- The Deep View
- Posts
- ⚙️ New ChatGPT Features
⚙️ New ChatGPT Features
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/10e1c6c8-bf68-496a-abe9-35712a89e1e7/DV_Newsletter_Banner__2_.png)
Good morning. Just as Ahsoka Tano wields her lightsabers with precision and flair, MidJourney's new inpainting feature is slicing through the AI world, transforming images with Jedi-like mastery. Ready to join the rebellion of creativity? May the force be with your designs.
In today’s newsletter:
🎨 MidJourney's Inpainting Feature: Innovating Image Editing or Playing Catch-Up?
🚀 OpenAI's GPT-3.5 Turbo: Personalizing AI with Fine-Tuning and Efficiency
🧠 Predicting Brain Cancer Outcomes with Precision: Stanford's AI Breakthrough
📈 Baidu Triumphs in Q2: Generative AI Fuels Revenue Growth and a New Era
🏥 Innovaccer's Affordable AI Suite: A Revolution in Healthcare Automation
👁️ AI-Powered Eye Scans: Unlocking Early Parkinson's Detection
NEWS
MidJourney's New Inpainting Feature: A Game Changer or Catch-Up Play?
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/c0d5be02-5c11-40b3-9de1-5f6059caf0a8/image.png)
Midjourney, one of our favorite AI image generators, has unveiled its latest feature, "Vary (Region)," allowing users to engage in inpainting—a process that edits elements within an image's canvas. This new tool enables users to select specific portions of an image and modify them using text prompts, such as transforming a part of the image or adding accessories to a portrait. While this marks a significant development for Midjourney, it's worth noting that Stable Diffusion, a free, competing open-source image generator, has supported both inpainting and outpainting functionalities since its v1.2 release (June 2022).
Here’s a breakdown by Midjourney themselves on how to use Vary Region 👇
The introduction of Midjourney's Vary tool reflects the growing trend in AI towards more refined and user-centric solutions. Users can simply create an image, upscale it, select the area they want to edit, and enter a prompt to generate new results. However, the tool is not without limitations. Midjourney's developers have acknowledged that the system may sometimes "fight your intent," and the feature is most effective when applied to larger sections of an image, ideally between 20% to 50%.
Though we prefer Midjourney as an overall text-to-image generator, Stable Diffusion's inpainting seems to be more accurate and user-friendly when compared side-by-side. Midjourney's Vary tool is great for large edits and adding new elements to a composition, but Stable Diffusion outperforms it in accuracy, & creativity, and works better with smaller areas of an image. This highlights the competitive nature of AI today, where various platforms offer overlapping functionalities, and even a paid tool like Midjourney, with a minimum annual subscription cost of ~$100, faces stiff competition from free alternatives.
TOGETHER WITH KOMBAI
Breakthrough in AI: World’s First AI Model that you can Prompt with Designs
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/f2989cc6-9965-4e40-ac7f-be93518bc36c/image.png)
Say hello to Kombai, an AI model that understands and codes UI designs like humans.
Because none of the existing LLMs/ GenAI models work on UI designs, they built a new ensemble model from scratch.
Remarkably, even for complex designs, it generates developer-quality code in just one click per component. You can even "prompt-engineer" the designs to steer it in the right direction.
They just launched their Research Preview, and it's free to use!
NEWS
OpenAI's GPT-3.5 Turbo Fine-Tuning: A Tailored Revolution in AI
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/0c1a0e65-75dc-47bb-8c51-efd8159cc5a9/image.png)
OpenAI has just unveiled GPT-3.5 Turbo fine-tuning, an update that enables you to further personalize your AI experience. From guiding your model's direction to presenting its output to setting its voice, this release is like having a personal team of experts tailoring your AI. Here's what's new:
Steerability: Guiding Your AI's Language Direction
Steerability ensures that ChatGPT follows specific instructions, like responding in a given language.
Example: If you're developing a multilingual customer support bot, steerability ensures that the model always responds in the language prompted by the user. If a user asks a question in German, the bot will respond in German, maintaining a seamless user experience.
Formatting: Neatly Presenting Your AI's Output
Formatting organizes ChatGPT, making sure it's presented in a consistent format. It's crucial for applications like code completion or composing API calls, where specific response formats are required.
Example: If you're a small business owner who needs to send out monthly newsletters, formatting ensures that your AI generates content that aligns with your specific layout and design preferences.
Tone: Setting the Voice for Your AI
Tone customization lets you fine-tune ChatGPTs voice to match your brand or audience. It's a great way for businesses with recognizable brand voices to maintain consistency across their AI-powered communications.
Example: If you're a non-profit organization focused on environmental conservation, you can set the tone of your AI to be passionate and inspiring, reflecting your mission and values.
Efficiency and Power: More Done, Faster
GPT-3.5 Turbo fine-tuning doubles the token limits and reduces prompt sizes, speeding up API calls and cutting costs. It's a powerful upgrade that handles large documents with ease.
