To draw a parallel with social media: if GPT3.5 is Facebook and GPT4 is Instagram, then GPT4o is definitely TikTok. This tool is ideally positioned not only to instruct but also to sell products, shape minds, and manipulate real world events.
Before I begin, I want to thank my readers who have decided to support my Substack via paid subscriptions. I appreciate this vote of confidence. Your contributions allow me to dedicate more time to research, writing, and building Educating AI's network of contributors, resources, and materials.
1. ChatGPT4o: Everywhere, Cheap, and Accessible
It's undeniable—OpenAI has taken another leap forward in the consumer AI market. Just this past Monday, I noticed an unexpected update while logging into my ChatGPT4-Turbo account. It reminded me of when GPT-2 suddenly appeared and outperformed competitors like Gemini, Claude, and ChatGPT, and it all started to make sense.
As I experimented with GPT4o, the first things that struck me were its remarkable speed, streamlined interface, and compelling persona. While testing some editing prompts, I didn't notice any dramatic improvements in intelligence, but the increased speed and new interaction modes certainly gave the impression of enhanced cognitive abilities.
Exploring its "omni-modality" feature, I observed significant strides made by OpenAI. Although switching between vision, text, and sound could be somewhat clunky, the audio-to-text modality was impressive. The ability to interrupt GPT4o mid-speech and the improved voice quality seem to contribute to an overall sense of advanced cognition.
In terms of accessibility, ChatGPT is aiming to reduce barriers and maintain its position as the preferred AI model globally. It costs half as much per token as GPT-4, and no longer requires an email and phone number to access, aligning with Perplexity's model. This may well be a strategic move as the company prepares to launch its AI search engine.
The underlying message is clear: OpenAI represents AI that is ubiquitous, affordable, and accessible.
Personally, when I'm not thinking as an educator, I find these changes beneficial. It's great not to pay more per token or to provide personal information unnecessarily.
2. ChatGPT4o: Unrestricted, Persuasive, Entertaining
However, when I put my teacher hat back on, I am met with several questions and concerns, driven by a vibrant discourse that unfolded quickly among the Substack AI and Education communities in response to these new models. Their insights and inquiries were not just about the capabilities of GPT4o or OpenAI, but also touched on broader, more holistic concerns about how we, as a society, are engaging with the rapid evolution and dissemination of AI technologies.
One significant area of concern involves access and safety. As a K-12 teacher, I find the accessibility of GPT4o troubling, as anyone with internet access can now use it. The irony here is palpable. We in the educational community have long emphasized safety and privacy with the advent of earlier models like GPT 3.5, and while it's beneficial that users no longer need to submit key pieces of identity information to use the technology, this change ushers in new, arguably more complex issues.
Another pressing issue concerns the long-term goals of OpenAI in the ed-tech space. Observers like Dan Meyer have pointed out that the promotional pitch for GPT4o emphasized its capability to solve educational problems. On the same day as the GPT4o launch, Google introduced a suite of Ed x AI software, suggesting that big money is anticipated in this sector. The market seems to be dividing between general versus specific tools. However, what OpenAI does not fully appreciate is that teachers desire general tools that are also safe and secure.
Lastly, there's the issue of overreliance on AI. As these systems become more lifelike, our interactions with AI will inevitably extend. Putting aside whether OpenAI could upgrade to GPT5 in this particular moment, the company has currently focused on making an application that is more persuasive—or, perhaps more accurately, more seductive.
To draw a parallel with social media: if GPT3.5 is Facebook and GPT4 is Instagram, then GPT4o is definitely TikTok. This tool is ideally positioned not only to instruct but also to sell products, shape minds, and manipulate social scenarios. Pair this tool with an AI search engine and an AI payment app, and the potential impact is staggering. Think the next Amazon!!!
Given these complexities and the ongoing evolution of AI, it's imperative that we as educators continually re-evaluate our approach to these technologies in the classroom.
3. K-12 Pause on Use of ChatGPT for Classroom Instruction
So, what do we do in K-12?
Do we give in to the allure of an unrestricted, free AI tool that increasingly resembles a social media influencer? Do we dive into OpenAI’s business plan despite not knowing its long-term trajectory?
Here are some possibilities I am considering at this moment:
OpenAI is trying to onboard schools and universities to ChatGPT4o before introducing its next line of specialized education-oriented products at premium prices. Think of something akin to Google’s LearnLM suite. This is highly unlikely given OpenAI’s commitment to general purpose models and the development of AGI.
OpenAI is attempting to increase its capture of data, inputs, and outputs in an effort to overcome its training set limitations that are hindering the development of GPT-5. Opening up access free to all will allow OpenAI to gather greater and greater quantities of data in order to push their product to the next level. OpenAI’s alignment with Reddit is another prong of this same strategy with Reddit representing a data goldmine for AI training.
Despite appearances, OpenAI is not prioritizing education. Instead, it is positioning its product as education-friendly to compensate for its increasing disregard for safety, security, and privacy. Notably, Ilya Sutskever, Chief Scientist of OpenAI, chose to leave the company this week concurrently with the announcement of ChatGPT4o. Sutskever’s departure marks the final step in a long process of the company’s abandonment of its altruistic non-profit roots and its commitment to safety and security above all else.
At this juncture, I advise schools to transition to products that offer more stringent privacy and security measures at the API level. In other words, OpenAI is too volatile a partner to build a reliable educational strategy around.
I believe that if schools do not make this transition soon, they will be failing their students, setting them up for future sales pitches expertly delivered by social influencer-like GPT.
To echo Marc Watkins, why are we over a year into the implementation and integration of AI, and we are still unwittingly handing over all our data and information to these massive corporations?
