Innovation Profs - 5/7/2024

Upcoming Innovation Profs Workshops

Professors Porter and Snider are teaching a one-day Intro to Generative AI Workshop at Drake University (or virtual) on Friday, June 7. Attendees will get hands-on with tools for creating text, images, audio and more. We will also discuss our framework for deciding how to deploy generative AI at your company. Sign up here.

Drake Professor and AI expert Eric Manley will lead a workshop on Integrating Large Language Models into Your Applications on May 28. This hands-on workshop for developers will introduce how to build language-aware applications in Python. Sign up here.

Drake Professor and data visualization expert Tim Urness will lead a Data Visualization Workshop on May 30 at Drake University (or virtual). Urness will teach you how to tell stories with with data, create dashboards, and make use of color, charts, and graphs effectively. Sign up here.

We’re also hosting a free lunch-and-learn on what’s new in generative AI on May 14. Sign up here.

Generative AI News

OpenAI and Stack Overflow partner to bring more technical knowledge into ChatGPT

OpenAI and Stack Overflow have formed a mutually beneficial partnership, with OpenAI gaining access to Stack Overflow’s AI along with developer feedback that will be used to improve model performance, while information and links from Stack Overflow will be included in ChatGPT’s responses to user prompts in response to coding-related questions. In addition, Stack Overflow will use OpenAI’s models to support Overflow AI, Stack Overflow’s own generative AI tool. The partnership aims to “foster deeper engagement with content” for both companies.

Microsoft says it did a lot for responsible AI in inaugural transparency report

Last week, Microsoft released its Responsible AI Transparency Report, covering its efforts towards developing and implementing responsible AI in 2023. Among Microsoft’s contributions highlighted in the report:

• developing 30 responsible AI tools;

• growing its responsible AI team;

• requiring a robust understanding of the risks associated with any new generative AI applications;

• watermarking AI generated images via its Content Credentials metadata; and

• providing Azure AI users with tools to detect harmful content.

Warren Buffett compares AI to nukes after seeing deepfake doppelganger

At the Berkshire Hathaway annual meeting on Saturday, Warren Buffett weighed in on the topic of AI during a Q&A session. Initially, Buffett stated, “I don’t know anything about AI, but that doesn’t mean I deny its existence or importance or anything of the sort.” However, he followed up this remark by comparing the potential misuse of AI in the financial sector to the proliferation of nuclear weaponry, a remark that apparently was tied to Buffett’s recent exposure to a deepfake of him that someone had created. Buffett continued: “I mean, scamming has always been part of the American scene, but this would make me, if I was interested in investing in scamming, it’s going to be the growth industry of all time. And it’s enabled in a way. Obviously, AI has potential for good things too but I don’t know how you, based on the one I saw recently, I practically would send money to myself over and over in some crazy country.”

Randy Travis gets his voice back in a new Warner AI music experiment

Back in 2013, Randy Travis lost his voice as the result of a stroke, but this hasn’t stopped him from producing new music using his own voice. “Where That Came From,” a newly released track, was created by having a different country artist, James DuPre, initially lay down the vocals, and then using an AI model, trained on 42 of Travis’ vocal-isolated songs, to transform DuPre’s voice into Travis’ voice. The result? You can judge for yourself by watching the video in the linked article.

Meta’s new multi-token prediction makes AI models up to 3X faster

Traditionally, large language models predict one token (roughly, a word) at a time in producing their outputs. What happens if we allow LLMs to predict multiple tokens at a time? Meta researchers, working in tandem with a team of researchers in France, found that in some contexts, this yielded a significant improvement in performance, tripling the speed of the model along with improved accuracy. It will be interesting to watch how new research developments in large language models will impact the everyday use of generative AI tools.

Quick Hits

Tool of the week: Fathom

The days of taking minutes or notes during meetings may be over. Fathom is a free tool that records, transcribes, highlights, and summarizes your meetings so you can focus on the conversation.

Give it a try at fathom.video.

Innovation Profs Homework

It’s been quite a while since we’ve assigned any homework in our newsletter.

We love creating this newsletter every week, but we need your help making it grow. Your homework this week is an easy one. Share this newsletter with someone you think would enjoy a weekly dose of generative AI news.

AI-generated image of the week

The Innovation Profs are hard at work preparing for our May 14 Hot GPT Summer virtual event. Sign up for free to learn what’s new in generative AI from some guys that sort of (maybe) look like this:

prompt: award-winning photograph of two 35-year-olds males - one with gray hair and short beard and one with brown hair and short beard - sitting in lawn chairs on the beach, wearing hawaiian shirts and shorts, working on their laptops, wide shot, Kodak Porta 800 Film --s 750 

Generative AI tip of the week: Video editing

Need to edit a video but don’t know how? Our favorite video editing tool is Descript. You simply upload the video, wait for the transcript, and delete the words you don’t want from the transcript. Descipt does the editing for you.

Get starting with Generative AI

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

• NightCafe image prompting tips

• Learn ChatGPT prompting

• Best AI image generators

• AI terms everyone should know

• Prompt Engineering: What to Know and Why It's Important 

What we found

This week we have two items of interest: a Sora-generated music video for the band Washed Out, and a report of President Biden’s first use of ChatGPT last spring.