Innovation Profs - 9/26/2023

Generative AI News

OpenAI’s new AI image generator pushes the limits in detail and prompt fidelity

DALL-E 3 is coming! Fully integrated with ChatGPT, DALL-E 3 reportedly will be a vast improvement over DALL-E 2, as well as all other currently available image-generating AI tools. New features include better performance on small details such as hands, images more aligned with the prompts on which they’re based, and the ability to better handle text within images. DALL-E 3 is also “designed to decline requests that ask for an image in the style of a living artist.” ChatGPT Plus and ChatGPT Enterprise users will have access to DALL-E 3 in early October.

ChatGPT Is Growing Eyes and Ears to Better Respond to Your Human Whims

Just yesterday (9/25), OpenAI announced that they’ll be adding new functionality to ChatGPT, namely image recognition, which will allow for users to engage in a conversation with ChatGPT about a given image, and several speech synthesis tools that will allow for text-to-speech and speech-to-text capabilities (these latter tools will only be available on the ChatGPT mobile app). Only ChatGPT Plus and ChatGPT Enterprise users will have access to these features.

In Artificial Intelligence Race, 'Claude' Just Got a $4B Boost

The LLM arms race continues, as Amazon just invested $1.25 billion in Anthropic, the startup that created the AI tool Claude, with an option to increase that investment to $4 billion. Alphabet, the parent company of Google, has also invested $300 million in Anthropic this year. As part of the deal with Amazon, Anthropic will use Amazon cloud services and Amazon chips for its software.

See How AI Generates Images from Text

The October 2023 issue of Scientific American features a helpful and brief explanation of how image-generating AI works. Take a look!

George RR Martin and John Grisham among group of authors suing OpenAI

A few weeks back we included a story about a group of authors, including Sarah Silverman, that was suing OpenAI over alleged copyright infringement. A similar lawsuit against OpenAI has been filed by the Authors Guild, which represents over 9,000 authors, along with 17 authors, including George RR Martin and John Grisham. Authors Guild CEO Mary Rasenberger had these strong words to say, “It is imperative that we stop this theft in its tracks or we will destroy our incredible literary culture, which feeds many other creative industries in the US.” We’ll continue to watch for and pass along any new developments on these lawsuits.

How Google taught AI to doubt itself

Google has just added a new feature to its Bard conversational AI tool, allowing users to “double check” a response given by Bard. Those parts of a response that can be substantiated by results on the web will be highlighted in green and can be clicked to access the supporting material, while those results that remain unsubstantiated will be highlighted in brown. It remains to be seen how effective this and similar features will be in addressing the tendency of LLMs to be “confidently wrong.”

Generative AI Workshop Oct. 17

Our next Generative AI Workshop will be Oct. 17 at Drake University (you can attend in-person or virtually). Come learn how to use generative AI tools ChatGPT, Midjourney, D-ID and more. You can attend in-person or virtually. Sign up here.

Know someone who wants to learn more about generative AI? Forward them this newsletter or send them this link to sign up.

Tool of the week: Google Bard

Just this week Google announced Bard Extensions, which allows Bard to work with a range of Google tools, including Gmail, Docs, Google Maps, and YouTube.

Innovation Profs Homework

Create an image in Midjourney. For the longest time, I had no success using the free version of Midjourney. But last week a few of my colleagues were able to make the image creation tool work without paying.

Your homework this week - if you have not done this already - is to create an image in Midjourney. Follow the instructions here (but skip No. 2 that says to subscribe to a plan). Not sure what to make? Here are some prompts to try.

Reply to this email and let us know if you were able to get the program to work for free. And send us what you created. See what we created below…

AI-generated image of the week

The entire world (or at least our social media feeds) are filled with content about Kansas City Chiefs tight end Travis Kelce and superstar singer Taylor Swift. What we don’t have are many photos of the two together. Midjourney to the rescue.

Which of these images do you like best? And what happened to Taylor in the second one?

Prompt: Taylor Swift and Travis Kelce eating ribs in a Kansas City restaurant, happy, photo, 8k, realistic --ar 16:9

Generative AI tip of the week

Chain of Thought prompting is a prompting technique in which the user provides an LLM with one or more examples of a specific reasoning process, with the aim of inducing the model to follow that same process on a different example or set of examples. You can learn more about Chain of Thought prompting at the link below.

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

What we found

X (formerly Twitter) user @EddIe4Dimension created these Nike branded images using a combination of Stable Diffusion and ControlNet. ControlNet allows you to add extra conditions to AI images such as a mask. We’ll be on the lookout for a tutorial to share with you on how to do this.