Fostering deep learning in humans and amplifying our intelligence in an AI World
A free 288-page report on advancements in AI and related technology, their effects on education, and our practical support for AI-amplified human deep learning
Six weeks ago, Dr. Sabba Quidwai and I accidentally stumbled upon an idea to compare the deep learning revolution in computer science to the mostly lacking deep learning efforts in education (Mehta & Fine). I started writing, and as these things often go with me, I thought there were many other things that would be useful to think through and for educators to know, and we ended up with this 288-page report.
Since we believe that it is essential that educators at least have an understanding of the issues discussed, it is available for free at SSRN. If people find it useful based on conversations and download totals, we will do our best to keep it updated.
It’s not really possible to write a paragraph summarizing everything, so I made a TL;DR. It is also very well-indexed, so you can read the parts that interest you the most. The full table of contents is below.
We certainly welcome any feedback and do our best to update this biweekly with any corrections and relevant changes. The report is free, but if you find it useful and are able to make a $20 USD contribution to our continued efforts, we’d appreciate your support.
TL;DR
*Hundreds of billions of dollars are currently being spent developing all human intelligence capabilities in machines using AI deep learning approaches. This will eventually enable machines to surpass human intelligence in at least most of the domains in which humans are intelligent. We need to devote similar efforts to developing all intelligence capabilities in humans using deep learning instructional methods and then amplify those intelligence capabilities with AI tools.
*The current level of AI will transform work and life, even if there are no advances beyond the technology’s current level. Machines are already smarter than us in some ways, and we will never be able to compete against them in those particular ways.
*Expected short-term gains in AI (3-5 years), including the development of AI agents that can act autonomously, will dramatically change work and life.
*The world that today’s 9th graders and college freshmen will graduate into will be fundamentally different than the one we are currently living in, and we need to help them prepare for that world.
*Many individuals have a distorted view of the existing capacity of these systems relative to what they can already do and their emerging capabilities because they are only using a free version of the most popular AI model, ChatGPT3.5, which lacks many of the capabilities of more advanced systems that are commonly accessible. There are many more capable systems and hundreds of thousands of AI apps that build unique capabilities, some of which are not in any way good, on top of those models and even smaller models. And there are undoubtedly capabilities of which the general public is not yet aware.
*Developments in related technologies such as AI pins (due out next week) that can record a person’s entire day and then allow them to search and interact with it, synthetic biology, quantum computing, and brain-computer interfaces will even more radically change the world.
*AI literacy for students and educators is a borderline emergency. I say “borderline” to soften the blow. It’s actually hard to think of something more important for students to learn at the moment other than basic literacy so that they can understand how society is changing and the importance of avoiding using these technologies to do harm, which, as we’ve seen in multiple instances lately (Spain, New Jersey), is being done by teenagers at the click of a button.
*All or nearly all future ‘knowledge work’ will involve using these technologies, as studies show productivity gains of 30%–80% with even the existing, limited technologies and fly-by-night integration. I’ve seen people argue the productivity gain is as high as 400%.
*The fundamental grammar of the education system has barely changed in 100 years, so making the changes schools need to be relevant in this world will be very difficult, but we know that educators will rise to the challenge.
*Public schools are facing many pressures: decentralization (online schools, private schools, homeschooling (it’s grown so much that VCs are trying to get in the space), DAOs, students simply dropping out; politicization; violence; social media & mental health; teacher shortages; industry complaints about the lack of relevant skills in K-12 and university graduates; day-to-day challenges) that will make it harder for them to adapt.
*This will be magnified by technologies that allow students to learn almost anything they want 24/7 in computationally mediated, multimodal, personalized environments that we have already lost our students to, despite thinking that they crave human attention from adults. One high-tuition private school is already largely relying on bots to provide content teaching, turning teachers into general facilitators and focusing their teaching on developing the skills we recommend.
The education system focuses on and excels at what computers are already good at—learning information and analyzing it. Many of those trying to cash in on the AI revolution are providing solutions to schools that simply focus on helping them do that better, reifying the problem of a lack of AI-relevant adaptations as they help teachers better teach students to try to compete against machines in ways that are impossible while sucking money out of the public education system.
*A significant number of current academic assessments can be completed wholly by bots and there is no reliable way to detect AI vs. human output. As more and more tools allow students to produce writing in undetectable patterns and even their own voice, this is a race against machines that cannot be won. These technologies are everywhere, and schools will not be able to control their use, just as governments will struggle to control them with regulation given the rapid and non-linear developments coupled with the decentralized nature of the internet.
*Developing “AI Policies” is important to setting expectations, deterring behavior that threatens students and faculty, and managing liability, but these are not panaceas, as they are difficult to enforce for the reasons mentioned. AI policies should be developed, but they aren’t the solution everyone is hoping for.
*Instruction and assessment for an AI World, where students and professionals will live and work with the 24/7 assistance of bots that are as smart or smarter than them in many ways, needs to be reconfigured; it’s impossible to imagine how this cannot be the case.
*Students and teachers need to start working with AI so they can understand how it works, but AI is not the sole answer to the challenges we face. Without making changes to the educational system, AIs being sold to schools will just supercharge students’ abilities to do what computers will always be better at doing. We can’t say this enough times.
*Schools already have many academic deep learning approaches in place that focus on critical thinking, collaboration, and communication, as well as developing metacognition and the ability to learn on one’s own, which will enable students to thrive in this constantly changing world. These abilities will determine students’ futures. We lay out suggestions for how to do that in the paper.
*Human deep learning approaches need to be integrated with students learning how to use AIs to amplify their intelligence. Humans cannot compete with machines, but they can use machines to extend their own abilities. Humans have always used technologies for this; it’s just that today’s approaches require more radical applications.
*We need to focus on developing the “non-intelligence” human attributes, including courage and love.
*Leadership is needed now more than ever to begin instituting these changes. When significant change happened in the early 20th century, it was due to leadership, and it can happen again.
*Practical action, combined with visionary change, is needed, and suggestions are included.
*This report has a focus on K–12, but a lot of what is written applies to universities, and we included an appendix on universities. We hope that as universities work to adapt to the AI world, they will support K-12.
*Even if you do not agree with us, our 1100 end notes may prove useful to you.
*We do hope that more schools find a way to pay more attention to these issues.
Ironically, the paper that provided the technological foundation for many of these advances was simply titled Attention is All You Need. People in AI were paying attention, and now all the authors are part of multi-million and multi-billion dollar enterprises. They, and others, took advantage of it and other advances in AI to change all of our worlds, and there is no going back.`These advances will forever change how we and our students need to learn and work.
We realize that schools are facing a lot of challenges, but we hope this large, free report provides a foundation for understanding these issues and a roadmap to consider for the “future” that is already here. We need to act now to maximize the benefits and minimize the harm.