Authored by

Geetanjali Shrivastava
Co-founder, Chief Learning Officer

Impact of Large

Language Models

on Education

© Adaptiv.Me January 2023

Contents

1. Introduction 03

2. Teachers : the new super humans in a brave new world 04

3. What are large language models? 05

3.1 How do large language models work? 06

3.2 Examples of large language models 08

4. Impact of large language models on education 09

4.1 Automating creation of learning material 10

4.2 Summarizing complex concepts / textbooks. 11

4.3. Teaching programming and other subjects with repetitive

13
tasks

4.4. Generating assignments and grading frameworks 13

4.5. Generating personalized lesson plans 15

4.6. Providing additional support and mentoring outside
17
classroom

6. Future of Generative AI & Education 18

6.1 Adaptiv and the future of education 19

7. Conclusion 20

8. Disclaimer 21
Introduction
What you can expect to learn in this paper

As artificial intelligence (AI) and machine learning

continue to advance, content creators and even
educators have started exploring large language models,
in particular, and their potential to revolutionize the way
we learn and teach.

Large language models allow educators to offer

personalized and adaptive learning experiences and
automate content creation and assessment. There are
many potential benefits to incorporating language
models in education. By carefully considering the use of
this technology, educators can effectively leverage the
unique capabilities of language models to enhance the
learning experience for their students.

In this paper, we will explore the impact of large

language models on education, and how educators can
leverage this technology to become the superhumans in
this brave new world.

Impact of Large Language Models on Education

 Page 4

Teachers : super

humans in a brave

new world
Teachers have always been the heroes of the classroom,
but imagine a world where they are enabled to go beyond
their current responsibilities of planning and delivering
lessons. In a world where language models are increasingly
able to perform a wide range of tasks with high accuracy,
it is easy to fall prey to the idea of the death of the
educator. But it is important to remember that language
models are not capable of the same level of critical
thinking, creativity and emotional intelligence of a skilled
teacher.

With language models working as their assistants,

teachers will be able to focus on making lessons more
relatable by multiplying interactions with students. Instead
of fading away into oblivion, the opportunity to exercise
their unique skills of knowledge transfer and empathy will
allow teachers to become superhumans. To understand
how this can be achieved, let’s delve into the world of large
language models and their impact on education.

Impact of Large Language Models on Education

 Page 5

What are large

language models?
Large language models are computer programs that are
trained to predict the likelihood of a sequence of
words.They are typically trained on large datasets of text
and use statistical techniques to identify patterns in the
data and generate text responses based on prompts.

Language models have been widely used in natural

language processing tasks such as language translation,
text generation, and text summarization. With the
advances in machine learning and the availability of large
amounts of data, language models have become
increasingly powerful and are now being used to create
educational materials.

Impact of Large Language Models on Education

 Page 6

How do large

language models

work?
Large language models are based on the idea that certain
words are more likely to follow other words in a given
language, and that the likelihood of a particular word can
be predicted based on the context of the words that come
before it.

There are several different types of language models, but

the basic idea is the same across all of them. First, the
model is trained on a large dataset of text, which could be
anything from a collection of news articles to a database
of books. The model analyzes this data to learn the
statistical patterns and relationships between words.

Once the model has been trained, it can then be used to

generate text by predicting the likelihood of each word in a
sequence based on the words that come before it. For
example, if the model has been trained on a dataset of
news articles, it might predict that the word "president" is
more likely to appear after the word "the" than the word
"unicorn."

Impact of Large Language Models on Education

 Page 7

To generate text, the model typically starts with a seed

word or phrase and then uses its prediction capabilities to
generate a sequence of words that follow. The generated
text can then be fine-tuned by adjusting the input
parameters or by providing additional constraints to guide
the output. Language models can be trained on a large
dataset of text and can learn the patterns and structure of
the language. This allows them to generate text that is
similar in style and content to the training data.

Impact of Large Language Models on Education

 Page 8

Examples of large

language models
One of the most well-known large language models is
GPT (Generative Pre-training Transformer), developed
by OpenAI.

Here's how GPT introduces itself:

There are several other large language models, like Cohere

which use technology to generate new content that is
similar to existing content or data. Among its various
services, Cohere is also developing a multilingual text
understanding model that will empower content creators
to break language barriers and reach a much wider
audience.

Impact of Large Language Models on Education

 Page 9

Impact of large

language models on

education
ChatGPT has been in the news recently for the impressive
outputs based on prompts entered by users. Everyone
from professional content creators to students is
generating essays on the platform. Ethical concerns about
student submissions are increasing and educators are
sounding the alarm about the impact of this technology on
student essays, with worries about attribution oversights,
factual inaccuracies, and bias in assignments. But like any
technology in its early stage, this will also resolve as the
technology matures. Solutions that check for plagiarism
and detect content generated by OpenAI already exist,
and will only improve in their efficiency.

One of the key considerations when integrating GPT in

education is determining how it will be used. Will it be used
to supplement traditional teaching methods, or will it play
a more central role in the learning process? How will it be
incorporated into existing lesson plans and curriculum?
How will educators deal with its limitations and potential
dangers?

Impact of Large Language Models on Education

 Page 10

Answering these questions will help educators identify the

best ways to leverage the unique capabilities of language
models.

Large language models have great potential to make the

task of lessons creation easier and more efficient, and
improve access to education:

1. Automating creation of

learning material
Language models like GPT can be trained to generate
pedagogical materials, allowing teachers to save time and
effort, and focus on more high-level tasks such as
curriculum development and student support.

While this would require significant amounts of training

data and would likely be a long-term project, the potential
benefits are significant. By using GPT to generate
educational materials, it may be possible to significantly
reduce the cost of learning material, making education
more affordable and accessible for students.

