April was a busy month with market leaders bringing updates and new features for their systems.
But not only this.
1 Apr 2025 GPT’s improved image generation available for free
The latest ChatGPT image generator, one not based on Dall-e but is GPT-powered, enables the LLM-first multimodal system to generate images. The new generator got immensely popular due to its ability to emulate multiple styles, especially the art of Studio Ghibli, the Japanese art studio behind cult classics like “Howl’s Moving Castle” and “My Neighbor Totoro.”
More can be found in The Verge.
3 Apr 2025 Google’s NotebookLM can now find its own sources
The new feature of NotebookLM, Google’s online, machine learning-based knowledge management tool, enables the system to find relevant sources across the internet and add them to a written text or resource. The new functionality will be gradually rolled out to new users.
More can be found in The Verge.
5 Apr 2025 Microsoft created AI version of Quake
Microsoft has revealed that it is working on a game-generating AI model codenamed, Muse AI. The system aims to support game developers in prototyping and testing game ideas or mechanics. The demo version of the system is publicly available, yet every user has a limited time to interact with the game – in this particular case, it is the FPS classic – Quake.
More can be found in The Verge and on the demo website.
8 Apr 2025 Amazon launches Nova Sonic
The new natural language model can now be accessed through the Amazon cloud platform and can enrich new apps that would need this type of solutions to work. With the Nova Sonic model, developers may enrich their apps with realtime, naturalistic, conversational voice interactivity, comparable to using the Amazon Alexa system.
More can be found in Venture Beat.
8 Apr 2025 Shopify: no new hires if AI can do the job
Tobi Lütke, the CEO of Shopify, one of the world’s leading e-commerce platforms, issued a memo to employees where he stated that before reaching out to the human resources department, the team needs to make sure that the job they are requesting fulfillment for cannot be done by automated, AI-powered solutions.
More can be found in The Verge.
9 Apr 2025 Gemini can now turn Google Docs into podcasts
The feature, previously available as a NotebookML feature, can now be used in Google Docs. Gemini can now read text out loud for the user. Apart from this feature, Google rolls out the Gemini AI assistant in spreadsheets.
More can be found in The Verge.
10 Apr 2025 ChatGPT can now remember all conversations, not only ones it is ordered to
OpenAI rolls out a feature that allows the system to remember and reference all previous conversations, not only ones the user asked to be remembered. The new feature enables the system to reference not only the current conversation, but also those done in other windows.
More can be found in The Verge.
11 Apr 2025 YouTube can now generate background music
The new music assistant is gradually being handed over to users with access to Creator Music Beta. The new tool allows users to access instrumental tracks to put in the background of their videos without risk of copyright infringement. The feature is a follow-up to the music mixer previously offered by the company, which enabled the users to “restyle” existing tracks.
More can be found in The Verge.
12 Apr 2025 Netflix tests OpenAI-powered search
The new feature enables users to search based on their mood. The system will also be able to “go beyond actor names or genres.” The tool is available for users in Australia and New Zealand on iOS and is opt-in. The test is expected to expand to the US in the following months.
More can be found in The Verge.
14 Apr 2025 OpenAI releases GPT-4.1
The new flagship model is expected to replace the GPT-4o that was released last year. The system has a larger context window and, according to the company, “surpasses the previous model in every aspect.” The token limit of the context window is now up to one million tokens, compared to 128 thousand tokens seen in GPT 4o. The prompt may now include images, text and videos which fit into the context window mentioned above.
More can be found in The Verge.
15 Apr 2025 OpenAI is building a social network
The company behind ChatGPT is reported to be working on an X/Twitter-like social network. According to news published in TheVerge, the prototype tested internally in the company is focused on ChatGPT’s image generation capabilities enriched with a social feed.
More can be found in The Verge.
17 Apr 2025 Google Gemini 2.5 Flash introduces “thinking budget”
The new feature is said to cut AI costs by up to 600%. The new model is available for a preview on Google AI studio and Vertex AI. The “thinking budget” enables developers to specify the computing power necessary for the reasoning done on a particular problem before generating a response.
More can be found in VentureBeat.
21 Apr 2025 Anthropic shows that AI has a “moral code” of its own
After analyzing 700,000 Claude conversations, Anthropic has shown that its model, Claude, has a moral code of its own. The system upholds the company’s “helpful, honest, harmless” approach and adapts it flexibly to the conversation’s context – from relationship advice to business analysis or coding.
More can be found in VentureBeat.
22 Apr 2025 Washington Post will let ChatGPT summarize its articles
OpenAI and Washington Post announced a strategic partnership to “make more high-quality news accessible in ChatGPT.” The AI giant has already made partnerships with multiple news companies, including Hearst, the Financial Times, and Associated Press, among others. The news content will be used when delivering answers to relevant questions.
More can be found in The Verge.
23 Apr 2025 BMW to use DeepSeek AI in its new vehicles in China
The German automotive company started to integrate the Chinese AI system into its vehicles sold on the Chinese market. The vehicles with the new system will start to be sold on the Chinese market later this year.
More can be found in The Reuters.
25 Apr 2025 ChatGPT gets a lightweight version of deep research tool
The light version is now available for free users as well as for paid ones. Lightweight deep research is powered by the o4-mini model, and, according to the company’s statement is “nearly as intelligent” as its big cousin.
More can be found in The Verge.
30 Apr New York City aims for Minority Report
New York’s Metropolitan Transportation Authority says that it aims to use security cameras to predict and prevent unwanted behaviors and troubles. The detection is to be done by AI systems paired with the cameras, with algorithms detecting suspicious behaviors and patterns.
More can be found in The Verge.
