🧞‍♀️Create Magic with DALL-E

Hey TEDchats Fam!

Happy Friday! Ever stared at a blank canvas, your mind buzzing with ideas, yet your hands unsure where to begin? Imagine a world where your visionary ideas not only leap onto the canvas without a single brushstroke but can also be tweaked and transformed with just a few clicks. Welcome to DALL-E 3’s new upgrade!

👀 SNEAK PEEK

🚨FEATURE STORY

DALL-E's New Editing and ‘Inspirational’ Features

Imagine conjuring up a sci-fi polar bear on a tropical vacation with just a few clicks and then add an ice cold beer to the picture. This is no longer just a whimsical dream, thanks to the latest upgrades to OpenAI's DALL-E, which promise to transform our interactions with AI-generated art.

To Start: DALL-E 3 Gets a New Home

OpenAI's DALL-E has been at the forefront of AI-driven creativity, enabling its ChatGPT Plus and Business users to generate detailed images from simple text descriptions. With its third iteration, DALL-E 3 now stands quite literally alongside ChatGPT. Positioned right below the ChatGPT 'new chat' tab on the left sidebar, there is now a dedicated 'new chat' tab for DALL-E 3. Previously confined to the ChatGPT chat, DALL-E 3 now has its designated space, enhancing the overall experience and streamlining the creation of AI-generated masterpieces.

The First Problem with DALL-E — Now Solved!

Despite DALL-E's innovative capabilities, users often faced two frustrating challenges. Firstly, if their generated image was perfect except for an unwanted background blemish, their only option was to create an entirely new image, hoping it wouldn't deviate too much — a challenging endeavor. To tackle this issue, OpenAI has unveiled new editing tools that allow users to select areas and instruct the AI on desired changes (see OpenAI's post and demo video).

The Second Problem with DALL-E — Now Solved!

The second daunting challenge users face is what's known as 'the blank canvas syndrome'. How does one begin generating images amidst limitless possibilities? What should one do when inspiration fails to strike? OpenAI has introduced a solution: a series of 'inspirational styles' tabs positioned above the text input box. These tabs provide users with a preview of their style options, helping them overcome creative blocks and start generating ideas (see OpenAI’s post and demo video).

Turning Fantasy into Reality

The upgrades to DALL-E have made creating AI-generated art more accessible to casual users and enthusiasts alike. To address issues of trust and misuse, OpenAI has also announced the inclusion of watermarks and embedded metadata in DALL-E 3 images, signifying progress in safeguarding content integrity (learn more). Overall, the evolution of DALL-E hints at a future where AI and human creativity collaborate more seamlessly than ever before.

🧑‍💻 DAILY PROMPT

🍕 QUICK BITES

Anthropic Discovers Dangerous Jailbreaking Technique!

Anthropic, a prominent AI startup, has shown how a certain style of questioning compels large language models (LLMs) to reveal dangerous information, including instructions for creating forbidden materials like a bomb. By bombarding LLMs with 'How do I' questions, AI models gain momentum in answering, eventually even providing responses to hazardous inquiries. Anthropic's 'many-shot jailbreaking' technique worked for both their AI models and others, prompting them to alert other developers. This finding exposes AI's safety vulnerability and underscores the swift advancements and ongoing challenge of securing AI systems against misuse (see Anthropic’s announcement).

OpenAI’s ‘Custom Models’ Program for Businesses

OpenAI is pushing the boundaries of generative AI technology with an exciting expansion of its ‘Custom Models’ program. By introducing assisted fine-tuning and the development of custom-trained models, OpenAI aims to meet the growing demand for highly personalized AI solutions. This strategic move allows businesses to fine-tune OpenAI’s base models, like GPT-4, to their specific needs, enabling more accurate, industry-specific applications. OpenAI has also launched an inquiry page for organizations interested in collaborating with OpenAI researchers to train custom GPT-4 models tailored to their specific domain (see OpenAI’s announcement).

Microsoft 365’s Copilot Elevates to GPT-4 Turbo

With plans to “bring the power of generative AI to everyone across work and life”, Microsoft is integrating the advanced GPT-4 Turbo into Copilot for Microsoft 365 users. This significant upgrade will also include extra perks, such as the removal of daily chat limits and increased file upload capabilities. Later this month, users will also be able to enjoy enhanced image generation capabilities within Microsoft Designer, including priority access and an increased number of image generation boosts (see Microsoft’s announcement).

🧰 AI TOOLBOX 

🤯 MIND SPARK

Did You Know: AI originated in the 1950s from neuroscientists' efforts to understand the brain. A recent report in the renowned scientific journal Nature delves into this topic, examining the current interplay between AI and neuroscience (read the Nature report).

Eclipsing the Extraordinary with a Pixel Phone: On April 8, North America will witness a total solar eclipse, a rare event not to recur for another two decades. Learn how a Google camera expert recommends capturing this fleeting spectacle with a Pixel phone (read expert tips).

The Dawn of AI Gadgets: April 2024 marks the beginning of an era filled with AI-first gadgets, including Humane’s AI Pin, Rabbit’s r1 gadget, Brilliant Labs’ Frame AI Glasses, and Meta AI Ray-Ban Smart Glasses Updates, transforming how we interact with technology (see The Verge’s full list).

🎨 AI ART