Google Gemma, Gemini's Bug, OpenAI's Chinese Competitor, Windows Magic Eraser, and MORE!

PLUS HOT AI Tools & Tutorials

Feb 23, 2024

👋 Hey, I’m Stepan and welcome to a ✨ news edition ✨ of Creators’ AI. By subscribing, you directly support Creators' AI's mission to deliver top AI insights & practical knowledge without ads or clutter. Your subscription allows us to grow our dedicated team and curate the most important AI Tools, Stories, and Tutorials in one place. - Stepan

Welcome to our newsletter about the top news from the AI world! Compared to what's been happening over the last month (think OpenAI Sora, new Gemini updates, ImageFX & MusicFX, and much more!), this week has been relatively quiet. Nevertheless, we've gathered a few essential things for you and are ready to discuss them.

So today in today's newsletter: Google announces the Gemma 2B and Gemma 7B, Gemini loses its image generation feature, Windows gets its own AI for photo editing and more, plus valuable tools, a meme, and a tweet of the week. Let's go!

Are you laggard or early adopter? Find out your AI Adoption score!
Are You an AI Early Adopter or Laggard?
Stepan Ikaev
·
Feb 19
Read full story

This Creators’ AI Edition:

Featured Materials 🎟️
News of the week 🌍
Useful tools ⚒️
Weekly Guides 📕
AI Meme of the Week 🤡
AI Tweet of the Week 🐦
(Bonus) Materials 🎁

And try our new quiz! It will show your score in AI Adoption!

QUIZ

Featured Material 🎟️

Google launched two new open LLMs: Gemma 2B and 7B

The word “Gemma” and a spark icon with blueprint styling appears in a blue gradient against a black background.

A week after the launch of the latest Gemini version, Google announced Gemma, a new family of lightweight, state-of-the-art open models. We're talking about the Gemma 2B and Gemma 7B models. They were "inspired by Gemini" (because they are based on the same research and technology) and are available for commercial and research use.

Google claims that despite their smaller size, the 2B and 7B "significantly outperform larger models in key criteria" and are "capable of running directly on a developer's laptop or desktop computer." Based on the images from Google, it appears the company compared the platforms to Meta Platforms' Llama-2.

Users can fine-tune Gemma models based on their own data to tailor them to specific application needs, such as summarization or retrieval-augmented generation (RAG). Google noted that Gemma supports a wide range of tools and systems, from cloud services to Nvidia hardware platforms.

Notably, both models are offered under a commercial license regardless of the size of the organization, number of users, or type of project. However, like other major IT companies, Google restricts some usage scenarios. This applies to weapons development programs, viruses, and other dangerous things.

To ensure that Gemma is used for the public good, Google will also provide its users with "responsible artificial intelligence toolkits." According to Tris Warketin, director of product management at DeepMind, it's more challenging to place guardrails in open-source models than in more closed systems like Gemini.

The Responsible Generative AI Toolkit will allow developers to create their own guidelines or list of forbidden words when deploying Gemma in their projects. It also includes a model debugging tool that enables users to examine Gemma's behavior and fix problems.

Gemma 2B and Gemma 7B are available through Kaggle, Hugging Face, Nvidia's NeMo, and Google's Vertex AI. You can learn more about Gemma and access it on the model page. Researchers can also apply for Google Cloud credits of up to $500,000 to accelerate their projects.

Can you feel it? A wave of AI products built on Google's new models is coming. Let's wait a bit and see what's in store for us.

If you want to see what Gemini Advanced can do, here is the newsletter for you:
Our Gemini Test Drive: Honest Review
Stepan Ikaev
·
Feb 20
Read full story

News Of The Week 🌍

Google has disabled the Gemini features responsible for images

A bug has been spotted in Gemini just a week after the release of the new chatbot. According to CNBC, users on social media complained that the artificial intelligence tool was generating images of historical figures, such as the US founding fathers as people of color, calling it inaccurate.

Gemini now refuses to generate images on demand (though our tests showed the chatbot still generates images on requests for third-party topics). Google promises to refine its AI and release an improved version soon.

But at least the Gemini problem has caused a bunch of memes:

Hundreds of AI luminaries sign letter calling for anti-deepfake legislation

Members of the AI community are signing an open letter calling for strict regulation of AI-generated fakes and deepfakes. According to the statement, signed by more than 500 people working in machine learning and related fields, "deepfakes are a growing threat to society, and governments must impose obligations across the supply chain to stop the spread of deepfakes."

Among the signatories were prominent opinion leaders, as well as an OpenAI employee and several people from Google DeepMind. The letter does not mention Anthropic, Amazon, Apple, or Microsoft representatives.

Reddit says it’s already made $203M so far licensing its data

Reddit appears to be reaping impressive benefits from its partnerships with AI developers. The company's IPO prospectus filed with the U.S. Securities and Exchange Commission outlined how it benefits other businesses by licensing data.

First up is big tech, which uses more than 1 billion posts and more than 16 billion comments to develop chatbots and other models. In January of this year, Reddit entered into contracts with terms of two to three years totaling $203.0 million.

Windows is getting its own Magic Eraser to AI-modify your photos

The Photos app built into Windows 11 and 10 is being updated with a new "Generative erase" that uses AI to remove people and objects in images. If you've been following Android updates, you know that similar features have recently been implemented into Samsung and Google smartphones.

This tool allows you to erase unnecessary objects from a photo to make it more beautiful. But while it used to require a new smartphone from their Galaxy or Pixel lineup, now you just need a PC with an up-to-date version of Windows. Currently, Generative Erase is only available for developers, but other users will get it later.

OpenAI's competitor from China? Moonshot AI raised $1B for its LLM

Moonshot AI, an artificial intelligence startup founded less than a year ago that builds LLMs capable of handling long text and data input, has raised over $1 billion in a Series B round. If the information is confirmed, the latest infusion of capital values Moonshot AI at $2.5 billion, the largest funding for Chinese LLM developers.

Moonshot AI focuses on the development of large language models. Its main feature is that the startup is working on the ability to handle deployed context and responses. Last October, Moonshot launched its first chatbot, Kimi, which can support processing 200,000 Chinese characters in a single conversation.

Share this post with friends, especially those interested in AI stories!

Useful Tools ⚒️

Vadoo AI – Search widget trained on your videos

Dreamwriter – AI-powered content creation

Kraftful 2.0 – Copilot for product teams

Recaster AI – Transform your e-commerce into an intelligent commerce

Decktopus AI 2.0 – AI presentation assistant for teams

Decktopus calls itself the World's #1 AI-powered Presentation Generator. This platform works as a chatbot that creates presentations based on short descriptions. During the sign-up process, you choose the audience you want the presentation for and its visual style, and then you can enter a short query and get the finished project.