- OpenTools' Newsletter
- Posts
- 🤯Apple's Most Powerful AI Yet
🤯Apple's Most Powerful AI Yet
Reading time: 5 minutes
Today we will discuss-
🤝🏻U.S. and UK partner on AI safety
👨🎨How to edit images in DALL-E
📲Apple develops AI that understands screen context
⚙️9 amazing AI tools you might not have heard of
All this and more - Let's dive in!
👩🍳What’s cooking in the newsroom?
U.S. and UK sign MoU to partner on AI safety, testing
♨️News - On Monday, the United States and Britain announced a new partnership dedicated to enhancing the safety of artificial intelligence, in light of growing concerns surrounding upcoming advancements in the field.
The agreement, signed by U.S. Commerce Secretary Gina Raimondo and British Technology Secretary Michelle Donelan, marks the world's first bilateral agreement on AI safety. It builds upon pledges made during an AI Safety Summit at Bletchley Park last November and outlines how the two governments will collaborate to pool technical knowledge, information, and talent on AI safety.
🔐So how do they plan to make AI ‘safer’? As part of their official collaboration, Britain and the United States aim to conduct at least one joint testing session on a publicly available AI model. The countries plan on exchanging expertise by swapping researchers between the UK's new AI Safety Institute (AISI), launched in November, and its yet-to-launch American equivalent. The institutes will also collaborate on methods to independently evaluate private AI models developed by companies like OpenAI and Google.
Additionally, they plan to exchange information regarding the capabilities and risks associated with AI models and systems, as well as technical research on AI safety and security. Both nations are also seeking to establish similar collaborations with other nations to advance AI safety.
OpenAI now offers a way to edit images you create with DALL-E
🖼️News - OpenAI has released a new editor interface tool that allows users to make edits to DALL-E images by describing the desired changes through text prompts.
🪄Here’s how - So there are two main ways to edit images: one is by selecting and highlighting a specific part of the image and providing detailed instructions for that area, and the other is making general changes to the entire image through the chat sidebar.
On the desktop version, there's a new select button represented by a paintbrush icon that allows users to highlight and edit specific parts of an image. Then, within the editor text box, users have the option to add, remove, or update specific characteristics of objects within the highlighted section by providing prompts such as "add clouds," "remove cat," or "change the dog's expression to happy."
Users can undo and redo selections using the corresponding buttons above the image, or clear the selection to start over entirely. To save a creation, click the save button positioned at the top-right corner of the editor. On the ChatGPT mobile app, users can use the select tool to highlight areas with their finger and then provide a prompt in the conversation panel.
For general edits, simply enter a prompt in the chat sidebar, such as "make it black and white." This feature too is available on the ChatGPT mobile app. Happy editing!
Apple’s latest AI can ‘see’ and understand on-screen context
🤖News - According to a newly published research paper, Apple researchers have created an advanced AI system that understands vague on-screen references, contextual cues from conversations, and background context—making conversations with voice assistants feel more natural.
🧐How exactly does it work? The system, named ReALM (Reference Resolution As Language Modeling), uses large language models to understand what you're talking about, even if you're referring to things you see on your phone or computer screen.
Basically, how it does this is by looking at everything on the screen and figuring out what's on it and where it's located. Then, it creates a written description that mirrors how things look visually, which helps it understand better. Apple researchers claim that this method, along with fine-tuning language models specifically for handling references, outperforms GPT-4 on the task.
They also believe that ReALM is a big step towards making voice assistants truly hands-free, as it would allow users to ask questions about what's on their screen without needing to touch anything.
🙆🏻♀️What else is happening?
🔫AI-based games
Left 4 Dead - It is a first-person shooter set in a post-apocalyptic world, where a pandemic has turned most of the population into mutated creatures. As part of a survivor team, you must navigate hazardous environments, complete objectives, and reach safety while facing waves of infected enemies.
Half-Life - It is a first-person shooter game where you play as Dr. Gordon Freeman, a theoretical physicist stuck inside an underground research facility. The game has been widely praised for its realistic incorporation of AI, which creates a dynamic, engaging and highly challenging environment for the players.
Red Dead Redemption 2 - It is a survival thriller game set in the late 1800s, where you play as Arthur Morgan, a member of the Van der Linde gang as he navigates the rapidly changing wild west landscape. The game's advanced AI makes the game even more life-like and captivating for human players.
👩🏼🚒Discover mind-blowing AI tools
Careered - An online platform that helps users creates cover letters from resumes and job listings (Free)
AI Poem Generator - AI-powered tool that generates unique rhyming poems on any subject (Free)
ThreadBois - An online tool that helps users create viral thread headers (Free)
Mixpanel Spark - A tool that allows users to ask questions in natural language and get answers about their product, marketing, and revenue data ($24/month)
Avian.io - A data automation platform for marketing data, integrating seamlessly with Google Sheets, Looker Studio, or Data Warehouses ($79/month)
TheDream.ai - A tool that uses AI and deep learning to create perfect profile pictures, Instagram posts, LinkedIn photos, photoshoots, and Avatars in various styles ($9.99/48 photos)
Wav2Lip for Automatic1111 - A tool that generates lip-sync videos by combining a video and a speech file (Open Source)
Motionagent - An AI assistant that helps users convert their ideas into motion pictures (Open Source)
ormGPT - An AI-powered ORM (Object-Relational Mapping) tool that translates plain language into SQL queries for easy database interaction (Open Source)
👨🏭Prompt engineering tips
Zero-shot prompting - It is a unique technique that is particularly effective for tasks like sentiment analysis, spam detection, and text translation or summarization. Here, you provide no examples in your prompt to a chatbot. Instead, you simply ask a question and rely on the model's pre-trained knowledge and understanding of language to generate a meaningful response or perform a specific task.
Few-shot prompting - Under this technique you can provide a small number of examples within your prompt, usually ranging from two to five. The model uses these examples to generalize and generate responses or perform tasks based on the input. This approach allows the model to adapt to specific tasks or prompts more effectively.
Chain-of-thought (CoT) prompting - It is a method that involves encouraging the chatbot to explain their thought process step by step before giving a final answer to complex problems. This helps the model break down difficult tasks into smaller parts, making it better at solving tricky problems that it otherwise might struggle with.
Can you help us, please?Be our mentor and let us know your take on today's newsletter. Share your honest and detailed thoughts on how we can improve OpenTools for you or just drop a Hi. YES! We personally read all the messages. |