My Perspective on AI in Education: Changes Needed for Students to Thrive in a Radically Changing World

Dorothy, we aren't in Kansas anymore.

I have given AI in education a lot of thought. I --

• Co-hosted what was probably the first large webinar (200 people) on gen AI in education in January of 2023.

• I co-edited a 32-author book on the topic in the spring of 2023.

• I just finished co-authoring another 700+ page book (print edition now available).

  

• I co-authored a “guidance on guidance” document schools can use in developing AI guidance.

• I had a significant role in the development of SAUSD’s guidance.

• I’ve written 107 blog entries about AI in education.

• I’ve done presentations for K-16+ educators.

• I’ve appeared on many AI in education podcasts.

• I taught a college “Intro to AI” course.

• Most importantly, I’ve posted a lot about AI on social media :)
  

  Life-long learning works.

Anyhow, my point is that this reflection comes after a lot of thinking.

So, what do I think?

I think the most important thing we need do to is to prepare students to live in and work in a world of advanced AI, integrated robotics, and, eventually, synthetic life. This is a world where machines will likely at least achieve a level of intelligence in the ballpark of human intelligence and exceed our intelligence in certain domains (they already have). We are on the cusp of a new Cambrian explosion that will forever redefine how we live and work, and all of our students will live most of their adult lives in this world.

Preparing for, and thriving in, that world necessitates being a well-educated individual who is committed to being a lifelong learner.

Education is more important than ever.

Preparing students for this world should not be a controversial idea. It’s no different from preparing students in the 19^th century, at the end of the agricultural era, to thrive in the 20th century industrial era. We stopped teaching young people how to use oxen and horses to plow fields and taught them how to use steam-powered tractors and threshing machines. We stopped teaching young people to spin and taught them how to use a spinning jenny. Sure, using the spinning jenny meant they didn’t experience the full “struggle,” but humans are tragic creatures who will always find ways to struggle. We’ll be ok.

I contrast this educational North Star with another way to look at “AI in education:” focusing on infusing AI into teaching and student learning in the classroom.  There is a substantial debate about this that I won’t repeat here, but I think it begs the question: We should first think through what and how we want to teach to prepare students for an AI world before we rush into using AI as much as possible in the classroom. We certainly don’t want to use AI to help students learn how to use oxen and horses to plow fields.

This isn’t to say that it’s bad to use AI to do some of what we do now, and it is doing many incredible things. Using AI helps build AI literacy. AI can save underpaid and underappreciated teachers a lot of time. Individualized instruction can accelerate learning. It can provide lacking support to the neurodiverse. But it’s not the North Star; it’s the Kochab Star.  Once we figure out how to prepare students for the AI world, the second-brightest Kochab Star can help everyone prepare.

So, how can we prepare students for this AI world?

Help them develop foundational knowledge. There is a lot of talk today about knowledge being at everyone’s fingertips and just needing to know how to ask the right questions, but a person can’t ask the right questions without foundational knowledge. And we study history because culture binds us; if you don’t teach your students history, the Russians will.  Everyone needs to learn English math, science, and social studies.

Help them understand the changing world around them. AI is already impacting our perceptions of each other, interpersonal relationships, human-AI relationships, the economy, the military, the environment, and politics. With the acceleration of capabilities currently in the range of 10X, we can expect AI to have 10X the capabilities it currently has a year from now. 100X today’s capabilities the following year.

And even with no advances, area experts can make courses in an hour using AI.

AIs can help you celebrate your birthday.

Some call this AI impact: What impacts will AI have on society and how can we start helping students think about them?  Start talking about it with students; they are very concerned about their future. It will impact their future more than anything else they are learning.

They will see the drones flying overhead. For many of our lower SES students, this will impact them the most, as lower SES neighbourhoods correlate with higher crime neighbourhoods, and these are the first neighbourhoods where drones will be deployed. E-Hall Pass and other surveillance technologies monitor them at school and now drones monitor them in their neighbourhoods. Their phones monitor their every step as they walk home. [But we can’t let them learn with AI at school because we are protecting their privacy.]

