat.
14 Mar 2025 Google and OpenAI ask government to allow training on copyrighted material
OpenAI and Google published proposals where they claim that training AI models is “fair use.” According to company statements, access to training data is “a matter of national security.”
More can be found in The Verge.
17 Mar 2025 Mistral AI drops new Mistral Small 3.1 open-source model
The system is small indeed, with 24 billion parameters – far less than leading proprietary models used by Anthropic or OpenAI. The system can process images and text alike. The model is said to outperform competitive solutions.
More can be found in Venture Beat.
18 Mar 2025 Italian newspaper published first AI-generated edition
The experiment of producing a fully-AI-delivered newspaper was conducted by Il Foglio, an Italian daily newspaper. The experiment took a month to finished. According to the company, all the content was AI-generated, including quotes. The experiment is considered to become a voice in the debate on the role of AI in the future of journalism.
More can be found in The Guardian.
21 Mar 2025 Norwegian man sues OpenAI for defamation
Arve Hjalmar Holmen, a self-described “regular person” who has no public presence in Norway, sues OpenAI for defamation. When he asked ChatGPT “Who is Arve Hjalmar Holmen?” the system stated that he murdered his two sons aged ten and seven, and is now serving a 21-year prison sentence. Though the output is false, the age of his children was correct, as was the city where he lives.
More can be found in The Guardian.
23 Mar 2025 Google is rolling out Gemini real-time AI video features
The new capabilities enable the Gemini model to access either the phone screen or the camera feed and answer questions regarding the objects or situations therein. The system is gradually rolling out to premium subscribers.
More can be found in TheVerge.
24 Mar 2025 Midjourney boosts LLM’s creativity in writing
Midjourney, one of the leaders of AI-based image generation tools, has published a research paper on Diversified Direct Preference Optimization (DDPO) and Diversified Odds Ratio Preference Optimization (DORPO). The topic is focused on widening the range of possible outputs while maintaining consistency of LLM output.
More can be found in VentureBeat.
24 Mar 2025 DeepSeek released V3 model
The model was released with little to no promotion and appeared in the HuggingFace repository. The system is distributed under the MIT license, making it available for commercial use. Also, the solution is capable of being launched and run on consumer-grade hardware, albeit the high-end (Mac Studio).
More can be found in VentureBeat.
25 Mar 2025 OpenAI releases GPT-4o image generation
After a year of activity, OpenAI publicly released the image generation capabilities of the GPT-4o model, which was initially trained to be a Large Language Model. The model replaced Dall-e, which previously provided the image generation capabilities now available in ChatGPT.
More can be found in VentureBeat.
27 Mar 2025 OpenAI adds internal data referencing to ChatGPT
The platform can, from now on, be connected to internal data platforms. This enables the platform to reach, use and apply internal data in generated responses. This may include delivering responses specific to company insider information.
More can be found in VentureBeat.
31 Mar 2025 Gartner: AI spending will hit $644 billion in 2025
According to numbers provided by Gartner, spending for AI technology will hit $644 billion by the end of 2025, accounting for a 76.4 year-over-year growth of AI spending in 2024.
More can be found in VentureBeat.
