• AI Emergence
  • Posts
  • Google to launch “Phone Intelligence” ahead of Apple

Google to launch “Phone Intelligence” ahead of Apple

Along with: xAI has finally launched Grok 2 (but it’s not perfect)

Hey there, 

We just concluded India’s largest GenAI conference and the entire experience has been surreal. It is one thing to talk to the community members online and another to meet them personally, know about them, their work, their challenges, and to know how we are impacting their lives. Imagine doing that for close to 1000 people in 3 days! That is what I did during the DataHack Summit.

Let’s go through what the AI world has been up to this week

What would be the format? Every week, we will break the newsletter into the following sections:

  • The Input - All about recent developments in AI

  • The Tools - Interesting finds and launches

  • The Algorithm - Resources for learning

  • The Output - Our reflection 

Table of Contents

Google announced a bunch of updates during the Pixel 9 launch. There are big GenAI updates coming to the Pixel series as well as Android. Here are the updates - 

  1. Gemini Live: This feature allows users to interact naturally without needing specific commands. It's designed for fluid, real-time interactions, making the assistant more responsive and intuitive in day-to-day tasks.

  2. Context-Aware Assistance: Deep integration with Android enhances task management across apps like Keep, Tasks, and YouTube Music, providing smarter, contextually relevant suggestions.

  3. Improved Speed and Quality: The Gemini 1.5 Flash model enhances response speed and accuracy, ensuring a more seamless user experience. (source)

DataHack Summit - India’s most futuristic GenAI conference was held in Bengaluru, drawing a crowd of over 1,200 AI practitioners and visionaries. The event featured sponsorship from leading technology firms, including Intel, Google Cloud, Merkle, American Express, and LTIMindtree, among others.

The summit provided a perfect blend of knowledge sessions and entertainment with its AI playground, engaging a diverse audience of AI professionals. Key sessions included discussions on the latest advancements in AI technologies such as the RAGs, Diffusion model, and Large Language Models (LLMs). (source)

xAI launched Grok-2 and Grok-2 mini, upgraded versions of its chatbot that are currently limited to Premium and Premium Plus subscribers on X, with plans to extend access via an enterprise API later this month.

One of the notable features is the image generation capability. These models include a prompt-based image generation feature powered by Black Forest Lab's Flux 1 AI, allowing users to create and share images directly on the X platform.

But there have been some serious concerns, particularly in spreading misinformation.

Early examples of generated images show recognizable political figures in controversial scenarios, raising concerns about potential misuse.

Sakana AI and researchers from Oxford and British Columbia have come up with AI scientists, a groundbreaking system designed for fully automated scientific discovery. 

This system enables Foundation Models, including Large Language Models (LLMs), to independently conduct research across various subfields of machine learning. 

The AI Scientist streamlines the research process from generating ideas and writing code to conducting experiments and composing scientific papers, including peer review. (source)

Flux.1, developed by Black Forest Labs, is gaining recognition for its impressive capabilities and open-source availability. 

It is said to be surpassing all previous diffusion models, promising high-quality, accessible AI tools for image generation. 

It offers three versions- Pro, Dev, and Schnell, tailored to different performance needs and accessible even on high-performance laptops, making it suitable for a wide range of users from hobbyists to commercial entities. In fact, Flux.1 does not require internet access for operation, enhancing its usability. (source)

Researchers from Meta and Oxford have created an AI model named VFusion3D that can generate detailed 3D objects from single images or textual descriptions. 

This model represents a significant advancement in scalable 3D AI, with potential applications across virtual reality, gaming, and digital design. 

The researchers who spearheaded this project developed a novel method utilizing pre-trained video AI models to produce synthetic 3D data, overcoming the challenge posed by the scarcity of 3D training data. VFusion3D has outperformed previous models by converting 2D images into 3D in seconds. (source)

SingularityNET is set to launch its first supercomputer this September, aiming to advance toward Artificial General Intelligence (AGI). 

It will facilitate complex AI architectures that mimic human brain functions. The launch is expected by early 2025, featuring hardware from Nvidia, AMD, and Tenstorrent. 

This system will support the development of deep neural networks and large language models, allowing for extensive AI training and operation without constant internet access. SingularityNET will use its AGIX token for access and data contributions, positioning itself as a significant player in the quest for AGI.(source)

Cognition's Devin, an AI-driven software engineer powered by GPT-4, amazed users with its ability to autonomously write and edit code. However, just five months later, Cosine, a startup from Y Combinator, has introduced its own AI engineer, Genie, which significantly outperforms Devin. 

Genie scored 30% on the SWE-Bench benchmark, far exceeding Devin's 13.8% and even surpassing Amazon’s Q and Factory’s Code Droid. Designed to mimic human software engineers, Genie can handle coding tasks such as bug fixing and code refactoring autonomously or in collaboration with users. (source)

YouTube is testing a new feature called "Brainstorm with Gemini" that integrates Google’s AI technology to assist creators with generating video ideas, titles, and thumbnails. 

Announced on the Creator Insider channel, this feature is currently available to a select group of creators as part of a limited experiment. YouTube aims to gather feedback before deciding on a broader rollout. 

This new tool aims to differentiate YouTube from other social media platforms by providing unique AI-driven support. Creators can choose between the new Gemini-powered feature and an existing AI inspiration tool to get content ideas and outlines. (source)

Tool: Mockey.ai

Have you ever wished to see your custom t-shirt designs on models before launching your brand? Mockey.ai is a tool that makes this possible by automatically showcasing your design on a model. It provides t-shirt mockups that can be used for selling on e-commerce platforms, all without the need for expensive photoshoots.

Problem Statement: Suppose you are starting your t-shirt brand and need an efficient way to get the mockups of your designs on real-life models to better present them to potential customers.

How to Access the Tool:

  1. Visit Mockey.ai and go to upload a design.

  2. Add the design that you want to print on your t-shirt and then The tool will automatically generate images of models wearing your design.

  3. Download the generated images and use them for marketing or product listings.

This tool is ideal for entrepreneurs and brands who want to enhance their product presentation, providing an easy and visual way to showcase designs without the need for professional photoshoots.

Check what I generated by clicking here.

  • In the recent episode of Leading with Data, I had a conversation with Gaurav Agarwal, the Founder and CEO of RagaAI, he is a recognized thought leader in the AI industry. We discussed his views on building high-performing teams and the evolution of AI technologies, as well as his experience in leading innovation at major tech companies like NVIDIA and Ola Electric Mobility.

  • This course by Sharon Zhou and Andrew Ng is about fine-tuning large language models (LLMs) to enhance their performance for specific tasks and applications. It is designed for practitioners and researchers in the AI field who are interested in learning advanced techniques for customizing LLMs.

What were you up to this week? What was the new tool you learned? If you attended the DataHack Summit - what was your top learning from it?

Look forward to hearing more from you.

How do you rate this issue of AI Emergence?

Would love to hear your thoughts

Login or Subscribe to participate in polls.

Reply

or to participate.