Power to the People!

Open Source LLMs stand and deliver, Google makes Bard (and search) smarter, and Data's Dark Side is explored in court.

The Big Stuff

Open Source LLMs Drive a New Wave of AI Innovation

Open source large language model (LLM) projects are rapidly gaining traction, democratizing access to powerful AI tools and creating a ripple effect across the artificial intelligence landscape. The rise of open source LLM projects, like those above, has far-reaching implications for the AI landscape. By making powerful AI tools open, these projects foster innovation, collaboration, and the development of diverse applications across industries. In some cases, these products are becoming more capable than the available commercial products. One example of this is MiniGPT-4, a multimodal LLM implementation similar to OpenAI's GPT-4 that has not yet been released.

Here's a brief overview of some of the most innovative projects currently making waves in the AI world:

RedPajama aims to create leading, fully open-source LLMs, kicking off with a 1.2 trillion token dataset based on the LLaMA recipe. Previously, LLaMA's license restricted commercial use, but RedPajama is set to change that, opening doors for widespread adoption and fostering a surge of creative applications.

StableLM is releasing a series of large language models under the CC BY-SA license (link). Starting with 3B and 7B parameter models, the project plans to roll out 15-65B parameter models in the near future, further expanding the reach of AI capabilities.

WebLLM brings the power of a 7B parameter model right to your browser, making LLMs even more accessible and user-friendly. This cutting-edge approach has the potential to transform how we interact with AI technology on a daily basis.

MiniGPT-4 is an ambitious project that combines LLMs with computer vision, enabling users to input images and receive relevant information or suggestions. For example, you can show it a picture of food and ask for a recipe, or upload an image of a plant to diagnose potential issues. This groundbreaking integration of technologies is set to redefine the boundaries of AI applications.

The open source LLM movement also encourages transparency and ethical AI development, ensuring a brighter future for artificial intelligence and its applications in our daily lives. Embracing this open source revolution will undoubtedly pave the way for a new era of AI innovation and impact.

Google teaches Bard to code, while prepping for significant overhaul of their iconic search engine, as major customers consider other search options.

Samsung is reportedly considering a switch from Google to Microsoft's Bing as the default search engine on its devices, sending shockwaves through Google. With an estimated $3 billion in annual revenue on the line, Google is scrambling to develop a new AI-powered search engine while upgrading its existing one with AI features in response to the growing threat from AI competitors like Bing. Under the project name Magi, Google aims to provide users with a more personalized search experience by anticipating their needs. Google has long been involved in AI research, but has been hesitant to fully embrace AI in search due to the potential for generating false or biased results.The emergence of AI-powered competitors, such as OpenAI's ChatGPT, has intensified Google's focus on modernizing its search engine. The company recently released its own chatbot, Bard, but it received mixed reviews, highlighting the challenges Google faces in maintaining its edge in the fast-paced AI industry. This week, Google released an update to Bard that will allow it to code (link). Many users are reporting that the code doesn't work (link).

Lawsuits Flying as People Realize "Data is the New Gold"

As the digital revolution forges ahead, data has become the modern-day equivalent of gold, unlocking unprecedented possibilities and challenges. Among the most contentious issues is the rise of AI impersonation, a phenomenon that has blurred the lines between creativity and deception.

There is even an emerging genre of AI-generated music that is getting major play on Streaming services like Spotify where the voices of artists like Jay-Z, Drake, The Weekend, and Kanye are "singing" on original (AI-created) tracks they had nothing to do with. The songs sound legit, and fans don't seem to care about the origin as they embrace these "new releases from their favorite stars." While there is a novelty factor here, it's not the gimmick that is driving the listens, it's that the music is REALLY GOOD.

While AI-generated content has dazzled audiences with its uncanny ability to emulate iconic figures, it has also sparked a flurry of lawsuits as individuals, companies, and even tech giants grapple with questions of authenticity, ownership, and ethics. As the world stands at the crossroads of innovation and integrity, these stories shed light on the emerging battle to define the guardrails of AI:

• Universal Music Group responds to 'Fake Drake' AI track: Streaming platforms have 'a fundamental responsibility to prevent the use of their services in ways that harm artists.' (link)

• Tom Brady threatened to sue comedians behind AI standup video (link)

• Family of F1 legend Michael Schumacher plans legal action over fake AI interview (link)

Meanwhile, some of these companies are beginning to realize the value of their data sets, and they are starting to charge AI companies to train on their data sets:

• Stack Overflow Will Charge AI Giants for Training Data (link)

• Reddit will start charging AI models learning from its extremely human archives (link)

• Elon Musk threatens to sue Microsoft for illegally training on Twitter's data (link)

More Big News

• The inside story of ChatGPT's astonishing potential - Greg Brockman, co-founder of OpenAI gives a TED Talk (link)

• Microsoft reportedly working on its own AI chips that may rival Nvidia’s (link)

• Hugging Face and AWS partner to release AWS Inferentia2 (link)

• Reimagining our video and audio tools with Adobe Firefly (link)

Smaller But Still Cool Things:

• Things you can do right now with AI that you no longer need to pay a marketer for. (link)

• AI is already taking video game illustrators’ jobs in China (link)

• Democratizing the future of education (link)

• Artificial Intimacy: How AI-Generated Pornography is Changing Society - No nudity, but NSFW (link)

Going Deeper

• Nick St. Pierre shares his guide for structuring his prompts in Midjourney (link)

• Improve your prompt skills with AI Daily's prompt engineering guide (link)

• How to use Zapier and ChatGPT to transcribe audio and then do anything (link)

• Improving Document Retrieval with Contextual Compression (link)

• LongForm: Optimizing Instruction Tuning for Long Text Generation with Corpus Extraction (link)

• Expressive Text-to-Image Generation with Rich Text (link)

• Top-down design of protein architectures with reinforcement learning (link)

• Supporting Human-AI Collaboration in Auditing LLMs with LLMs (link)

• Is ChatGPT a Good Recommender? A Preliminary Study (link)

Tweets of the Week

Eye Candy

Do you have 30 seconds for a quick survey to help us improve Everday AI?

We'd love your feedback! Click here.

Do you like what you're reading? Share it with a friend.