4. ChatGPT4o: Disabilities, Accessibility, Opportunities
That said, I am closely listening to disability advocates as they review these new products. I realize that I tend in my own reviews to take my able-privilege for granted, and hope to do better in my future articles.
For these advocates, ChatGPT4o’s multimodality is truly groundbreaking. This model can provide detailed descriptions of visual content for individuals who are visually impaired, offering them a new way to experience and interact with their surroundings. Its read-aloud capabilities can assist those with reading disabilities, making information more accessible. Additionally, voice recognition features can support individuals with mobility impairments, enabling hands-free operation and interaction with technology. Finally, this model has EL teachers particularly excited about the possibility of real-time translation due to its low latency and super quick responsivity.
The significance of these advancements cannot be overstated. ChatGPT4o does have the potential to enhance accessibility and promote inclusivity. And yet, one must weigh the costs and benefits carefully in this and any other situation involving amazing new AI services and opportunities.
I am excited through ChatGPT4o architecture to migrate into safer, more protective climes where our students can take full advantage of its many offerings without necessarily availing themselves to the full force of OpenAI’s unfolding business plan and data collection enterprise.
Keeping this complication in mind, we now turn our attention now to some great new products that are built on top of GPT architecture but offer K-12 students more secure spaces to explore and develop.
5. Better AI-Tools for Classroom Instruction
Before we begin, please note that I am not affiliated with or being paid by any of these companies. I endorse these products because I personally find them valuable and useful.
For teachers seeking alternative, safe access points to AI for writing, Lex is an excellent option for students. It offers mediated use of the GPT-3 or GPT-4 architecture, focused specifically on writing and editing tasks.
Lex
Lex, established in 2022, has garnered significant endorsements over the years. Notably, it has received strong support from technical writing professor and prompt wrangler Lance Cummings. He highly recommends this product for his college courses, praising it as a simple, no-frills AI-infused writing space.
What sets Lex apart is its unique mode of interactivity. Users are no longer confined to the global input-output mode of dominant large language models like ChatGPT-4. Instead, Lex offers a writing space reminiscent of traditional word processors, complete with many expected features.
Users can isolate portions of text for AI processing, allowing for various editing options such as: 1. Continue writing, 2. Generate text from a prompt, 3. Get feedback on your writing, or 4. Insert a random word.
Additionally, users have access to a Lex AI Assistant in a separate window alongside the writing space. While the application may not be ideal for lengthy writing projects, it works perfectly well for shorter essays. Text can be easily copied to any other word processing format, including PDF.
PowerNotes
For teachers and administrators seeking secure environments and backend engagement with student writing, I recommend exploring the AI-infused version of PowerNotes.
PowerNotes predates the arrival of AI as a browser-based Chrome extension research and writer tool that supports the use of full-text articles from databases such as EBSCO and ProQuest. Teachers can use PowerNotes to set up writing assignments, guide students through research steps, assist in organizing research materials, and produce and submit assignments for feedback.
Practically speaking, PowerNotes functions as a closed ecosystem, with final grade reporting easily integrated into various learning management systems.
This aspect of PowerNotes becomes particularly valuable in the context of the 3rd AI era and the AI-infused version of this product. If teachers choose to enable AI interactivity at any point in the writing and research process, they can monitor that use on the backend.
Screenshot of a Sample Class: Not Real Students!!!
By assigning and submitting all writing assignments through PowerNotes, schools have a secure option for managing the increased reliance on AI for homework assignments.
While I appreciate the enhanced security and monitoring capabilities that PowerNotes offers, I am also reflecting on the implications of this greater safety. It's interesting that increased protection comes with increased accountability, although monitoring within PowerNotes is a choice made by the educator and school. I am carefully weighing these pros and cons, considering both the potential benefits and the ethical considerations involved. More on these dynamics in later posts.
6. Conclusion
In light of these observations, it's clear that we must pause and rethink the use of OpenAI products in our classrooms, except for rare cases where accessibility needs demand it. The rapid consumerization of AI, epitomized by GPT4o's transformation into an AI salesperson, calls for caution.
We need more secure, responsible access points for our students, shielding them from the increasing commercialization of AI technologies. By embracing alternatives like Lex and PowerNotes, which offer safer and more controlled environments, we can leverage AI to enhance learning while safeguarding student privacy and maintaining educational integrity. It's time to excise our students from the escalating consumerization of AI and prioritize their security and ethical engagement with these powerful tools.
Nick Potkalitsky, Ph.D.
Check out some of my favorite Substacks:
Terry Underwood’s Learning to Read, Reading to Learn: The most penetrating investigation of the intersections between compositional theory, literacy studies, and AI on the internet!!!
Suzi’s When Life Gives You AI: An cutting-edge exploration of the intersection among computer science, neuroscience, and philosophy
Alejandro Piad Morffis’s Mostly Harmless Ideas: Unmatched investigations into coding, machine learning, computational theory, and practical AI applications
Michael Woudenberg’s Polymathic Being: Polymathic wisdom brought to you every Sunday morning with your first cup of coffee
Nat’s The AI Observer: A fascinating investigation into the emergence of higher-order reasoning in advanced AI systems, complemented by amazing coding experiments
Michael Spencer’s AI Supremacy: The most comprehensive and current analysis of AI news and trends, featuring numerous intriguing guest posts
Daniel Bashir’s The Gradient Podcast: The top interviews with leading AI experts, researchers, developers, and linguists.
Daniel Nest’s Why Try AI?: The most amazing updates on AI tools and techniques
Riccardo Vocca’s The Intelligent Friend: An intriguing examination of the diverse ways AI is transforming our lives and the world around us.
nice post!
OpenAI seems willfully indifferent at negative second-order consequences of their shiny products these days.
If they are not interested in AI alignment research or AI copyright infringement, I doubt they care about prioritizing education.
The opening metaphor is great, thanks for sharing many useful cases.