AI technology keeps improving fast. This week brought several interesting updates! Here’s a look at what’s new and how you might use it.
Meta AI Gets More Social
Meta recently held their first event just for AI called Llamacon. A key announcement was the new Meta AI app. This app is a rebranded version of the Meta View app for Ray-Ban glasses. Now, it includes a standalone chat feature where you can talk directly with Meta’s Llama AI model.
One unique part of the Meta AI app is its social feed…. You can share your AI conversations and creations. Other users can comment, like, and share these posts. This feed can be a good place to find inspiration for prompts. The app can also create images using Meta’s Emu AI. The image quality is noted as “pretty solid”.
If you use Ray-Ban Meta glasses, conversations can now move easily between the glasses, the app, and a web app.
Heads up on privacy and ads: For Ray-Ban glasses, camera use is always on unless you turn it off. Photos and videos taken by voice command are saved to your phone and not used for Meta’s AI training3. However, storing voice recordings in the cloud cannot be turned off. You can delete recordings anytime, but they can be stored for up to a year to help Meta improve its products.
Also, Meta plans to add ads to the Meta AI app in the future…. This might include product recommendations within chat responses. The focus is on growing the app first before rolling out the business side.
Google’s Helpful AI Tools
Google also rolled out several updates. Their AI mode in Search is becoming available in the US for Labs users. It provides AI-generated answers with links and images, similar to other AI search tools…. A small test is also rolling out for some users outside of Labs.
The Gemini app on Android now lets you edit images directly. You can upload photos and ask Gemini to add things like sunglasses or hats.
For creators or researchers, Notebook LM (a tool that summarizes documents you provide) now offers audio overviews in over 50 languages. It can create a mini-podcast from your data in different languages.
Google also launched a new app called Little Language Lessons. It’s designed to help travelers learn basic vocabulary and phrases for specific situations like eating at a restaurant…. It includes features like Tiny Lesson for vocabulary, Slang Hang for idioms, and Word Cam to identify objects in photos.
Recraft: Master Your Image Style
Recraft is an image generation and editing platform with many tools. They just launched a huge new style library. You can browse endless styles or search for specific ones like “comic book”….
A cool feature is the ability to save your favorite styles. Even better, you can create custom styles by mixing multiple saved styles and adjusting how much each one influences the final image. You can then share these custom styles with others.
These updates make it easier to rapidly test different looks or ensure brand consistency by applying the same style to all your generated images.
One thing to keep in mind that I do not like is that if you use a free account the images are owned by Recraft and you cannot use them commercially. Then what’s the point?
OpenAI & Other Creative Updates
OpenAI rolled back a recent update to GPT-4o because its personality became too overly supportive8. They are working on a better balance8.
ChatGPT Search has been improved, including a better shopping experience8. You can ask for product recommendations, and it will show suggestions8. These product results are independent and not ads8. Search is also now available in ChatGPT within WhatsApp8….
An interesting observation with GPT-4o image generation is that if you repeatedly ask it to recreate an image exactly, it gradually changes the image over many tries10…. The final image can look very different from the start11.
For artists and creators, several tools added fun features:
MidJourney released Omni Reference…. This lets you include an image in your prompt to influence the output, like injecting your face into a picture of a Viking. It works well for faces and requires version 7.
Kling AI added an Instant Film effect. You can upload a portrait photo, and it turns it into an animating Polaroid-like image. This works for photos with multiple people or animals.
Higgsfield AI launched Iconic Scenes. You can upload a selfie and put yourself into famous animated movie moments.
Crayo introduced GPT Paint. This allows you to visually edit images by drawing shapes or arrows and adding reference images or text. For example, you could draw an arrow from a photo of boots to a dinosaur image and ask it to put the boots on the dinosaur.
Suno (music generation) released version 4.5 with minor updates like expanded genres and improved voices.
Other Noteworthy AI News
Duolingo announced it is becoming an “AI-first company”. They plan to use AI to handle repetitive tasks, allowing human employees to focus on creative work and problem-solving. The goal is to remove bottlenecks, not replace staff.
For drivers, Lyft launched an AI earnings assistant. It uses real-time data to help drivers optimize their routes and times to potentially earn more.
In transportation, the company Aurora has started using driverless trucks for customer deliveries on public highways in Texas.
These are just some highlights from a busy week in AI! While core AI models see steady improvements, the tools bringing AI to creative tasks and specific business needs are making exciting progress