Impact of Large Language Models on Education

 Page 11

2. Summarizing complex

concepts / textbooks:
With the increasing importance of micro-learning in L&D,
large language models can play a very important role in
generating summaries of complex concepts, and could
help students who are struggling to grasp a concept or
who need a refresher on a topic.

Automated generation of
summaries could also be useful
for GenZ learners, who have a
limited attention span and
prefer more casual and
informal content. By enabling
educators to quickly generate
concise, easy-to-understand
summaries of key concepts,
large language models could
help students improve their
understanding and increase
retention of complex concepts.

Impact of Large Language Models on Education

 Page 12

Impact of Large Language Models on Education

 Page 13

3. Teaching programming

and other subjects with

repetitive tasks
Zero-shot learning is a machine learning technique which
allows a model to classify objects, or predict outcomes for
which it hasn't been trained already.

While teaching programming, this can be used to

automatically generate code examples or exercises for
students, based on their previous knowledge and learning
goals. For example, a zero-shot learning model could be
trained to analyse a previous exercises completed by a
student, in order to generate a new set of exercises
customized to the student's skill level.

4. Generating assignments

and grading frameworks

Another potential use for zero-shot learning is to use it to
generate assignments and give feedback to students
based on a framework.

Impact of Large Language Models on Education

 Page 14

The model could analyze the student submission and

compare it to a framework associated with a set of
correct, or incorrect solutions, and accordingly provide
feedback to the student.

While this could potentially liberate educators from the

onerous tasks of creating and correcting assignments, it
would not completely replace the teacher, since human
review will still be needed to ensure accuracy of the
assignment and its evaluation.

It is important to note that zero-shot learning is still in its

early stages of research and development, and still needs
a lot of training to mature fully.

Impact of Large Language Models on Education

 Page 15

5. Generating personalized

lesson plans
By using natural language processing and machine
learning algorithms, GPT can understand and respond to
the unique needs and abilities of each student, providing
tailored learning experiences that are more effective and
engaging. A teacher could input a list of learning goals and
GPT could generate a customized lesson plan with
accompanying materials such as readings and activities.

Automating the creation of personalized lesson plans will

reduce the burden of teacher workloads, and allow them
to focus on student mentoring. When you take this to
scale, you can see how it could potentially open the doors
to a new era of democratized access to 1:1 mentoring,
which was denied to so many students before.

Impact of Large Language Models on Education

 Page 16

Impact of Large Language Models on Education

 Page 17

6. Providing additional

support and mentoring

outside classroom
Language models can be trained to provide students with
additional support and guidance outside of traditional
class time, and can be an effective way to reinforce
learning and encourage engagement. Generative-AI
powered mentors can answer students’ questions about
learning material, guide them to additional resources and
provide career mentoring by providing information about
career paths, important skill requirements and emerging
trends in the job market.

Impact of Large Language Models on Education

 Page 18

Future of Generative

AI & Education
At the dawn of the Internet, education experts had fears
about unhindered and uncensored access to information,
rampant plagiarism, and the threats to the status of the
teacher as the gate-holder of knowledge. Decades later,
the world finds itself in a very similar situation with
Artificial Intelligence and large language models
threatening the role of educators.

With each passing month, large language models will

improve in their efficiency with access to more training
data, more computing power, and better models. Sequoia
predicts that by 2023, generative AI will be able to put
together scientific papers and visual design mock-ups;
and by 2030, it will write, design, and code better than
human professionals in the field. Instead of adopting the
ostrich policy, educators should seize the opportunity to
upskill and learn to leverage the powers of large language
models to improve the future of education.

Impact of Large Language Models on Education

 Page 19

Adaptiv and the future of

education
Adaptiv is an AI-powered platform which helps educators
and industry experts create pedagogical material suited to
the content consumption patterns of GenZ. We believe
that this will help us solve the skills gap that is leading to
wide-scale unemployment, career disillusionment and
billions of losses. This will not only help existing teachers,
but also allow others to share their knowledge and
participate in the upskilling of the future workforce.

By enabling automated lesson creation, Adaptiv will enable

transfer of information from subject-matter experts, and
allow people from the industry to guide students and
young professionals about the emerging requirements of
the new economy.

Impact of Large Language Models on Education

 Page 20

Conclusion
Large language models are not a panacea for the
challenges facing education. While they have the potential
to improve the quality of education and increase access to
learning resources, they are only as good as the data they
are trained on. If the data used to train the models is
biased or lacks diversity, the generated materials may
perpetuate these biases and reinforce existing inequities.

Overall, the potential benefits to incorporating language

models in education outweigh its limitations and dangers,
and the possibilities are constantly evolving as these
technologies continue to advance. By carefully considering
how these solutions will be used, providing the necessary
support and training, and designing personalized and
interactive learning experiences, educators can effectively
leverage the unique capabilities of language models to
enhance the learning experience for their students.

Impact of Large Language Models on Education

 Page 21

Disclaimer
The information shared in this whitepaper is not all-encompassing or
comprehensive and does not in any way intend to promote any services of
products. The primary purpose of this whitepaper is to explore the potential
of this new technology and propose use-cases that will allow stakeholders to
make an informed decision.

Prior to any decision regarding the use of large language models in

education, we strongly advocate a careful study of this whitepaper and
other documents on the subject.

Certain statements and information featured in this whitepaper are

forward-looking statements that are based on and take into consideration
certain known and unknown contingencies and risks which in eventuality
may cause the estimated results or may differ factually and substantially
from the featured estimates or results extrapolated or expressed in such
forward-looking statements herewith.

For any information about the examples cited, or Adaptiv, please contact:
[email protected]

Impact of Large Language Models on Education