Innovation Profs - 2/13/2024

Your weekly guide to generative AI tools and news

Our next Introduction to Generative AI Workshop is March 14 at Drake University (and virtually). Attendees will learn about current generative AI tools and gain ideas of how those tools can be used in their professional lives. We'll also discuss how they should not be used, terms of service and ethical issues related to AI tools. Sign up by Feb. 29 to get the early bird rate.

Generative AI News

Google debuts more powerful “Ultra 1.0” AI model in rebranded “Gemini” chatbot

Last week we reported that Google was planning to rebrand Bard as Gemini. This week, Google made the name change official. In addition, Google released its most sophisticated Gemini model, Ultra 1.0, which is now available as part of a Gemini Advanced subscription. In keeping with ChatGPT Plus and Microsoft CoPilot Pro, a Gemini Advanced subscription costs $20 per month.

For those of you having trouble following these Google AI developments, here’s a recent timeline:

  • Google Bard announced in February 2023 and released in March 2023, powered by the language model LaMDA.

  • Google Bard upgraded to run on the model PaLM 2 in May 2023.

  • Google Bard upgraded to run on the model Gemini Pro in December 2023.

  • Gemini Nano and Gemini Ultra announced in December 2023.

  • Google Bard rebranded as Google Gemini in February 2024.

  • Google Gemini Ultra launched as part of a Gemini Advanced subscription in February 2024.

Apple releases ‘MGIE’, a revolutionary AI model for instruction-based image editing

Apple has entered the generative AI conversation! Apple researchers have released a new image-editing tool called MGIE, which stands for “MLLM-generated image editing” (where MLLM stands for “multimodal large language model”). MGIE allows users to edit images via text prompts. Want a bluer sky in your image? Just ask for the sky to be bluer. Users can similarly crop, resize, or add filters to images via text prompts. Note that MGIE is not a commercial tool but rather has been made available as open-source on GitHub.

Microsoft brings new AI image functionality to Copilot, adds new model Deucalion

Ever heard of Deucalion? According to Greek mythology, he’s the son of Prometheus and the Greek counterpart of the biblical Noah, who helped repopulate the world after a global flood. Deucalion is also the name of a language model that underlies Microsoft Copilot’s “Balanced” mode, which is the middle ground between the “Creative” and “Precise” options that users can choose when feeding prompts to Copilot. In a post on X, Microsoft’s Jordi Ribas, Head of Engineering and Product for Bing and Copilot, described Deucalion as “a fine tuned model that makes Balanced mode for Microsoft Copilot richer and faster.”

Meta says it will identify more AI-generated images ahead of upcoming elections

Previously, Meta has been labelling as AI-generated any image on its platform that was generated by its own AI image tools. Now, in anticipation of the 2024 presidential election and other elections worldwide, Meta will start labelling images generated by any AI-powered image generation tools, including those from Google, OpenAI, Microsoft, Adobe, Midjourney, and Shutterstock. According to Nick Clegg, president of global affairs at Meta, “We’re working hard to develop classifiers that can help us to automatically detect AI-generated content, even if the content lacks invisible markers.”

AI Companies Take Hit as Judge Says Artists Have “Public Interest” In Pursuing Lawsuits

In response to a class action lawsuit of visual artists against AI image generation tool providers Stability AI, Midjourney and DeviantArt, the companies offered a First Amendment defense that the lawsuit was targeting their free speech. Just this past Thursday, U.S. District Judge William Orrick rejected this argument and further found the case to be in the public interest. Back in October, this same judge had dismissed a number of claims that the artists brought against the companies, including claims of copyright infringement and breach of contract.

Quick Hits

Tool of the week: Google Gemini AI for Mobile Devices

Along with rebranding Bard as Gemini and releasing Gemini Ultra, Google has also made its Gemini AI available as an app for both Android and iPhone.

AI-generated image of the week

Hands are still sometimes a challenge for image creation tools. As is counting. So we decided to test Midjourney’s ability to do both by creating an image celebrating the Kansas City Chiefs’ fourth Super Bowl victory.

Prompt: football player's hand wearing four kansas city chiefs super bowl rings

Generative AI tip of the week

This week’s tip is a bit more technical than the ones we typically share. Retrieval-augmented generation (RAG) is a way to augment language models with an external knowledge base to provide up-to-date responses and improve accuracy. Learn more about RAG in the article linked below!

Get starting with Generative AI

New to generative AI? Here are some places to start…

What we found

We’re torn about whether the story below is fact or fiction. It’s true that ChatGPT can help boost productivity, but using it to navigate the dating scene probably isn’t one of the use cases most of us have in mind…