When Access to AI Outpaces Attention
It’s been another week of exhausting major AI deployments, so I thought I’d spend a few minutes doing a run down of some of them. But this is a list with a point—I don’t think we’re talking about the right things when we discuss equity and access to AI systems. No one has the bandwidth to keep up with the pace of these deployments and that makes the access discussion moot, at least until we get a slowdown and a period of stability.
You can listen to an AI generated NPR-like discussion of this post below. I used the “pasted text” of this post as a reference point and loaded it into Google’s NotebookLM. Funny how it referred to it as just that—pasted text. I then used the new Audio Overview feature to generate a podcast with two AI voices discussing the article.
HeyGen’s URL to Video Feature
You might have seen the AI avatars out in the wild on socials and some of what this technology is capable of is simply wild. I stumbled on HeyGen Labs' suite of use cases and was pretty shocked at how far along the technology is. I used their URL to Video feature to create a trailer of sorts for a presentation I did about AI’s impact on reading.
Generating it was simple. I used my previous post No One is Talking About AI’s Impact on Reading as the source before editing the output and making it fit the needs of my presentation. It probably took me longer to write down this explanation than it did for me to generate the video. It’s not perfect by any means, but I cannot believe this is where we are now just 22 months from the public release of ChatGPT.
AI Political Misinformation
How far have we fallen in our public discourse, when a presidential candidate uses generative AI to amplify a disgusting and racist rumor? Social media has made us collectively jaded to the point that we laughed when Trump used the debate and his socials to promote the widely debunked claim that Haitian immigrants in Ohio were stealing and eating pets. That’s not to say the left is any less gleeful when spreading rumors about JD Vance’s amorous encounter with a couch. But if you cannot see the difference between the two, then there’s little I can do to sway you.
Trump actively used fake images to malign Harris as a communist, but I never dreamed he and his campaign would have leaned in so far into deep fakes to support the misinformation about immigrants eating people’s pets.
OpenAI Releases Their Newest Model
OpenAI released a new “reasoning” model, one that may be more of an improvement in technique rather than a technical breakthrough. The new model is part of a series of large and “mini” language models called o1. The breakthrough is this new model uses some version of chain-of-thought prompting each time you query it. In my limited tests, the new model takes quite a bit longer to produce an output, often cycling between a dozen or more iterations behind the scenes before generating an output. It’s now in private preview for paying customers.
Somewhat refreshing was the hype-lite Tweet from Sam Altman upon announcing the release of o1:
Hume.AI Tries to Recreate “Her”
Hume.AI released a major update to their Empathic Voice Interface and in the process chose to mimic some of the most cringe-inducing scenes from OpenAI’s demo of their advanced voice model. Seeing three tech bros hovering over a cell phone while a synthetic voice talks with them with a feminine and flirtatious tone is nauseating.
From a technical standpoint, EVI 2.0 offers more personas to chat with and lower latency in its response. It is certainly an improvement and I’m sure we’ll see movement toward putting this into the hands of students as a tutoring assistant, but as I wrote about before, a synthetic stand-in for empathy isn’t something that I’d want to interact with.
Google’s NotebookLM Update Turns PDFs into Podcasts
This is an odd update, but one I like on some levels. Google Labs has quietly been developing a slew of products and it looks like NotebookLM is starting to become one of their main testbeds for many of these emerging use cases. The most recent update allows a user to upload a document or links and allows Google’s Gemini to summarize and synthesize the information. With one extra click, you can then turn those synthesized summaries into a generative podcast, like NPR.
I honestly don’t know what to think about this. From my experience, it’s not like students rush to podcasts when I’ve assigned them as options, so giving them a reading they aren’t particularly excited about and having them use AI to generate a podcast may not be the best tool to engage a young student. But, I do see some potential here. Asking a FYW student to include three different peer reviews of a rough draft they’ve written, then listen to an AI overview and consider if the output helped them synthesize commentary or gave them reason to push back, could be a powerful pedagogical tool.
The Big Picture: What Does Equity & Access Mean in the Tsunami of AI Deployments?
I’m serving as a faculty mentor for AAC&U’s AI Institute and it has been wonderful hearing from other faculty at different institutions talk about how they are dealing with AI in education. Equity and access come up time and again, but what does access truly matter when you don’t have time to process any of these changes or use cases? The well-meaning intention behind the discourse is to give equitable access to as many learners as possible. However, that side steps the elephant in the room--we are all involved in a massive public experiment no one asked for.
It's as if the world suddenly gave everyone access to a powerful, multi-functional vehicle that can be a car, boat, or airplane at will. We've done this without providing proper training, establishing traffic laws, or considering the broader implications on our infrastructure and society. Educational institutions and individuals are scrambling to adapt, but the rapid pace of change makes it nearly impossible to keep up with the evolving landscape.
Some of these developers are treating AI deployments like iPhone releases and don’t quite grasp that the world isn’t too keen on adopting these tools at that pace. This artificially creates a massive equity issue that runs far deeper than simple access. The goal of these developments has nothing to do with actual adoption and is far more related to creating excitement to raise capital.
That’s a recipe for continued chaos. Giving everyone access to the tools without any critical framework or time to process how these use cases impact daily life transcends the equity discourse and is the conversation no one appears to want to tackle. Because doing so inevitably leads to questions about regulation and anything that makes it more challenging for these companies to raise money is simply a nonstarter. I think it is time to push back on that.
This point nails the central problem: "we are all involved in a massive public experiment no one asked for." And at the same time, you helpfully digest and feature some of the most dramatic developments in AI (for just this week!!). One point I and others made at the AAC&U AI Institute was that it's impossible to keep up, even for those of us spending hours a day attempting to do so! Your critique are your practical guidance are both useful in navigating this rapidly shifting terrain.
I feel exhausted after reading this post. 😅