AI Weekly Rundown (October 7 to October 13)
Major AI announcements from OpenAI, Google, Adobe, Microsoft and more this week.
Hello, Engineering Leaders and AI Enthusiasts!
Another eventful week in the AI realm. Lots of big news from huge enterprises.
In today’s edition:
✅ OpenAI’s GPT-4 Vision might have a new competitor, LLaVA-1.5
✅ Perplexity.ai and GPT-4 can outperform Google Search
✅ Microsoft to debut AI chip and cut Nvidia GPU costs
✅ Anthropic’s latest research makes AI understandable
✅ Google Cloud launches new generative AI capabilities for healthcare
✅ SAP’s new generative AI innovations for spend management
✅ Adobe reveals 100+ new AI features & 3 models
✅ Docker offers developers 2 new AI solutions
✅ ElevenLabs breaking language barriers with AI dubbing
✅ Tesla's Dojo Supercomputer Finds Home
✅ Replit bringing AI for all developers
✅ OpenAI plans developer-friendly updates
✅ OpenAI reveals how it develops models like GPT-4
✅ Google SGE can now generate images and drafts
Let’s go!
OpenAI’s GPT-4 Vision might have a new competitor, LLaVA-1.5
Microsoft Research and the University of Wisconsin present new research that shows that the fully-connected vision-language cross-modal connector in LLaVA is surprisingly powerful and data-efficient.
The final model, LLaVA-1.5 (with simple modifications to the original LLaVA) achieves state-of-the-art across 11 benchmarks. It utilizes merely 1.2M public data, trains in ~1 day on a single 8-A100 node, and surpasses methods that use billion-scale data. And it might just be as good as GPT-4V in responses.
Perplexity.ai and GPT-4 can outperform Google Search
New research by Google, OpenAI, and the University of Massachusetts presents FreshPrompt and FreshAQ. FreshQA is a novel dynamic QA benchmark that includes questions that require fast-changing world knowledge as well as questions with false premises that need to be debunked.
FreshPrompt is a simple few-shot prompting method that substantially boosts the performance of an LLM on freshQA by incorporating relevant and up-to-date information retrieved from a search engine into the prompt. Its experiments show that FreshPrompt outperforms both competing search engine-augmented prompting methods such as Self-Ask as well as commercial systems such as Perplexity.ai.
FreshPrompt’s format:
Microsoft to debut AI chip and cut Nvidia GPU costs
Microsoft plans to unveil its first chip designed for AI at its annual developers’ conference next month. Similar to Nvidia GPUs, the chip will be designed for data center servers that train and run LLMs, and is codenamed Athena.
Microsoft’s data center servers currently use Nvidia GPUs to power cutting-edge LLMs for cloud customers, including OpenAI and Intuit, as well as for AI features in Microsoft’s productivity apps.
Anthropic’s latest research makes AI understandable
Unlike understanding neurons in a human’s brain, understanding artificial neural networks can be much easier. We can simultaneously record the activation of individual neurons, intervene by silencing or stimulating them, and test the network's response to any possible input. But…
In neural networks, individual neurons do not have consistent relationships to network behavior. They fire on many different, unrelated contexts.
In its latest paper, Anthropic finds that there are better units of analysis than individual neurons, and has built machinery that lets us find these units in small transformer models. These units, called features, correspond to patterns (linear combinations) of neuron activations. This provides a path to breaking down complex neural networks into parts we can understand and builds on previous efforts to interpret high-dimensional systems in neuroscience, ML, and statistics.
Google Cloud launches new generative AI capabilities for healthcare
Google Cloud introduced new Vertex AI Search features for healthcare and life science companies. It will allow users to find accurate clinical information much more efficiently and to search a broad spectrum of data from clinical sources, such as FHIR data, clinical notes, and medical data in electronic health records (EHRs). Life-science organizations can use these features to enhance scientific communications and streamline processes.
SAP’s new generative AI innovations for spend management
SAP announced new business AI and user experience innovations in its comprehensive spend management and business network solutions to help customers control costs, mitigate risk, and increase productivity.
SAP will also embed Joule, its new generative AI copilot, throughout its cloud solutions, with availability in its spend management software planned for 2024. It has also unveiled SAP Spend Control Tower, which offers advanced AI features and the ability to see across all SAP spend solutions.
All these new AI innovations are being developed with security, privacy, compliance, ethics, and accuracy in mind.
Adobe reveals 100+ new AI features & 3 models
At Adobe’s annual MAX creative conference, They reveal over 100 new AI features across Photoshop, Illustrator, Premiere Pro, and beyond, including magical new generative AI features powered by 3 new foundational Adobe Firefly models for images, vectors, and design. And these 3 new models are:
Firefly Image 2 Model: It is the company's take on text-to-image generators; the major perk is the increased quality of renditions, higher resolutions, more vivid colors, and improved human renderings.
