AI is changing the world, and businesses are no exception. In this week’s roundup, we’re taking a look at some of the latest AI tools, and news that can help you stay ahead of the curve.
Let’s dive in!
- AI in development and programming
- AI in creative fields
- AI in various applications
- AI in unintended consequences
- AI research and insights
AI in development and programming
GitHub’s code referencing feature for Copilot
GitHub Copilot is an AI-powered tool that helps developers write code. A new feature, code referencing, identifies instances where Copilot’s suggestions match publicly available code on GitHub. This information can help developers make informed decisions about using code suggestions, such as for learning, dependency management, or giving credit. Code referencing is currently in private beta, but it has the potential to improve the developer experience and help developers connect with the open-source community.
BigCode project chronology
The BigCode project has made significant progress since its announcement on September 26, 2022. Key milestones include the release of the Stack dataset, the introduction of language models SantaCoder and StarCoder, and outreach to the BigCode community and the NLP research community. The project is on track to achieve its goals of advancing code-related technologies and promoting community engagement.
Stable Beluga – fine-tuned instruction models from Stability AI
Stability AI and its CarperAI lab have released two new large language models, Stable Beluga 1 and Stable Beluga 2. These models are based on the LLaMA foundation models and have been fine-tuned using synthetically-generated datasets. They are designed for various applications, including intricate reasoning, understanding linguistic nuances, and solving complex problems.
The Stable Beluga models are available for open research under a non-commercial license. They have the potential to significantly impact the AI community by enabling new applications and advancing research in natural language processing.
IDX – a new tool for developers
Project IDX is Google’s new online development tool for developers. It is like a digital workshop where developers can create apps and software effortlessly. Developers can access Project IDX from anywhere, whether they are at their desktops at home or on the go.
AI in creative fields
HeyPhoto – a new tool for photographers
HeyPhoto is an AI-powered photo editing service that lets you enhance your selfies and group photos. You can modify facial features like gaze direction, nose shape, skin tone, age, and hairstyle. The service supports JPG, PNG, and WEBP images. It also offers features like face swaps and photorealistic headshots.
CapCut – a tool for video editing
CapCut is a free, easy-to-use video editing app for mobile and desktop. It offers a variety of features for editing videos, including trimming, cutting, and adding music, text, and effects. CapCut is a popular choice for creating short-form videos for social media, such as TikTok and Instagram Reels. It is also a good option for beginners who want to learn how to edit videos.
GamifAI – gamification with AI
GamifAI is an AI tool that simplifies game development. It can handle aspects like art and code, freeing up developers to focus on their creative ideas. GamifAI is designed to democratize game development, making it possible for anyone to create games without extensive resources.
AI in various applications
AI can steal data by analyzing the sound of keystrokes
Researchers have developed a new cybersecurity threat that can steal passwords and sensitive data by analyzing the sounds of keystrokes. The attack uses a deep learning model to recognize and decipher the sounds produced by typing on a keyboard. The researchers achieved high accuracy rates, around 95% when using a microphone next to the keyboard, and 93% accuracy when recorded over platforms like Zoom.
eBay’s new search feature inspired by window browsing
eBay’s new visual search feature uses computer vision to let users find items similar to their preferences using images. This is inspired by window shopping and is more effective than traditional textual search. The process involves generating vector embeddings for items and matching them based on their visual similarity. This new search experience promises to improve customer engagement.
AI in unintended consequences
PaknSave AI bot spits out recipe for chlorine gas
A New Zealand supermarket chain, Pak’nSave, introduced an AI-powered recipe suggestion tool called Savey Meal-Bot. The tool generates recipes based on the ingredients users have at home. However, the AI inadvertently generated a recipe for an “Aromatic Water Mix” that included dangerous ingredients like bleach and ammonia. This could have resulted in the creation of deadly chloramine gasses.
The incident highlights the unintended consequences of using AI to generate recipes. The AI is not always able to recognize harmful combinations of ingredients, and it can sometimes produce unusable or bizarre recipes. The supermarket chain has acknowledged the challenges of using AI in this context and is working to improve the safety and accuracy of the Savey Meal-Bot.
AI research and insights
Interview with Anthropic CEO Dario Amodei
In a podcast interview, Dario Amodei, CEO of Anthropic, discusses various topics related to AI, including scaling, neural circuits, alignment, implications, and integration. The conversation covers the reasons for AI scaling success, mechanistic interpretability of neural circuits, alignment strategies for training models, cybersecurity concerns, and AI’s impact on society and decision-making.
New chat research from Stability AI
A new open-access large language model (LLM) with advanced capabilities, including intricate reasoning and solving complex problems, has been launched. The creators invite AI safety researchers and developers to help improve the model’s safety and performance.
Generative Agents – open source project on GitHub
The Generative Agents project creates interactive simulations of human behavior using computational agents. The project’s repository contains code and resources for creating these simulations, as well as detailed instructions for setting up the simulation environment, running simulations, replaying simulations, and customizing the behavior of the agents.
Neuralangelo – new project from NVlabs on GitHub
Neuralangelo is an open-source project for high-fidelity neural surface reconstruction. It is based on the Imaginaire library from NVIDIA and was presented at CVPR 2023. The code can be set up using prebuilt Docker images or a conda environment. Data preparation and surface reconstruction instructions are also provided.
By providing personalized recommendations, optimizing pricing, and automating customer service, AI can help businesses improve their bottom line. If you’re not using AI in your eCommerce business, you’re missing out.
Contact our experts today, and we’ll help you choose the right AI solutions for your needs and implement them effectively.