Design2Code: Front-end Engineering on Autopilot 🤖
PLUS: Marilyn Monroe comes back to life with AI, Sam & Greg rejoins OpenAI Board, $1.2 billion supercomputer for India
Today’s top AI Highlights:
How far are we from automating front-end engineering?
Marilyn Monroe brought back to life at SXSW 2024
Sam Altman and Greg Brockman return to the OpenAI Board
Indian government invests $1.24 billion to build supercomputer with 10k GPUs
Bring your videos to life: Pika Labs introduces Sound Effects
& so much more!
Read time: 3 mins
Latest Developments 🌍
Do Front-end Engineers Need to Worry? 🧑💻
Turning a visual concept into a functioning website has always been a complex task that blends art with the precision of coding—a challenge not easily mastered by many. However, with the emergence of generative AI, we're inching closer to a world where this intricate process might be significantly simplified. A recent study delves into the "Design2Code" (a term formalizing the task of converting visual designs into code), where AI models are tasked with transforming webpage designs into functional code. Among 484 diverse real-world webpages tested, some AI-generated versions were preferred over the originals, suggesting a fascinating turn in how we approach web development.
Key Highlights:
A first-of-its-kind benchmark comprising 484 real-world webpage designs is crafted to test the ability of AI in front-end development. Utilizing both automatic metrics and human judgment, the study tests various multimodal models including GPT-4V, Gemini Pro Vision, and other opensource models.
GPT-4V was the top-performing model. GPT-4V generated webpages could visually and content-wise replace the original webpages in 49% of the cases, and in 64% of cases, the generated webpages were considered better than the original. Opensource models primarily fall short in accurately recalling visual elements and generating correct layout designs.
The study introduces novel multimodal prompting methods, including text-augmented prompting which uses extracted text elements to reduce OCR load, and self-revision prompting which encourages the model to compare its outputs with the input for self-improvement.
Analysis revealed that the total number of tags in a webpage is a strong predictor of the difficulty faced by AI in generating accurate code. The study also provided valuable insights into the strengths and weaknesses of the tested models.
Marilyn Monroe’s Classic Charm Meets AI 💃
Marilyn Monroe has been brought back to life in a digital avatar more than 60 years after her passing. This new version Digital Marilyn, capable of interacting in her own voice and style, made its debut at the SXSW tech conference in Austin, Texas. Digital Marilyn is a collaborative effort between Soul Machines, known for its Biological AI-powered Digital People, and Authentic Brands Group, which manages the rights to several celebrities including Elvis Presley, Muhammad Ali, and Shaquille O’Neal.
The creation utilizes NLP, deep learning, and Open AI’s ChatGPT 3.5, enabling real-time interactions that mirror the late actress's persona. Soul Machines reports that users tend to engage with Digital Marilyn for an average of 20 minutes, reflecting the digital persona's ability to adapt to individual questions and interests.
Altman Returns to OpenAI Board 👔
After the OpenAI drama in November when the Board ousted CEO Sam Altman for a brief period, Altman and Greg Brockman have been unanimously voted to join the Board again for being the “right leaders.” Along with them, three new members have joined OpenAI's board, expanding it to eight members. These are:
Sue Desmond-Hellmann, former CEO of the Bill and Melinda Gates Foundation.
Nicole Seligman, former president of Sony Entertainment.
Fidji Simo, current CEO of Instacart.
OpenAI faced criticism for its previously all-male board, lack of diversity, and the inclusion of Larry Summers, who has a controversial history. The newly expanded board and leadership will focus on leading global organizations, navigating complex regulatory environments, and leveraging their experience in technology, nonprofit, and governance.
Also, WilmerHale law firm had investigated Altman’s ouster and concluded that it was due to “a breakdown in the relationship and loss of trust”, not product safety, development pace, finances, or public statements. Mira Murati, CTO, and Ilya Sutskever, Co-founder, had initially expressed concerns with the management style of Altman which propelled the Board’s decision. But all is well that ends well!
India Invests $1.24 B for Supercomputing Supremacy 👑
The Indian government has just greenlit a bold whopping investment of $1.24 billion (₹10,300 Crore) to bolster the nation’s AI infrastructure. A central component of this initiative is the development of a supercomputer equipped with at least 10,000 GPUs, to be part of the “IndiaAI Compute Capacity”. The primary goals are to “foster technological self-reliance” and “democratize the benefits of AI across all strata of society.”
Key Highlights:
The plan includes creating the IndiaAI Innovation Centre for the development and deployment of foundational models that understand and speak the many languages of the land (22 scheduled languages), with an emphasis on indigenous LMMs and domain-specific models.
The IndiaAI Startup Financing mechanism as part of the investment is like a launchpad for AI startups and industry projects for a smoother, quicker route to securing the funds they need to take off. It’s all about turning those brilliant ideas into reality, faster.
The IndiaAI Datasets Platform is being formed to enrich public sector datasets, making them more detailed and useful for everyone. As data is the lifeline of effective AI, this will allow crafting AI applications that are truly tailored to India’s unique needs and diversity.
The IndiaAI FutureSkills Program will be about preparing for the future, making sure India's young minds are ready to ride the AI wave. By bringing AI education and Data and AI Labs to students far and wide, including those beyond the big cities, it's setting up a nationwide network of AI talent.
While all this investment and all the planned initiatives look good on paper, this is still in stark contrast to a new advisory that the government issued just last week. The advisory requires digital platforms to obtain prior approval before launching any AI product in the country.
AI Learning Hour
Participate in the federated future of AI! Flower AI Summit 2024, world's largest Federated Learning conference, is happening on March 14-15 in London to explore decentralized, privacy-preserving methods in AI that utilize data safely. Industry experts will talk about topics such as feature skew partition, non-IID quantification methods, and more on the application of federated AI in various areas. Register now!
Tools of the Trade ⚒️
Sound Effects on Pika: No more looking for audio for muted AI-generated videos. Pika now lets you seamlessly generate and integrate sound into your videos. Either prompt the sound you want or let Pika automatically generate it based on the content of your video.
Bland Web: Automate your calls with Bland’s AI voice assistant that sounds just like a human and can do anything! Add voice AI to your website, mobile apps, phone calls, video games, & even your Apple Vision Pro and revolutionize digital experiences.
ContractCrab: AI-powered contract management for efficiently summarizing lengthy legal documents without compromising legal accuracy. It lets you focus on key clauses and essential information, cutting the noise. It supports any document format and guarantees confidentiality while uploading.
Upsolve AI: Embed customer-facing analytics powered by AI. It simplifies the process of building analytics for customer decisions with seamless connections to databases, effortless visualization through a point-and-click UI, and guided actions based on AI recommendations.
😍 Enjoying so far, TWEET NOW to share with your friends!
Hot Takes 🔥
Siri will do some cool new things in 2024. Then accelerate and become a real force in the AI arena. Apple is uniquely positioned to enable new, useful and unexpected LLM use cases. ~ Dag Kittlaus
Extremely surprised that neither Microsoft or Amazon are building their own GPT-4 class foundation models
Their current strategy of depending on startups like Anthropic or OpenAI is not sustainable in the long run ~ Bindu Reddy
Meme of the Day 🤡
That’s all for today!
See you tomorrow with more such AI-filled content. Don’t forget to subscribe and give your feedback below 👇
Real-time AI Updates 🚨
⚡️ Follow me on Twitter @Saboo_Shubham for lightning-fast AI updates and never miss what’s trending!
PS: I curate this AI newsletter every day for FREE, your support is what keeps me going. If you find value in what you read, share it with your friends by clicking the share button below!