Safety and Transparent Pricing: Trust and Clarity - Skip this part if you’re not interested in the costs associated with GPT-3.5 Turbo Tuning
OpenAI has been able to preserve the default model’s safety features through its Moderation API and a GPT-4 powered moderation system.
Fine-tuning costs are broken down into two buckets: the initial training cost and the usage cost:
• Training: $0.008 / 1K Tokens
• Usage input: $0.012 / 1K Tokens
• Usage output: $0.016 / 1K Tokens
For example, a gpt-3.5-turbo fine-tuning job with a training file of 100,000 tokens that is trained for 3 epochs would have an expected cost of $2.40.
New Models and API Endpoint
OpenAI is introducing babbage-002 and davinci-002 as replacements for the original GPT-3 base models. These models can be fine-tuned with a new API endpoint, offering more extensibility and support for the future evolution of the fine-tuning API.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/4b502f3a-d876-4868-831e-f6f1402c3c7e/Rapid_Fire_AI_News__1_.png)
In Monday’s edition of the newsletter, we compared two images of strawberries. This comparison seemed to be the easiest for y’all considering that this poll yielded the highest percentage of correct answers. Congrats!
Here are a couple of your thoughts on why you voted for a given image over the other! (Please let us know if this is something you’d like to see included in future editions too 😁) ⬇️
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/e381cd9a-3b3d-400e-b393-55552f11c5b8/Untitled_design__10_.png)
Last week’s results ⬇️
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/c8fa751b-e0f0-497b-a380-b4db8fb84465/image.png)
Today’s game will hopefully be a bit more difficult…
Best of luck! 😉
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/dc8ebbc1-f1be-4524-92f9-fd13f1959f2c/image.png)
Image 1
Which Image of Pillows is Real?Please select which image you think is real. |
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/23cd0c17-f22b-428c-af16-b4c8273855a1/image.png)
Image 2
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/fde392a0-fd53-4041-a40a-890ab3a81032/1__3_.png)
🧠 Predicting Brain Cancer Outcomes with Precision
Stanford Medicine scientists have created an AI model to predict the severity of glioblastoma, a type of brain cancer. By analyzing stained images of tumor tissue, the model helps physicians understand the genetic makeup of tumors and assess remaining cancerous cells, potentially aiding drug developers and surgeons in targeting the cancer more effectively.
📈 Baidu Triumphs in Q2, Cheers Generative AI Progress: A New Era?
China's Baidu Inc exceeded Q2 revenue estimates, driven by post-pandemic recovery in advertising and intensified efforts in generative AI. With investments in autonomous driving and LLMs, and awaiting regulatory approval for its Ernie bot, Baidu's revenue reached $4.67 billion, signaling a promising future in AI.
🏥 Innovaccer Unveils Affordable AI Suite: Revolutionizing Healthcare Automation
Health data analytics company Innovaccer has expanded its AI assistant, Sara, into a suite of health AI products. The suite includes solutions for insights, care management, point of care, and experience center, aiming to make healthcare operations more efficient and patient-centric, with an emphasis on affordability and accuracy.
👁️ AI-Powered Eye Scans: A New Horizon in Early Parkinson's Detection
Scientists from London's Moorfields Eye Hospital and the UCL Institute of Ophthalmology have found that AI-powered eye scans could detect Parkinson's disease before symptoms appear. By analyzing optical coherence tomography (OCT) scans, physical differences were identified in the eyes of people with Parkinson's, seven years before clinical presentation, offering a promising pre-screening tool for early intervention.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/45e3ed1b-8465-4463-9fcf-07206439ee49/Rapid_Fire_AI_News__5_.png)
Have cool resources or tools to share? Submit a tool or reach us by replying to this email (or DM us on Twitter).
SPONSOR THE DEEP VIEW
📈 Get in front of 50,000+ AI Enthusiasts!
The Deep View is currently one of the world’s fastest-growing newsletters, adding thousands of AI enthusiasts a week to our incredible family of over 50,000! Our readers work at top companies like Apple, Meta, OpenAI, Google, Microsoft, and many more.
If you want to share your company or product with fellow AI enthusiasts before we’re fully booked, reserve an ad slot here.
* This is sponsored content
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/c0294cd3-1aaa-4a7b-8780-f347bc431e61/image.png)
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/bb859100-f74d-48fc-a72e-286a1dc695cb/image.png)
Prompt: A stained glass image of a bear in a forest --ar 16:9
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/59312f58-ef8d-44b6-bbf7-7874f327a9c1/image.png)
Prompt: a futuristic black motorcycle concept sheet --ar 16:9
What'd you think of today's email?Your feedback helps us create better emails for you! |
That's a wrap for now, we hope you enjoyed today’s newsletter!
Should you have any captivating projects or concepts, don't hesitate to connect with us by replying to this email or dropping us a direct message on Twitter: @thedeepview
We appreciate your continued support, and we'll catch you in the next edition.
-Bot Buddy and The Deep View Team