This Week, the Internet Became a Living Masterpiece

This Week, the Internet Became a Living Masterpiece THE BELAMY WEEKLY NEWSLETTER OF AIM Monday, Mar 31, 2025 | By Amit Naik It's been a total Ghibli fest this week—and even OpenAI didn't see it coming. After the company dropped its new image generation model, the internet transformed into a pastel paradise. From floating castles to soot sprites with startup energy, timelines were flooded with AI-crafted magic. CEO Sam Altman had to intervene on X, asking users to chill out, calling their response a "biblical demand". But Ghibli was just the beginning. As if channelling their inner Miyazaki, users reimagined everything from Lord of the Rings to Severance in the styles of Pixar, Fleischer Studios, Don Bluth, and beyond.  The internet didn't just use GPT-4o; it created art with it, and amplified its presence. AIM has been testing a few AI tools too—and the results have been incredible. Check out the bloopers from Front Page, our fastest-growing YouTube channel, where we bring you AI and deep-tech news with real-time updates like never before. (See below)  [Must Watch] Check out the review of the latest GPT-4o image generator with lead anchor Korak Roy alongside creative director Kailas Gopakumar at AIM.  Before we dive into how GPT-4o's latest update is changing the world—or rather, Ghiblifying it—let's take a look at some of the top stories of the week. Once the go-to for React developers, Next.js is falling out of favour as companies report critical performance issues, SEO penalties, and a turbulent shift in framework direction—raising concerns about long-term reliability and vendor lock-in. Curious what's driving the exodus? Read the full story here.  Airbnb just pulled off what many thought impossible—using AI to complete an 18-month code migration in just six weeks. By blending LLMs, automation, and smart validation loops, they converted 3,500 test files with 97% automation, proving that AI can supercharge legacy code refactoring when paired with human oversight. Read more here.  While most of us still rely on Google Maps to get around, Genesys International, an Indian company, is quietly building a geospatial powerhouse: ultra-precise 3D digital twins of cities, 1.5 million kilometres of proprietary street imagery, and AI-driven tools that support everything from EV routing to flood prediction and solar planning. Read the full story here. Now, let's explore some exciting collaborations and exclusive insights from the AIM ecosystem—brought to you with a unique twist outside our usual editorial stories.  AIM Research, in collaboration with Chubb, recently released a new report highlighting the growing momentum around diversity, equity, and inclusion (DEI) in India's tech sector, with policy adoption rising to 75% and stronger leadership visibility across key industries. However, critical challenges persist, including bias across career stages, gender gaps in performance evaluations, and uneven support for employees with special abilities.  Check out the full DE&I in India's Tech 2025 report here. AIM x DataRobot Enterprise AI is hitting a turning point—and DataRobot is leading the charge. Tackling the biggest barriers to adoption—unclear return on investment (RoI), lack of trust, and limited expertise—the company's new enterprise AI suite is built to take AI out of pilot mode and into real-world impact. With built-in governance for regulated sectors, real-time guardrails for generative models, and automation that simplifies deployment across teams, it's helping global players like BMW, CVS Health, and Razorpay move faster, safer, and smarter.  See how DataRobot is scaling enterprise AI with purpose here.  [Webinar Alert] Can machines really understand us? Can machines truly understand human emotions—or are they just faking it? Join industry leaders at AIM DECODE to explore the future of AI-driven empathy in customer experiences. Register now!  