Choose one of the
account to be activated!

If you already have an account Login here

If you have any doubts consult our FAQ or write to us at [email protected]

No products in the cart.

Andrea Migliorini
AI won’t do your homework. A student using AI will
Feb 13, 2023
5 Minutes

You might have already heard about New York Public Schools’ ban on ChatGPT. This reaction is just one of the latest surrounding ChatGPT, which has become a buzzword in the blink of an eye. In recent months, teachers and educators have had very different reactions to ChatGPT, spanning the gamut from enthusiastic statements to the opposite.

 

One thing we know for sure: AI can provide educators with the chance to train and identify new skills. The question is: will this also be a chance to reinvent the way we assess knowledge?

Key fact: How does ChatGPT work?

First things first. What is ChatGPT and how does it work? ChatGPT, aka Generative Pre-training Transformer 3, is an artificial intelligence chatbot able to conduct conversations with users and generate human-like writing on a (potentially) infinite number of topics. Magic? Not really.

TO SUM UP
It’s all about Large Language Models, that is, models using advanced algorithms to process large amounts of data. Here are some key tech facts to better understand how this works and where it comes from.
Process language

Many languages. It's been a paramount advancement in the branch of Natural Language Processing (NLP).

Mirror a brain

A smart brain. ChatGPT’s algorithm works like a "neural network": a type of computer program based on the way human brains work.

Be trained on data

Lots of it. It’s built on 175 billion parameters (different kinds of texts). To get the idea, that’s more than 20 times the world population. 

Answer questions

Millions of questions. ChatGPT's algorithm uses data to learn the patterns and structure of language, generating fine-tuned human-like answers.

TO SUM UP
ChatGPT is an AI chatbot based on billions of human text parameters. It can generate human-like responses on every possible topic you can think of thanks to a machine learning model known as neural networks. And yes, the results, while sometimes slightly inaccurate or utterly wrong, are quite impressive.
AI in Education: Not only ChatGPT

If you think the AI issue includes ChatGPT only, you might need a little catch-up. Indeed, ChatGPT is just one AI tool out there. In today’s market, one can find things like chatbots, speech recognition devices, and adaptive learning technologies in the AI education landscape. However, none of them had the impact ChatGPT had on a broad audience. 

 

For example, neural networks – the same type of technology ChatGPT is based on – have been trained and used for other purposes too, much more related and designed for education. In a recently published study led by Iddo Drori, Associate Professor of Computer Science at Boston University, scientists explore the possibilities of how neural networks can create and solve complex problems in different mathematical modalities (numbers, equations, plots), designing and self-cracking university-level mathematics courses – no joke, we’re talking about MIT level problems.

The importance of neural networks
Why is Drori’s study groundbreaking?
Well, before this research, no one believed neural networks could solve math problems, while this particular network in action had 81% accuracy - which is very similar to human-level results.
How can Drori’s study benefit education?
If an algorithm is able to create and solve complex problems, it could be a great way to generate tailored tests and learning experiences that address specific learners' needs, provide feedback and help teachers to focus on the actual teaching.
The need to rethink how we assess knowledge

As The Atlantic CEO Nicolas Thompson said in a recent video published on his LinkedIn profile: we need to understand how to leverage AI in Education, rather than ban it: “Chat GPT is an incredible learning skill. Just like you need children to know how to use Google, you now need them to know how to use ChatGPT”. The hardest problem this tool brings with it, as you can easily imagine, lies in cheating.

TO SUM UP
ChatGPT was the straw that broke the camel's back. However, tools like this are only the beginning of a revolution we will see in the ears to come, and we need to get familiar with them.

Let’s get back to where we started. The main reason behind the ban from New York Public Schools can be summarized as follows: if students can ask ChatGPT questions and then copy-paste answers, their critical thinking skills are at risk. Another fear is that, as Stephen Marche puts it, this can signal the end of college essays. However, this is not bad news. And critical thinking is not on the verge of extinction – at least, not as things stand right now.

Bye bye mimicry

Let’s focus for a second on how we tend to assess knowledge at every level of education. Think back to your own school experience. Too many times, our knowledge was evaluated based on mimicry, with students studying and regurgitating information they’re going to forget in a week.

 

To overcome this situation, we can think of standard alternatives, like oral exams, live writing tasks, and software that detects cheating (like the one created by this guy at Princeton). However, radical innovation calls for radical change: we need to change our approach to the subject matter of assessments altogether. 

This is the chance to overcome mimicry-like tests. First and foremost, the future will require a forgotten skill: the ability to pose engaging and useful questions. In this case, this can translate into prompt-writing lessons. Did you know there are already prompt engineering job offers?

In a long-form published on Noema, Jacob Browning and Yann Lecun point out that these models’ understanding of language, while impressive, is shallow. They mention how we’re used to learning at school as an example: right now, they state, assessments are structured in a way that trains jargon-spouting students who don’t actually understand what they’re talking about, simply mimicking what professors and textbooks said.

TL; DR: We did it with Google, what about doing it again with ChatGPT?

Schools and educators fear AI will kill critical thinking and other standard skills, like writing. However, the rise of tools like ChatGPT could be a chance to improve and adapt to new standards in the way we assess knowledge, overcoming the mimicry-like model. Considering the role and bandwidth AI will have in the foreseeable future, schools and educators need to help students harness and leverage AI in their daily tasks, focusing on the new skills the Future of Work will ask for.

TO SUM UP
Here are the possible benefits AI tools like ChatGPT and neural networks can bring to the classroom: for teachers, new ways to assess knowledge and more time to actually teach; for students, to gain new skills and follow more personalized learning paths.
Stay in the know

Subscribe to the first non-boring newsletter dedicated to Education