Help them develop durable skills.  AI can and will be able to do a lot of things, including some of our jobs, for us but it can’t develop our communication, critical thinking, and collaboration skills. Just as it cannot exercise for us, if we want to develop these skills, we have to develop them ourselves. These are the number one job skills now and into the future.

Prompting thrives on context, so we need to learn how to communicate. And our students, especially our low SES students (and that number will grow as unemployment increases and wages decline) are going to need to learn to advocate for themselves, to claim a sliver of the AI economic pie. It’s hard to find any AI scientist or economist who does not think that at a minimum the rapid acceleration of AI capabilities will not produce short-term unemployment and wage suppression.

With all the world’s knowledge at our fingertips and the ability to generate knowledge synthetically, we need to be critical thinkers.

In a world where AI will be able to do a lot of things for us, we are going to need to work well with others, as we become our own bosses and business owners.

Just last week, AI Godfather Geoffrey Hinton revealed he has been pressing the U.K. government to establish universal basic income. Fellow Godfather Yann LeCun, who is more optimistic about employment over the long term, agrees there will be significant short-term labour market disruptions and that government support will be needed. What does the Godmother, Fei Fei Li, think? She seems to choose her words more carefully but points out that AI can do what a lot of “knowledge workers” do. Sam Altman? He thinks there will be significant short-term unemployment. Elon Musk thinks people will work if they want to, and that AI will be able to do everything. What does the third Godfather, Yoshua Bengio, think? He thinks AI will lead to significant unemployment and social disruption.

Do you think AI can’t do your job? Texas replaced 4,000 STARR test essay graders with AI. Klarna replaced 700 employees with AI. Work for translators and graphic designers is declining rapidly.

In 2 years, AI, generally speaking, will have 100X the capability it has now. MIT Research Scientist Lex Fridman noted today: “(The) next 10 years in AI progress will be wild and intense.”

I say, “generally speaking” because this assumes no efforts are made to specifically train it to do all or nearly all the tasks you do at work. People won’t ask how smart the AI is when it takes their job. Arguments about how AI can’t be as smart as a human won’t help them keep their job.

We are training machines to do what all our schoolteachers told us we needed to do to be successful: Develop our intelligence.

Help them develop metacognitive skills.  Metacognitive skills, which involve the ability to monitor and regulate one's thought processes, are invaluable. In an ever-changing world, the ability to monitor and regulate one's thought processes is invaluable for adapting to new situations and learning new skills. By critically evaluating their understanding and identifying gaps in their knowledge, individuals can seek out relevant information and develop new strategies for mastering new challenges.

Help them learn to work with AIs.  Yann LeCun frequently states that we will all work with/supervise an army of digital AI assistants. These aren’t only executive assistants but AI accountants, lawyers, engineers, and consultants.

This is something every leading AI scientist says, and it is why companies are in a race to develop assistants and AI agents that we are starting to see deployed. Both OpenAI and Google put on demonstrations of these emerging agents last week.

There are already bets about how long it will be before one person can run a $1 billion company.

Today’s students need to learn how to be their own bosses, and entrepreneurs who manage teams of highly intelligent AI agents to achieve objectives. This requires them to be well-educated.

Who is going to hire them to do what a team of AIs can do better than them at pennies on the dollar?

Foundational knowledge, communication skills, metacognitive skills, and an ability to work with AI will serve our students well.  They need foundational knowledge to ask Ais the right questions and understand the answers. They need critical thinking and communication skills to interact well with AIs, especially as AIs replicate more human communication patterns.  They need to be aware of their strengths and weaknesses to adapt and develop. They need to learn to work with AIs to both augment their intelligence and compete against others who are using AIs to do so.

