AI Weekly Update: Breaking Boundaries with DreamDiffusion, OpenFlamingo, and Unity Muse

Hey, what's up guys, Eddie here with a fresh scoop on the AI and tech universe. Let's dive right into it.

Microsoft is stepping up its game with a suite of AI-powered shopping tools integrated into Bing search and Edge. These ingenious tools include AI-curated buying guides which automatically sift through product specs and buying options based on your search queries. They also pack AI-powered review summaries, giving you the quick and dirty rundown of online product reviews.

On the subject of AI, Salesforce AI Research dropped the XGen-7B, an incredible new open-source 7B LLM, trained with a staggering 1.5 trillion tokens over an 8K input sequence length.

In a fascinating development, researchers have unveiled DreamDiffusion, a groundbreaking technique that allows the generation of high-quality images directly from brain EEG signals. This cutting-edge tech makes it possible to bypass the step of translating thoughts into text.

Microsoft, in an AI Skills Initiative, has announced a bunch of exciting offerings. We're talking free coursework in partnership with LinkedIn, a global grant challenge open to all, and amplified access to free digital learning events and resources for AI education.

OpenAI’s Flamingo model has an impressive counterpart. Stability AI released OpenFlamingo V2, an open-source reproduction, boasting more than 80% of the performance of the original Flamingo.

Meanwhile, Unity is bringing two AI-powered tools to the party. Unity Muse, a tool that churns out animations, 2D sprites, and textures in the Unity Editor using your text and sketches, and Unity Sentis, a tool that embeds an AI model into the Unity Runtime for your game or application, running flawlessly on any device where Unity does.

ElevenLabs is launching Voice Library, a platform for sharing AI-generated voices fashioned with their voice design tool.

In the education tech space, Merlyn Mind is on a mission to revolutionize the learning experience with the release of three open-source, education-specific LLMs.

Amazon's AWS is making waves with the launch of the Generative AI Innovation Center, a $100 million initiative that ropes in AWS machine learning and AI expertise to build and roll out generative AI solutions with businesses.

A slick new open-source text-to-video AI model, Zeroscope_v2 XL, is now available. This model delivers high-quality video at 1024 x 576 without those pesky watermarks.

MotionGPT is making its debut, a new motion-language model primed to handle a multitude of motion-centric tasks.

In the startup acquisition space, Databricks is set to bag the open-source startup MosaicML for a cool $1.3 billion. MosaicML recently unleashed MPT-30B, an open-source model licensed for commercial use that trumps the original GPT-3.

In the job market, US-based generative AI-related job postings spiked around 20% in May according to Indeed’s data.

For the coders out there, the DragGAN algorithm source code has been released, allowing interactive point-based manipulation on the generative image manifold, with a demo available on Huggingface.

The language model space is heating up as China’s Baidu outshines ChatGPT (3.5) and even outperforms GPT-4 in several Chinese language capabilities with its new foundation model, ERNIE 3.5.

In other news, Google is set to host the first-ever Machine Unlearning Challenge on Kaggle.

For Adobe customers, the company is offering to cover any claims in case of lawsuits over the use of content generated by Adobe Firefly, their generative AI image tool.

And finally, Google is launching generative AI coding features in Google Colab exclusively for Pro+ subscribers in the US.

There you have it! This is Eddie signing off, reminding you to stay on the pulse of the AI revolution.