Updated Report: Teaching Alongside (Ro)Bots, Agents, Movies, AGI, Professional Development, More

We cannot teach students and train teachers the same way we taught them for the last two industrial revolutions and expect them to succeed in the next industrial revolution.

I updated our “Humanity Amplified” report and this update covers a few key areas.

The role of the teacher (Chapter 5). I’ve spent the past few months working with others on a basic “AI Literacy” curriculum. It is designed to be co-taught by a bot and a teacher. The first part is essential because the number of individuals who can teach AI literacy without the help of a bot is going to be very small. So, the idea is to have the teacher and students learn together from the bot, with the teacher being a caring instructional guide who knows how to teach better than a bot but doesn’t necessarily know a lot about AI.

When thinking through that, it occurred to me that AI literacy is just the first thing that will have to be taught that way. As AI rapidly advances knowledge at a rate faster than any previous advances in human history, even foundational knowledge students need to learn, it’s not practical to help every single teacher upskill in that area.

When watching a lecture by Dr. Manolis Kellis at MIT the other day (you should check it out, as it’s only 15 minutes and rather mind-blowing), he explained how AI is completely changing the way science research is done, including even the role of the hypothesis.

Science teachers will need to learn this, but there is no way to effectively upskill them all in a short period of time.

So how can they learn these things, by collaborative teaching with bots. We cover some different ways to do this in our report (Chapter 6), but teachers will need training. Shockingly, as noted in a new Education Week article, teachers are receiving almost no AI training at all.

Towards the end of his lecture, Kellis talks about how AI is changing medicine and the growing role it will play in treatment and diagnosis. He says the best doctors will be the most caring doctors and those that have the best bedside manner, not necessarily those that have the most knowledge.

I’ve been saying the same about teachers for months: AIs will soon exceed their knowledge, ability to share content, and ability to adapt instruction to individual students. Vinood Khosla explains:

Having a personal tutor for every child on the planet, that's an AI tutor, can pay way more attention than any physical tutor can. And my wife works on a nonprofit called CK 12 that is building AI tutors and they map gaps in a kid's learning far better than a personal teacher can. And then they can tutor to the gaps in learning.

The best teachers will be the ones who are the most caring and have the most rapport with students. As with doctors, content knowledge will no longer have much of a premium.

Are we just going to stick our heads in the sand and pretend this is not true or are we going to provide professional development for teachers to work in a new way?

Our school districts should not be leaving our teachers out of the next industrial revolution.

“AGI” (Chapter 2) Artificial General Intelligence (AGI) is a concept in the field of artificial intelligence (AI) that refers to a machine's ability to understand, learn, and apply its intelligence to solve any problem that a human being can, with equal or better efficiency. Unlike narrow AI, which is designed to perform specific tasks (like facial recognition or playing a game), AGI encompasses a broader, more adaptable form of intelligence that mirrors human cognitive abilities. There is a growing expectation among researchers, technologists, and futurists that while it may not be possible to define AGI or decide precisely when we’ve achieved it, strong breakthroughs in AI capabilities are on the horizon, potentially within the next 3-8 years, that will put us in the ballpark of different definitions of AGI ((Kurzweil (2029); Altman (2030 or before) (minute 1:32); Musk (2025-2028)); Huang (2030) based on the capabilities of the systems.

And, as I always say, the exact definitions and related years do not matter: We are looking at a massive expansion of capabilities.

Agents (Chapter 2) As I’ve been discussing, AI agents are rapidly advancing. The key characteristics of AI agents include autonomy, the ability to make decisions and perform actions without human intervention; adaptability, the ability to learn from and adjust to new data or environments; and goal-oriented behaviour, the capability to take actions to achieve specific objectives.

Agents could, for example, complete a student’s science project (thanks to ChatGPT4 for detailing my example).

They can also write essays through a process rather than from a zero-shot prompt.

Outside of student science projects and essay writing, these agents will propel the development of new AI models and related advances.

It also changes the comparisons we make between models. For example, ChatGPT3.5 using an agent workflow, substantially out-performs ChatGPT4. Since agents can reflect, use tools, plan, and collaborate, they’ll be able to do more and more of what we do.

And they will persuade us.

Yes, it’s exponential. Will “GPT5” create “GPT6”?

Robots (Chapter 2). Bots and robots differ significantly in their operational environments and functionalities. A bot is essentially a software application designed to automate tasks, existing entirely within a digital framework. Bots interact with users or other digital systems through software interfaces, lacking any physical form. In contrast, a robot is a physical machine capable of performing tasks autonomously or with minimal human intervention, often equipped with sensors and actuators to interact with the physical world. Robots are programmed to execute a variety of tasks, ranging from simple repetitive activities to complex operations requiring real-time environmental adaptation. While bots automate digital processes, robots extend the capability of automation to the physical realm, making tangible impacts on the environment around them.

