WN #20

OpenAI’s Spring Update

Since OpenAI burst onto the scene with the release of ChatGPT, the rest of the industry has seemingly been playing catch up. OpenAI’s models have become a benchmark standard against which all other models are judged, and they have made a habit of being the first to release new capabilities to the public. As a leader in the AI space, when OpenAI announces updates, it grabs the attention of the entire industry. This week’s hyped “Spring Update” was no different. Here’s what they announced and how it affects you as an educator:

First, OpenAI’s flagship model, GPT-4, which powers ChatGPT’s premium offering, has gotten an upgrade. Dubbed “GPT-4o”, the new and improved model is not only faster than GPT-4, it improves its text, voice, and vision capabilities as well. Don’t have a premium subscription to ChatGPT? No worries, because OpenAI has made almost all of ChatGPT’s premium features available to free users (up to a usage limit). This includes:

• GPT-4o - access to a top-of-the-line AI model

• Vision - upload images for analysis, interpretation, transcription, etc. (See “The Lab” section below for a few education-specific examples)

• File uploads - upload text documents, PDFs, spreadsheets, etc.

• Data analysis - Generate charts, visualizations, summaries, interpretations, insights, etc.

• Web search - Ask it to search the web for up-to-date info rather than be stuck with a knowledge cut-off date.

• Custom GPTs - Use customized versions of ChatGPT that have been optimized for specific tasks (available for free through the GPT Store).

(Note that image generation, creation of Custom GPTs, and higher usage limits are still premium features at the moment.)

In addition to rolling out GPT-4o and making its premium features more accessible, OpenAI also announced major upgrades to voice and vision capabilities that will be coming in the next few weeks. Though free users already had access to voice conversations, the new version will be able to do things like provide real-time translation and use a more human-like range of tones. It will also be able to see and discuss images you upload, as well as real-time video from your phone’s camera or from your desktop if you are a Mac user and use the new macOS ChatGPT app. This latest feature has perhaps the most profound potential to impact education, as seen in this video of a student completing his math homework with ChatGPT following along as he works.

Regardless of how you feel about these new features, they will be accessible to students and educators alike, so we must be aware of them and their potential implications. We play an important role as educators in helping ensure these tools are used to enhance rather than impair education. Jane Rosenzweig, Director of the Harvard Writing Center, said it best:

“Students need to learn ABOUT these tools and not just be handed access to them.”

The same goes for educators.

AI literacy will go a long way towards determining how helpful or harmful new AI features may be to the education space.

Reflecting Upon AI’s “Freshman Year”

The Atlantic put together a collection of student and educator insights, perspectives, and experiences from “AI’s Freshman Year.”

The 6 Types of Conversations with Generative AI

There are plenty of ways to prompt and use an AI chatbot, and it’s possible that two people may have completely different experiences with the same chatbot because they are using it differently. This article provides some helpful ways to think about our interactions by breaking down the 6 types of conversations people have with generative AI, along with some tips and insights for each.

The Rapid Evolution of AI Music Generation

AI song generator Suno recently got a new update, making their technology even more impressive, though they also gained a competitor with the launch of Udio. Many artists worry about the implications of AI song generation, but there are some who are ready to embrace it. Rolling Stone recently did a deep dive on Suno if you are up for a read. This is a great topic for classroom discussions on AI ethics and implications.

📌 Mark Cuban’s Free AI Bootcamps for High Schoolers

Mark Cuban wants to give high schoolers across the US the opportunity to boost their AI literacy and knowledge. His foundation is putting on free AI bootcamps this November in 18 cities across 11 states (CA, FL, GA, IL, IN, MI, MN, MO, NE, PA, TX) and Washington, DC. To apply, students just need to be in the 9th-12th grade next school year and live within driving distance of the site. No prior knowledge or experience needed. Apply here.

📌 Learning Opportunities for Educators

Looking to build your own AI knowledge? Perplexity, a popular AI chatbot that searches the web and cites sources, recently launched a short, online course to help educators make use of their platform. Similarly, Adobe launched a course to show how their generative AI features can be used in the classroom. You can also check out Google’s “Generative AI for Educators” course or Microsoft’s “AI for Educators” modules. All of these courses are free and self-paced.

📌 Meta’s New AI Model

Meta released its latest AI model, Llama 3, which you can chat with for free over at meta.ai or on WhatsApp (which is a big deal if you know how popular WhatsApp is worldwide). One cool feature is real-time image generation, which you can check out below:

Utilizing AI’s Vision Capabilities

With vision capabilities now accessible to free users of ChatGPT (and also available on other platforms like Claude, Copilot, and Gemini), I wanted to give a few examples of how you might use this feature as an educator.

This first example does a good job of demonstrating the overall capabilities of the vision feature. I uploaded a “one-pager” a student created about the French Revolution with a lot going on. I gave ChatGPT a rather vague prompt, but the purpose was just to see how well it could pick out various elements. It had no problem.

I also like using the vision feature as a quick way to give context to a conversation. In this example, a fellow educator was asking her followers on X for advice on turning standards into aligned learning targets across grade levels that are also scaled on a level of mastery. This seemed like something AI would be able to help with, so I simply took a screenshot of the thread from X, uploaded it to Claude, and asked it to help.

And don’t forget you can combine capabilities as well! Moving back to ChatGPT, in this example I used both the vision and data analysis features together. I first uploaded an image of a data table (monthly rainfall totals and average storm totals at LAX in the 90’s), then asked ChatGPT to visualize it.

More Ideas…

Do you use PowerPoint? Here’s how Microsoft’s Copilot AI can help you with your presentation. (read about it) (watch a Youtube video)

• This Polaroid-style camera uses AI to print out poems in place of photos, while this company wants to use AI to make photos out of your memories.

• Entertainment giant A24 used images that were clearly AI-generated to promote its new film, “Civil War,” and people were not happy about it. With a production budget of around $50 million, the promotional images are shockingly amateur.

• While A24’s AI images are getting jeered, other AI generations are being celebrated, such as this exhibition at an art gallery in Beverly Hills earlier this year that showcased images generated by OpenAI’s DALL-E image generator.

• “Beware of Dog” signs just got a little scarier.

That’s all for this week! If you appreciate the content in this newsletter, consider subscribing for free, or sharing with people in your network who may find value in it. If you are looking for more, feel free to check out the archive to catch up on any editions you may have missed.