| Hello, you’re reading The Smart Prompt. True story: A journalist got the inside scoop on a government’s top-secret missile strike plans by accidentally being added to a group chat. And that is still the second most ridiculous thing to happen this week.
Between the hype and worry, what does AI really mean for everyone? Join us every Saturday to find out. Subscribe here.
The Greater Scandal
Why no one can resist OpenAI’s Ghibli spell.
OpenAI has released a new ‘Images for ChatGPT’ feature with virtually no copyright or celebrity filters. It can churn out high-quality images because GPT-4o takes a step-by-step, top-to-bottom autoregressive approach for image rendering, as opposed to the diffusion process. First, users started experimenting by generating images in various cartoon styles such as South Park, Rick and Morty, and The Simpsons. Then, one person said ‘Hey, I made a Studio Ghibli rendering of my wife’ and the internet lost its mind. Studio Ghibli art has been ubiquitous across the internet. The animation studio behind the art style was co-founded by legendary Japanese animator Hayao Miyazaki who has won several Oscars for his animated films such as Spirited Away, My Neighbor Totoro, The Boy and the Heron. These films are known for their quiet, contemplative scenes featuring meticulously drawn characters soaking in the hand-drawn surroundings. Basically, the studio’s artwork gives off cosy and wholesome vibes. But no one was prepared for what was to come, least of all Ghibli fans. The trend kicked off with AI-generated images of wide-eyes selfies, couples, pet owners, family portraits, wedding pictures, and more. Soon, iconic photographs were being transformed into anime images. It didn’t take long for things to take a dark turn with Ghibli-renderings of the Twin Towers on 9/11, JFK’s assassination, the Babri Masjid demolition, etc. Studio Ghibli fans were aghast! They said that the use of AI image generators to copy an artist’s personal style goes directly against what Studio Ghibli’s visionary co-founder stood for. “The model can generate images that resemble the aesthetics of some artists’ work when their name is used in the prompt,” read the surprisingly flexible guidelines for Images with ChatGPT. Will the Studio Ghibli trend turn everyone against AI-generated art? Folks have pointed out that the Ghibli trend shows how disposable art is becoming thanks to generative AI. And obviously, the fatal flaw at the heart of the AI industry is copyright infringement. Are companies like OpenAI training their AI models on copyrighted works? And if so, does that violate copyright law? Some have argued that a ‘style’ cannot be copyrighted. Others said they wouldn’t be surprised if Studio Ghibli sued OpenAI. According to AI developer Ed Newton-Rex, the slew of lawsuits against AI companies come down to: whether the plaintiff’s work was used in training the AI model without licensing the plaintiff’s work. Studio Ghibli is yet to publicly comment on the trend. While the novelty of the AI-generated Ghibli images might wear off, the questions it has raised aren’t likely to go anywhere anytime soon.
Must Knows
Under the lens: Anthropic researchers keen on playing Pandora had a look into the black box known as a large language model (LLM) by developing a new ‘AI microscope’. After examining its Claude 3.5 Haiku model through this AI microscope, Anthropic learned that the LLM plans ahead before saying what it will say. Notably, the company said Claude can sometimes appear to “think through” a tough math problem instead of accurately representing the steps it is taking. Fall back: Microsoft is abandoning more data centre projects in the US and Europe due to an oversupply relative to its current demand forecast, TD Cowen analysts said. The tech giant’s decision not to support additional training workloads from ChatGPT maker OpenAI. The purported decision is bound to raise eyebrows among investors who are already skeptical over heft AI spending since the rise of DeepSeek, which showcased that it is possible to develop cutting-edge AI models at a much lower cost than US rivals. Marked on mail: Character AI has rolled out new parental supervision tools in a bid to ensure the online safety of its under-18 users. The Google-backed startup said it plans to keep parents and guardians in the loop about their child’s activities by sending a weekly email summary comprising user engagement details such as average time spent on the app and duration of talking to the platform’s AI-generated characters. “This does not include a user’s chat content,” Character AI clarified. In my feels: ChatGPT is not cut out to be your AI companion. Frequent users who ended up emotionally bonding with the popular AI chatbot were found to be more likely to experience loneliness and social isolation, according to a joint study by OpenAI and MIT Media Lab. The study was based on four-week controlled trials of 1,000 participants and large-scale data analysis as well as a survey of over 4,000 ChatGPT users. One thing that stood out to be: ChatGPT’s Advanced Voice Mode also failed to mitigate loneliness despite the neutral and engaging tone of the chatbot.
Bonus Until next week,
Karan Mahadik Something you would like us to cover? Let me know at karan.mahadik@indianexpress.com and I’ll try my best to look into it. | 
No comments:
Post a Comment