There have been stunning advances in robotics, as the bots, which are now capable of learning on their own have become the brains of the robots. FS-1 is capable of speech-to-speech reasoning. Nvidia offer similar robots.

See: NVIDIA Robotics: A Journey From AVs to Humanoids

The army is considering integrating robot platoons into brigades.

Driverless cars, which are really robots on wheels (Musk) are now operating in a 45-mile radius in LA.

And their structure as “humanoids” means they can learn more as embodied individuals.

Video. Sora capabilities were demonstrated about a month ago, but they continue to improve.

Immersive environments and wearables (Chapter 2). Generative AI has rapidly expanded the production of immersive environments and has accelerated the development of wearables to make those environments more accessible.

Tutoring “(ro)bots” (Chapter 3. It doesn’t take much of an imagination to see where this is going, both across society and in education.

We are seeing the emergence of bots that are more and more intelligent (both generally and as a result of training on domain-specific data, fine-tuning, HRLHF, HRLAIF, RAG, and other methods) and able to share such knowledge with students on screen, in immersive environments, and soon, even physically.

One company is even working on cloning professors, something Dr. Kellis did discuss in a previous podcast.

These (ro)bots will be able to teach our students, whether we like it or not. Will they be able to operate at the same intellectual level as a human? No, they don’t have all of our capacity. But will they be able to do a lot? Yes. Maybe they’ll be able to do 85% at 5% of the price…And, eventually, maybe they’ll be able to do 110% at 1% of the price.

What does this mean for teachers? It means they need training in AI so they can work as facilitators between AIs and students who still need human support and will need it (I think) even in a world of AGI.

Unemployment (Chapter 3). Yes, AI is coming for our jobs. It can and will be able to do an overwhelming number of our work tasks. If it can even do some of your work tasks, then your remaining tasks have more limited economic value in most instances. Some of the remaining ones may not be worth your employer keeping you around to do.

Of course, there will always still be a role for humans, at least for a while. As this article notes:

In short, there is still a role for humans--such as company founders--when it comes to making executive decisions. ChatGPT can provide creative solutions, but it cannot do the math or the customer interviews needed to make a compelling case for picking the best option, or winning the capital and people required to implement an effective strategy.

It is getting much better at math, but, yes, we still need people for “executive decisions.” And if we are focused on helping our students and teachers learn to make those, then we are contributing to their possible success in an AI World.

Learning loss (Chapter 3). A study that shows a link between the use of ChatGPT and declines in academic performance is getting a lot of press. The "methodology" is horrific, but at the same time it makes an obvious point: If you wait until the last minute and have an AI do your work, you won't remember (really "learn") anything. So, while the study doesn't prove anything given its atrocious methodology, it makes an obvious point: If don’t learn, you won’t learn.

What to do about learning loss (Chapter 4)? As I've been saying since last May, the current issue in education isn't so much how to use AI (that's important but the easier part), but how to redesign instruction so that when students work with AI, they will learn MORE and what they learn will be RELEVANT to in a world of AI. This may include some more low-stakes, in-class formative assessments (Dr. Ethan Mollick), but it also includes “instructional redesign” (Jerry Crisci), with elements like debate (Me + others), work portfolios (Dr. Sabba Quidwai), hands-on STEM, entrepreneurship, games, and other programs leading the way.

Yes, students need to be learning about things, and they need to be learning about them in some traditional ways, but they really need to be learning how to do things, and how to do those things with AI. This is the new world, the new “industrial revolution.” This also applies to teachers. Without students and teachers learning how to do things with AI, the relevance of our educational institutions will decline by the day.

Bots will know more than any human teacher or doctor could possibly know and they’ll always be able to teach the newest and most relevant knowledge.

We cannot teach students the same way we taught them for the last two industrial revolutions and expect them to succeed in the next industrial revolution.

How are you redesigning learning to prepare students for an AI World?

Want to learn more?

Our Guidance overview for districts has been moved to a new document.

This free (version 1.0) 63-page AI Guidance Strategy Guide for Schools includes:

* A summative review of 10 major AI Education guidance documents issued globally.

* Coverage of major developments such as AI wearables that are not covered in any other guidance document.

* Identification of the combined strengths and weaknesses of the documents

* Many concrete suggestions for step-by-step actions you can take to develop an AI Guidance and begin implementation in your school.

Check out our free “Humanity Amplified” report at SSRN (chapter references above are from the report).

Also, check out our 1,000 page book —

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4621210