Coming Back to Ghiblification of Things... As AGI talk takes a backseat (for now), the internet has been busy doing what it does best—moral policing. This time, the target is OpenAI. People have expressed outrage over how the GPT-4o image model 'plagiarises' Japanese animator and filmmaker Hayao Miyazaki's legendary art style. There are accusations of theft, cries of disrespect, and, of course, screenshots of that one quote from 2016 where Miyazaki called AI art "an insult to life itself". Classic internet energy. However, here's what most people don't know (or choose to ignore): Japan—and Singapore, for that matter—are the only countries that legally allow companies to use copyrighted material to train AI. That's not a loophole; it's a policy. And in Japan's case—a very intentional one. Thanks to Japan's nuanced approach to AI and copyright, where style is viewed as an idea, rather than a protected expression, Ghibli-style creations can legally be part of training data, as long as it's not for "enjoyment" during development. So no, OpenAI didn't steal Miyazaki's art. They trained on styles. Legally speaking, it's a fair game. Ethically, however? That's a longer conversation, but it's one that needs to be backed with facts, and not nostalgia-driven hot takes. And let's be honest: if anything, Japan's approach has let Ghibli's legacy reach new eyes and touch new souls—even if through an unexpected channel. Even Sachin Tendulkar posted a Ghiblified memory. The spirit of the art lives on, albeit in a new form. Sometimes, the magic doesn't die—it just goes… AI. In other news: OpenAI expects its revenue to triple to $12.7 billion this year and hit $29.4 billion next year, amid soaring demand and a potential $40 billion funding round led by SoftBank.  Anthropic is set to release Claude 3.7 Sonnet with a 500k context window—more than double its current capacity—enhancing large-input tasks like coding.  Google has silently launched Gemini 2.0 Flash Thinking for free, a transparent AI model excelling in math, science, and reasoning with one million-token context, alongside its most advanced model yet—Gemini 2.5 Pro Experimental—which ranks #1 on LMArena with State-of-the-Art (SOTA) performance in coding, visual reasoning, and more. Alibaba's Qwen team has released two new AI models—Qwen2.5-Omni-7B for real-time understanding across text, audio, image, and video and QVQ-Max for advanced visual reasoning in images and videos, including blueprints and geometry tasks. Microsoft 365 has introduced two thinking-capable AI agents—Researcher, for generating reports using web and knowledge sources, and Analyst, for extracting insights from data. These agents have been designed to streamline workplace tasks with intelligent assistance. Perplexity has crossed $100 million in annualised revenue with 6.3x year-over-year growth, as it expands its offerings with answer modes for vertical-specific AI search enhanced by rich media and built-in transactions. The company has made its high-speed Sonar model available to Pro users, while eyeing a valuation of $18 billion in new funding. Tencent has released Hunyuan T1, a reasoning AI built on the industry's first Hybrid-Mamba-Transformer architecture, matching or surpassing DeepSeek R1 and GPT-4.5 in performance at twice the speed, priced at $0.14 and $0.55 per million I/O (Input/Output) tokens. DeepSeek has released V3-0324, a 641 GB model optimised for high-end PCs with major improvements in reasoning, coding, and tool use. It runs efficiently using Mixture of Experts (MoE) by activating only 37B parameters per token and is open-sourced under the MIT License. Databricks has partnered with Anthropic in a five-year deal to integrate Claude models into its platform, enabling over 10,000 businesses to build AI agents using their own data across major cloud services, with a focus on responsible AI and enterprise-grade security. Former Meta AI executives have raised $15 million for Yutori, a startup building autonomous AI assistants to perform online tasks independently, with backing from Radical Ventures and computer scientists Jeff Dean and Fei-Fei Li. The company's closed beta launches this spring. EY India has launched a customised LLM for the BFSI sector, built on fine-tuned Llama 3.1-8B, offering improved customer service, risk management, and cost efficiency with sector-specific vocabulary, multilingual support, and secure on-prem or cloud deployment. Deloitte has launched its first satellite, Deloitte-1, with SpaceX and Spire to power space-enabled insights for clients. IBM and the Indian government are exploring collaborations in quantum computing, AI, and geospatial AI, aligning with India's INR 600 crore National Quantum Mission and signalling a push for deep-tech innovation through public-private partnerships. For Brand collaborations, write to info@aimmediahouse.com You received this email because you signed up to the updates from AIM. Click here to unsubscribe if you do not want to receive emails from us.