Firefly Vector Model: With this brand-new addition, Users can leverage gen AI and use a simple prompt to create "human quality" vectors and pattern outputs.
Firefly Design Model: It has text-to-template capability, allowing users to use text to generate fully editable templates that meet their design needs.
Enjoying the weekly updates?
Refer your pals to subscribe to our newsletter and get exclusive access to 400+ game-changing AI tools.
When you use the referral link above or the “Share” button on any post, you'll get the credit for any new subscribers. All you need to do is send the link via text or email or share it on social media with friends.
Docker offers developers 2 new AI solutions
Docker announced the launch of its GenAI Stack and AI Assistant at DockerCon. The GenAI Stack is a generative AI platform that helps developers create their own AI applications, while Docker AI assists with deploying and optimizing Docker itself. The AI assistant is currently available through an application to a Docker early access program.
This is the first AI offering for Docker, which is commonly used to build popular AI tools. They have collaborated with upstream communities to provide trusted AI/ML images, resulting in a significant increase in downloads and sharing through Docker's Hub registry service.
ElevenLabs breaking language barriers with AI dubbing
ElevenLabs, a voice AI platform, has launched a voice translation tool called AI Dubbing. It can convert spoken content into another language within minutes while preserving the original speaker's voice. It aims to break down language barriers and make content accessible to a global audience.
This new feature follows the recent launch of ElevenLabs' Projects tool, which supports streamlined long-form audio creation. The AI Dubbing feature supports voice translation across over 20 languages, Automatic detection of multiple speakers, Background sounds & noise splitting, and more.
Tesla's Dojo Supercomputer finds Home
Tesla building a Bunker-like structure at Tesla's Giga Texas facility, Sparked rumors that it could be used for housing operations for Tesla's Dojo supercomputing cluster.
The Dojo cluster trains the company's AI neural network for its Full Self-Driving system. However, it is unclear if the claims are true, as there have been no permits or plans for a Dojo center at the facility. Tesla CEO Elon Musk has previously mentioned the possibility of using Dojo to sell cloud services to other companies.
Replit bringing AI for all developers
Replit, a software development platform, is launching "Replit AI for All" to make AI-driven software development accessible to a wider audience. They are incorporating GhostWriter into their platform, renaming it 'Replit AI' and making it available to all users.
They have also introduced an open-source generative AI LLM called replit-code-v1.5-3b, trained on 1 trillion tokens to improve code completion. Replit AI is now accessible to over 23 million developers, with basic AI features available for free and more advanced features for Pro users.
OpenAI plans developer-friendly updates
OpenAI reportedly plans to launch major updates for developers next month, enabling them to build software apps cheaper & faster. The updates will include memory storage in developer tools, potentially reducing costs by up to 20 times.
OpenAI also plans to unveil new tools like vision capabilities for image analysis and description. The company aims to expand beyond being a consumer sensation and become a hit developer platform.
OpenAI reveals how it developed GPT-4 model
If you're looking for a simple, straightforward breakdown of how and what goes on at OpenAI, here’s an explainer revealed by the maker of ChatGPT. OpenAI explains how it develops its foundation models, makes them safer, and much more.
Developing an advanced language model like GPT-4 requires:
Pre-training: to teach models of intelligence, such as the ability to predict, reason, and solve problems by showing a vast amount of human knowledge over months.
Post-training: incorporating human choice into the model to make it safer and more usable.
Before publicly releasing GPT-4, OpenAI spent 6 months on post-training. During which, it developed techniques to teach the models to refuse to respond to requests that may lead to potential harm. OpenAI made GPT-4 82% less likely to respond to such requests compared to GPT-3.5. OpenAI also used this time to increase the likelihood of producing factual responses by 40%, making it more conversational, and improving its performance on low-resourced languages.
Google SGE can now generate images and drafts
Google is bringing new capabilities to its AI-powered Search experience (SGE).
Image generation: Now, SGE can whip up images if you type a description in search (below is an example). Every image generated through SGE will have metadata labeling and embedded watermarking to indicate that it was created by AI. Google is also developing a tool called About This Image that will help people easily assess the context and credibility of images.
Written drafts in SGE: To avoid longer-running searches for writing ideas and inspirations, SGE will write drafts for and also make them shorter or change the tone. From there, it's easy to export your draft to Google Docs or Gmail.
That's all for now!
Subscribe to The AI Edge and gain exclusive access to content enjoyed by professionals from Moody’s, Vonage, Voya, WEHI, Cox, INSEAD, and other esteemed organizations.
Thanks for reading, and see you on Monday. 😊