Innovation Profs - 4/23/2024

Your weekly guide to generative AI tools and news

Innovation Profs
April 23, 2024

Generative AI News

OpenAI winds down AI image generator that blew minds and forged friendships in 2022

Farewell to DALL-E 2. As of last week, OpenAI has sunset the image generation tool, launched in April 2022, no longer allowing new users to purchase credits to use the DALL-E 2 to create images. Savvy readers will know that DALL-E 2 has been dramatically surpassed in performance by DALL-E 3, so this move should not be surprising. Still, it does feel like the end of an era, with the quirkiness and low fidelity of DALL-E 2 images now being permanently replaced by the much slicker but often over-stylized images produced by DALL-E 3.

Mitt Romney, senators release proposal to limit potential risk of artificial intelligence

The question of the risks posed by AI has now reached the Senate, as a group of senators, including Utah senator Mitt Romney, has pass along to the Senate working group on AI a plan to mitigate potential risks posed by AI, “including biological, chemical, cyber and nuclear threats.” According to Romney, “AI has the potential to dramatically improve and transform our way of life, but it also comes with enormous risks to national security and our humanity at large. My colleagues and I have spent the last several months developing a framework which would create safeguards and provide oversight of frontier AI models aimed at preventing foreign adversaries and bad actors from misusing advanced AI to cause widespread harm. It is my hope that our proposal will serve as a starting point for discussion on what actions Congress should take on AI — without hampering innovation."

"Anthropic CEO says that by next year, AI models could be able to 'replicate and survive in the wild'"

In an interview on Ezra Klein’s podcast, Anthropic CEO Dario Amodei predicted that AI models could soon become fully autonomous and capable of replication: “Various measures of these models are pretty close to being able to replicate and survive in the wild.” When pressed by Klein as to when this might happen, Amodei gave a window of “anywhere from 2025 to 2028.” In a follow-up remark, Amodei did show some shades of doubt: “I could be wrong. But I think it could be a near-term thing.”

Microsoft’s AI Copilot Is Starting to Automate the Coding Industry

In our workshops and other training sessions, we’ve primarily discussed everyday uses of generative AI for working professionals, with a strong emphasis on no-code applications. However, on the coding front, generative AI has been significantly impactful. As described in the linked article, Microsoft’s GitHub Copilot, which provides AI-powered coding assistance, has amassed 1.3 million customers, including 50,000 businesses ranging from small operations to massive corporations. With an improvement in performance expected in the coming years, software developers will continue to assimilate AI into their workflow, but even now, there appears to be a clear payoff in using the tool: “Engineers say Copilot saves them hundreds of hours a month by handling tedious and repetitive tasks, affording them time to focus on knottier challenges.”

Adobe claims its new image generation model is its best yet

Adobe’s image generation tool Firefly has thus far not fared well against competitors like DALL-E 3 and Midjourney, but that may be changing. This week Adobe announced the release of Firefly Image 3, a new model that purportedly outperforms the previous version, Firefly Image 2 (you can see comparisons in the linked article). The tool is now available in beta mode through Adobe Photoshop, as well as Adobe’s standalone Firefly web app. New features include Generate Background, Generate Similar, and Enhance Detail, providing new ways to modify existing images and create new ones. You can learn more about the tool in a press release on Firefly Image 3 put out by Adobe just today.

Quick Hits

AI’s ability to write for us—and our inability to resist ‘The Button’—will spark a crisis of meaning in creative work

We’ll increasingly create documents mostly with AI that get sent to AI-powered inboxes, where recipients respond primarily with AI.

fortune.com/2024/04/18/ai-button-ethan-mollick-co-intelligence-book-excerpt

Hugging Face releases a benchmark for testing generative AI on health tasks | TechCrunch

Hugging Face has published a new leaderboard and benchmark for evaluating generative AI models on health-related tasks and questions.

techcrunch.com/2024/04/18/hugging-face-releases-a-benchmark-for-testing-generative-ai-on-health-tasks/?guccounter=1&guce_referrer=aHR0cHM6Ly9uZXdzLmdvb2dsZS5jb20v&guce_referrer_sig=AQAAAKrY6lqJIsaW3vJ9NaF36FYPb3dDJKdv9vsBoFBUh1P60uO7xy33h65PLGhOd9QndVyKc_J3Tec_B5NejZCVhrvHPPoI2NfAd_EyShWbAIdBKKPyG79K8ZWHJLvntP3BZ9-4S6UbLj3GHRCXnIorpbb-O6ZSbIu9eYLxcKOCDC41

To build a better AI helper, start by modeling the irrational behavior of humans

MIT and other researchers developed a framework that models irrational or suboptimal behavior of a human or AI agent, based on their computational constraints. Their technique can help predict an agent’s future actions.

news.mit.edu/2024/building-better-ai-helper-starts-with-modeling-irrational-behavior-0419

Exclusive: Google finds AI agents pose fresh ethical challenges

Giving more autonomy to AI-powered assistants offers tantalizing benefits — along with fresh ethical dilemmas we're only beginning to explore, Google DeepMind researchers say in a new paper, shared first with Axios.

www.axios.com/2024/04/19/ai-agents-assistants-ethics-alignment-google

Microsoft's new AI tool is a deepfake nightmare machine

VASA-1 can create videos from a single image.

www.creativebloq.com/news/microsoft-ai

How One Author Pushed the Limits of AI Copyright

Elisa Shupe was initially rebuffed when she tried to copyright a book she wrote with help from ChatGPT. Now the US Copyright Office has changed course—but there’s a catch.

www.wired.com/story/the-us-copyright-office-loosens-up-a-little-on-ai

AI art is only a threat if we let "prompt-jockeys" take control

Tools like Adobe Firefly need to be accepted and harnessed, not feared, or generic AI art will take over.

www.creativebloq.com/news/ai-art-is-too-generic

Why vector databases are having a moment as the AI hype cycle peaks | TechCrunch

The proliferation of large language models and generative AI has created fertile ground for vector database technologies to flourish.

techcrunch.com/2024/04/20/why-vector-databases-are-having-a-moment-as-the-ai-hype-cycle-peaks

Tool of the week: Meta AI

Meta AI assistant is now integrated into the search box of Instagram, Facebook, WhatsApp, and Messenger. AI assistant uses a version of Meta's Llama 3, although a better version is coming in the future.

The new tool integrates real-time search results from both Bing and Google. It can write both text and create images (see example below).

Look for it in the search bar of Meta apps or use it at meta.ai.

AI-generated image of the week

Meta’s new AI image generation tool is available at meta.ai. Two features make this one interesting. First, it will create images as you are typing. Here’s what it made for the prompt below before I had even pressed the return button (before I typed a unicorn it already showed a boy on a motorcycle):

And here are the actual results:

Prompt: imagine a boy riding a unicorn photo

But wait, there’s more… The animate button will turn all four of your results into animated GIFs like this one (it actually dowloaded as an MP4 and I converted it to a GIF):

Generative AI tip of the week:

Did you know that ChatGPT can create a Microsoft Excel document for you? You will need to tell it what you want on the document and then to create the document. I asked it to create a document to track spending.