Open AI is not Higher Ed's friend
OpenAI is Not Your Friend: A Warning for Higher Education
As I write this, I already feel a mix of anxiety and solarpunkism thinking about higher education's threat and opportunity in the age of AI.
There seem to be two roads I anticipate these institutions to take.
They are not only false choices, but disastrous.
The first one is the "quick fix" approach.
The Allure of Quick Fixes
Higher education institutions may be tempted to turn to tools like OpenAI.
After all, this is the direction many enterprises are already turning.
While not optimal, for many of these organizations, it may still make sense to work with the OpenAI models for their own internal purposes.
For example, companies are improving operations, customer support, even simpler coding tasks, by leveraging external learnings for internal use.
The hosted services like OpenAI could make sense because the flow of learning is primarily inward.
These AI solutions, like the cloud services of a decade ago, offer a seemingly "easy" path to modernization.
But building a strategy around these hosted solutions is a mistake for those that have high value content that could be of interest externally.
The Race for Data Dominance
We are in a race for data dominance.
The institution that consumes and controls the most data will win.
In the past, this concept of data dominance came from the belief that more data would permit better decision making through machine learning: better marketing, pricing, personalization.
I believe the jury is still largely out whether this thesis is true.
However, unstructured, semantic data is the lifeblood of AI that needs to generate responses to human prompts. And OpenAI's ChatGPT has illustrated two powerful tropes.
The first is the human interface: chatbots, which emerged as a popular tech a few years ago (I, myself, dove into building NLP-based chatbots and hit many limitations).
But for some interactions, chat is a preferable interface versus, say, grids on a spreadsheet or forms in a web browser.
However, part of that experiment revealed that it's also not always the case.
Sometimes, it actually is better to book a flight through a filterable table.
Often, it's better to mainpulate data in a grid of a map.
But...Chat GPT also showed that IF the ability to provide useful responses that are "human friendly", chat can be more powerful.
In the case of being able to retrieve responses that can be understandable and useful, the more data a model has, the more powerful it becomes.
When higher education institutions adopt well-known solutions like OpenAI, they will give away their most valuable asset: their data.
This not only risks the quality of their outputs but also distances researchers from their own work.
The Cost of Disintermediation
Adopting these solutions leads to disintermediation. Researchers and educators lose direct access to the data and tools they need, relying instead on third-party providers. This can be detrimental, leading to a loss of institutional knowledge and control over educational content.
A study by the Brookings Institution highlights how data ownership and control are critical for organizations looking to leverage AI effectively:
"When institutions lose control over their data, they also lose the ability to innovate and adapt quickly" -- Brooking Institution
Licensing data to external AI providers means surrendering control over valuable research outputs. Universities might license their academic papers to OpenAI, thinking it will make their research more accessible. However, this could result in their own content being repackaged and sold back to them, diminishing their competitive advantage.
Dangerous DIY
Embracing Open-Source Solutions
The alternative, though challenging, is for higher education to embrace open-source solutions and build their own AI systems. This approach, while daunting, can be more cost-effective and flexible. Open-source AI platforms like TensorFlow and PyTorch offer robust tools for developing custom AI solutions tailored to specific institutional needs oai_citation:2,openai-not-your-friend-higher-ed.mp3.txt.
The Lure of Licensing
Imagine a scenario where all the data is licensed to OpenAI.
The institution loses control, unable to leverage its own research effectively. This echoes IBM’s mistake with Microsoft, where licensing allowed Microsoft to capture immense value, leaving IBM behind.
Higher education must avoid a similar fate by retaining control over their AI development.
Reintegration: A Strategic Imperative
The solution is to reintegrate AI development within higher education, creating competitive products internally. This requires a shift in mindset, from viewing AI as an outsourced service to a core competency. Universities must build, maintain, and improve their AI systems in-house.
Understanding the Customer
A major hurdle for universities is the gap between understanding their customers and developing effective product strategies. Traditional research-focused mindsets clash with market-driven product strategies. Universities need to adopt a more customer-centric approach. A report by EDUCAUSE emphasizes the importance of aligning IT strategies with institutional goals:
"Institutions must align their IT strategies with broader institutional goals to effectively serve their educational communities" oai_citation:3,openai-not-your-friend-higher-ed.mp3.txt.
Overcoming the Ivory Tower Mentality
The “ivory tower” mentality, where institutions remain isolated from real-world changes, can hinder innovation. Universities must stay in tune with shifts in technology and market demands. Platforms like YouTube and online course providers have disrupted traditional education models by offering flexible, on-demand learning opportunities. Higher education must adapt to these changes to remain relevant.
Organizational Dynamics and Cultural Challenges
Organizational dynamics and cultural inertia can hinder innovation. Those within universities who understand the need for strategic, customer-focused approaches often face resistance from colleagues who are accustomed to traditional methods. The urgency and ability to pivot quickly are crucial, yet these traits are often lacking in higher education.
Building a Culture of Innovation
Creating a culture of innovation within higher education requires strong leadership and a commitment to change. Universities need to invest in training and development to build internal capabilities for AI and digital transformation. Collaboration across departments and with industry partners can foster a more innovative environment.
The Critical Juncture
We are at a critical juncture. Higher education must act swiftly to retain control over their data and research. The future of higher education depends on staying ahead of the curve and avoiding the trap of outsourcing critical assets to external AI providers.
Moving Forward
In upcoming discussions, we'll explore implementation details, risks, and costs associated with building in-house AI capabilities. For now, remember: the future of higher education depends on strategic control of data and a proactive approach to AI development.
By addressing these challenges head-on, higher education institutions can ensure they remain at the forefront of innovation, providing valuable education and research in an increasingly digital world.