Alter teaching and learning to make all this possible. Last spring, in the first talk I ever did on AI, I told the people they need to hear from are not just people who understand the trajectory of AI, but people who can help them alter teaching and learning to move it more into project-based learning, interdisciplinary learning, performative assessment, entrepreneurship education, etc. Students need to learn and demonstrate what they know by doing. Jerry Crisci calls this ‘instructional redesign.” Leon Furze makes a strong case for moving to performative assessments, such as oral presentations and debates, arguing most current assessments are no longer reliable. This idea is emerging in AI guidance documents (NC, SAUSD). Sabba and I wrote a whole book about it. Last week I attended a webinar by individuals at the state and district levels who were working on AI in education, and they all made the same point – teaching and learning has to change along these lines.  The idea that we can primarily grade students on their ability to complete any product will fade, as AI, and emerging teams of AI agents, can/will be able to complete any product we ordinarily ask students to complete.

As Adaptive Intelligence summarizes:

Some people say some schools will never change this radically. That’s true. But schools that don’t will lose their relevance and shouldn’t be surprised when students stop attending them.

Integrate some stuff. Everyone is starting to recognize that it’s hard to add more and more to the curriculum. We keep doing this because there is more to learn and no one wants to cut anything, so let’s start by integrating some instruction about AI. For example, in history, when we teach about unemployment triggered by the Industrial Revolution and the Luddite backlash against it, we can teach a lesson about similar issues in AI.

Cut some stuff. Remember that the curriculum you are teaching now was not designed to prepare students for the AI World. This doesn’t mean it is no longer relevant, but it does mean that you can take a chance and substitute a lesson about the AI World.

Is there really nothing you can skip? Keep in mind that all students don’t get 100s and that the curriculum was designed to prepare students for a different world.

Look beyond “intelligence.” As Ilya Sutskever, one of the world’s leading AI scientists who studied under Geoff Hinton, notes:

In education, we’ve essentially been selling students a future where developing their intelligence determines their success. But while intelligence will still matter, it’s probably no longer going to be as important. AIs will be smarter than the smartest among us. Those “less intelligent” will close the “intelligence gap” with the “more intelligent” using AIs. It’s already happening.

We are going to need to develop our other human qualities. There isn’t any reason students can’t do that in school.

Invest in preparing students, invest in people. In our book, we observe that hundreds of billions of dollars are being invested to develop machine intelligence while investments in human intelligence continue to decline. If we want to develop human intelligence in a way that makes it possible for us to work with machines and each other as the world changes, we are going to need to invest in doing just that.

But what if this is all a sham?

Some think AI is like crypto and that is all a scam.

It is not.

AI is already being used across society and it’s starting to alter how we live and work.

The world’s most valuable companies, with a combined market cap of $7T+, have bet their futures on it.

Lonely (mostly) men are dating it.

It has helped students avoid suicide.

Its flown military jets without a pilot at the controls. Its killed people in the Ukraine.

Its convinced you to buy products. Its stopped your car from backing into another car.

It allowed the lowest skilled workers to close the gap with the highest skilled workers.

Its been grading written student work all over the world.

Its taught courses in Hong Kong.

It has been doing student work since at least December ‘22.

Some people think (this is controversial) tools like Sora and models like Chamelon indicate it is developing an understanding of the physical world.

It can do a lot of things and we need to learn to work with it and relate to it.

I haven’t even brought up brain-computer-interfaces, quantum computing, or embodied robotics.

Conclusion

The most important role of education in the age of AI is to prepare students to live and work in a world that will be fundamentally reshaped by artificial intelligence.

This requires helping students develop foundational knowledge, understand the impacts AI will have on society, cultivate durable skills like communication and critical thinking, learn to work effectively with AI systems, and adapt to the changing nature of “work.”

Educational institutions must alter teaching and learning to focus more on project-based learning, interdisciplinary studies, performative assessments, and entrepreneurship to equip students with the competencies needed to thrive alongside increasingly capable AI.

With machines poised to match or exceed human intelligence in many domains, developing students' human qualities, and ability to work with AI will be essential. Substantial investments in education are needed to rise to this challenge and ensure today's students are prepared to shape an AI-driven future. The accelerating capabilities of AI make this an urgent imperative for education systems worldwide.

If they aren’t preparing students for the AI World, what are they preparing them for?