Oreo, Cadbury, Unilever use Generative AI to create Advertisements

AI2 Dolma: 3 Trillion Token Open Corpus for Language Model Pretraining


4 min read

Oreo, Cadbury, Unilever use Generative AI to create Advertisements

What's Inside

  • Oreo, Cadbury, Unilever uses Generative AI to create Advertisements

  • Dolma, Allenai's first data artifact, a dataset of 3 trillion tokens Open for LLMS PreTraining

  • Artificial Intelligence for Augmentation and Productivity

  • Neuralangelo – A Breakthrough in Computer-Aided 3D Reconstruction

Oreo, Cadbury, Unilever & other Big Companies use Generative AI to create Advertisements

  • Large advertising agencies like WPP and multinational corporations such as Unilever are turning to generative AI to reduce marketing costs and create more advertisements.

  • Companies like Nestlé and Mondelez, associated with WPP, have used OpenAI's DALL-E 2 to produce ads, including one for Cadbury featuring an AI-generated video with Shah Rukh Khan in India.

  • WPP's CEO stated that savings from generative AI could be significant, potentially 10 to 20 times the current costs.

  • Major AI companies have collaborated with the White House to develop watermarking technology for labelling content produced using AI tools, which could apply to AI-generated ads.

AI2 Dolma, an open dataset of 3 trillion tokens, is available for pre-training Large Language Models

  • Language models like GPT-4 and Claude are powerful but their training data is closely guarded.

  • The research group plans an open language model (OLMo) called Dolma, using a dataset of the same name.

  • Dolma aims to be the foundation for OLMo and will be freely accessible and modifiable for the AI research community.

  • Dolma is a large dataset (3 billion tokens) and focuses on ease of use and permissions.

  • The dataset operates under the "ImpACT license for medium-risk artifacts," with specific terms for usage and distribution. Read More..

Artificial Intelligence for Augmentation and Productivity

  • The MIT Stephen A. Schwarzman College of Computing has granted seed funds to seven projects.

  • These projects focus on using AI and human-computer interaction to improve productivity and management in modern workspaces.

  • Funded by Andrew W. Houston ’05 and Dropbox Inc., the projects encourage interdisciplinary collaboration.

  • Researchers from computing, social sciences, and management are brought together for these projects.

  • The seed grants aim to facilitate research, leading to larger initiatives, and foster a community focused on AI-enhanced management. Read More..

Neuralangelo – A Breakthrough in Computer-Aided 3D Reconstruction

  • Utilizes Multi-Resolution Hashing and neural rendering for detailed and complex structure reconstruction.

  • Differentiates by not requiring specialized cameras, working with basic RGB captures like smartphone images.

  • Offers scalability for reconstructing large areas or buildings through step-by-step and multi-level process.

  • Provides versatile 3D models applicable in VR/AR, gaming, robotics, cultural heritage digitization, etc.

  • The public availability of its code encourages further development and adaptation by the research community.

  • Neuralangelo's impact: democratized and scalable 3D reconstruction from common image sources, potential to reshape industries and the real-virtual world interaction.

Product Launches & Funding

NCSOFT Announces its AI LLM ‘VARCO’

  • NCSOFT, a global game developer and publisher, introduces 'VARCO LLM,' its AI language models.

  • VARCO signifies 'Via AI, Realize your Creativity and Originality.'

  • VARCO LLM are categorized as Foundation, Prompt, Dialogue, and Generative models, with performance based on parameter size.

  • VARCO is the first Korean language model available through Amazon SageMaker JumpStart, offering a free trial. Read More...

Electric Sheep Raises $500K Funding

  • Electric Sheep receives a capital injection led by AI-focused VCs, including Dasein Capital (US), and backed by Spatial Capital (US) and Look AI Ventures (Czech Republic).

  • The funding is intended for building and improving Spotlight, an AI rotoscoping tool designed for Hollywood-level detail.

  • Electric Sheep aims to streamline post-production workflows, enabling artists to focus on impactful visual effects and storytelling.

The UK Government announces new AI funding for carbon-intensive industries

  • The UK government has allocated GBP3.75 million for AI solutions to accelerate industrial decarbonisation.

  • Twelve green AI initiatives will receive GBP1 million to decarbonise and boost renewable energy generation.

  • An additional GBP2.25 million is dedicated to AI innovations targeting emissions reduction in energy sectors.

  • The government's Digital Catapult agency gets GBP500,000 to establish the Centre for Excellence on AI innovation for Decarbonisation (ADViCE).

  • ADViCE aims to facilitate collaboration between technology experts and carbon-intensive sectors to address decarbonisation challenges.

With Cubby, your AI travel companion, Ascott tests generative AI adoption

Ascott's Cubby - AI Travel Buddy

  • Ascott introduces a generative AI-powered web chatbot named "Cubby" to enhance travel planning and booking.

  • The chatbot is built on Microsoft OpenAI and Azure Services.

Cubby serves as a 'travel buddy' and offers travel insights, destination highlights, accommodation recommendations, attractions, shopping, adventure activities, and more. Read More..

Arthur launches Bench - A robust way to evaluate LLM

  • Bench is an open-source evaluation tool designed for comparing LLMs (Language Models), prompts, and hyperparameters for generative text models.

  • The tool aids businesses in assessing LLM performance in real-world scenarios to make informed decisions about integrating AI technologies.

  • Arthur Bench supports model selection, validation, budget optimization, and privacy control.

It enables the translation of academic benchmarks into practical performance measures for LLMs. Read More..

Cover Picture Credit: hpc2018

For sharing any interesting details, please reach out to us through a direct message on Twitter: Saran