By Patrick Gallant
Given that the economy is changing at an incredible pace, many teachers are naturally concerned about their jobs. Will AI replace them? Or will they still have a role to perform?
Fortunately for existing and prospective teachers, education is one of the most stable sectors from a job security standpoint. Generative AI is impressive, but its inherent flaws make it impossible to replace human teachers in the near future of teaching.
That said, educators stand to benefit from AI. AI language models like GPT-3 can help with a variety of tasks, such as lesson planning and role playing.
Here's what we'll cover:
- Why Employment in Teaching will Continue to Rise
- How Generative AI Can Help Teachers
- Problems with AI
- Why You Should Become a Teacher
Will Employment in Teaching Continue to Rise?
According to a recent McKinsey report, education is one sector which will continue to see job growth in the years to come. This is especially true of the primary and secondary education sectors.
One reason for this has to do with demographics:
- The Baby Boomers, the largest generation in US history, is rapidly aging into retirement.
- Their senior positions are being filled by members of Generation X, one of the smallest generations in US history.
- At the same time, Millennials- the second largest generation in US history- are in their peak childrearing years.
These factors suggest that there is an increasing supply of students and a decreasing supply of teachers to teach them.
It is true that most people will have to switch careers multiple times. Somewhat paradoxically, this fact actually makes education more stable. This is because having a broad education, as well as regular training in new fields, is more important than ever.
Another reason why education is a stable career choice is that soft skills are becoming more valuable. In a world increasingly dominated by computers, communication skills and empathy can still only be taught by fellow human beings.
Finally, AI itself opens up more opportunities for education. Although AI tools can handle some tasks, people still need to be trained to push their buttons.
How Can Generative AI Help Teachers?
Generative AI tools like ChatGPT and DALL-E are impressive for what they do. But despite the recent hype, they are far from capable from doing everything- it would be unwise to use ChatGPT to pick your stock portfolio, for example.
AI tools are most useful when applied to very specific purposes. When it comes to education, here are a few specific uses:
- Small administrative tasks
- Imagination work
- Teaching fact-checking skills
- Personalization
Small administrative tasks
Teaching involves so much more than just what happens in the classroom. Teachers have to write lesson plans, quizzes, student reports, and more.
This is where generative AI can help take care of some of the work. Already, tools like ChatGPT can write formulaic text documents like lesson plans and quizzes. In any case, teachers should take care to edit and proofread anything they make using AI.
Here are some prompt ideas for this purpose. In any case, it's a good idea to experiment and find what works with you:
- Write a lesson plan that introduces the US Bill of Rights.
- Write a few question prompts for a test on the quadratic formula.
- Design a rubric for grading student essays.
Imagination work
Generative AI often works at its best when it performs imaginative functions like role-playing and brainstorming. This is because, in these tasks, it doesn't need to be completely factual- it just needs to sound plausible.
This is especially useful for language learning, where students can practice their skills by conversing with AI.
Here are some prompt ideas for imagination tasks:
- You are a scientist on a mission to Mars. Describe the challenges and discoveries of space exploration.
- You are a nutritionist. Advise me on healthy eating habits and create a personalized meal plan.
- Brainstorm some ideas for games to play in the classroom.
- Brainstorm some possible prompts for a class discussion on politics.
Teaching fact-checking skills
Fact-checking is a more important skill now than ever. And the tendency of AI programs to make stuff up only makes fact-checking even more important.
For example, ChatGPT once reported that a man crossed the English channel on foot in 2020.
Once students have become convinced of this problem, you can have them use ChatGPT as an exercise for fact checking. Make the program report on a topic, then have students do their own research to verify every single one of its claims.
Personalization
Adapting lessons to fit each student's particular needs is ideal, but time-consuming.
Fortunately, some AI education tools help design personalized learning paths, based on dynamic feedback. DuoLingo and Rosetta Stone, for example, continuously redesign your learning pathway depending on which questions you get right or wrong.
Problems with AI
Despite how impressive they are, AI systems have serious problems. We will still need human decision makers for most important tasks, for now and in the foreseeable future.
There are a few big reasons why AI language models like the ones used by ChatGPT can't fully replace humans:
- Making stuff up
- Limited knowledge
- Bias
Making stuff up
This is commonly called "hallucinating", although this term is somewhat misleading. It's not clear that AI language models are perceiving anything at all when they function, whether factual or not. They are simply guessing the next word in the sentence.
The problem is that language models are not really connected to reality the way we are. They are trained on a ton of text information that was assembled at a specific point in time, and are not able to really look at anything that is happening right now.
Limited knowledge
Despite how much information AI language models are trained on, they are typically lacking in a lot of important information. Since GPT-3 was trained in 2019, ChatGPT originally lacked any information about Covid-19, for example.
This is related to the previous problem of making stuff up. AI language models don't really represent reality- they represent language.
That's why they don't really know what they do know and what they don't know. So they will often make stuff up if you ask a question that involves information they have no access to.
Bias
Since AI language models are based entirely on the information they are trained on, they will function with any biases inherent in the training data. This is a problem, since any large set of data will include some biases, and it can be hard to root them out.
For example, asking language models different political questions will yield different results, depending on the model. ChatGPT has been shown to provide more left-libertarian leaning answers, while Meta's LLaMA provides more right-wing authoritarian answers.
AI image generation programs are perhaps even more strikingly biased. Programs like Midjourney are known to generate images based on ethnic and gender stereotypes.
Why you should become a teacher
Don't let the rise of AI turn you off from becoming a teacher. Education as a profession is probably one of the most stable and rewarding careers you can pursue.
Think of how computers affected education, when calculators started to become widely available decades ago. We no longer teach as much arithmetic. Instead, we teach programming.
Most likely, something similar will happen here. Students may need some skills less. But they will also need AI-related skills more.
And teachers need training, too. If you are thinking about becoming a teacher, take a look at our one-year certified teacher accreditation program.