Artificial intelligence for education (abstract).
The disruptive power of artificial intelligence (AI) is radically changing education. In the process, a variety of exciting AI applications are applicable today or in the near future.
1. the creation and processing of texts
2.-4. Conversion to other media formats such as images, audio or video.
5. predictive analytics: the better and actionable data analysis.
6. personal learning coaches
Moreover, artificial intelligence sets completely new standards and demands on us humans in terms of our understanding, our critical faculties and our ethical principles. (7.)
In Silicon Valley, the release of these new AI tools is being hailed as an "iPhone moment," the dawn of a new era. Artificial intelligence (AI) is expected to give a new boost to the tech industry, which in 2022 will be battered by mass layoffs, share price losses and an underlying sense of saturation. - Once again, the Bay Area is in a gold-rush mood.
The disruptive power of Artificial Intelligence has been unleashed in almost every industry since early 2023. First and foremost in education.
More and more people use ChatGPT to learn new things.
Thousands of videos on YouTube today already explain how to use this new technology to learn languages, programming, history, general life questions and much more. Just 12 months ago, this was not an issue at all.
In this article, we would like to specifically address the AI applications that we at MaxBrain, as "Learning Experts", expect to provide the greatest benefit to education.
However, the topic of artificial intelligence for education is vast:
One could write a book from almost every one of the following paragraphs. Moreover, many of our forecasts will very likely prove to be wrong, one-sided or short-sighted.
Nevertheless, it's worth taking a look at these fascinating prospects for the future. We look forward to hearing what you think of this article.
Overview: Applications of artificial intelligence in education
The following graphic illustrates the key applications that artificial intelligence offers for education and training, teachers, and learning platforms:
In the following, we will discuss some of the most exciting parts in this review:
1. create texts faster thanks to artificial intelligence
One obvious application of artificial intelligence is in faster, better, or automated text creation.
Anyone querying ChatGPT for known topics will immediately receive an answer that is qualitatively in the top 10% of possible answers.
Helping teachers write with artificial intelligence is multi-faceted. And the development of large language models (LLMs) promises groundbreaking, as yet unforeseeable innovations in the coming years.
Today, in the first half of 2023, the following applications are available in impressive quality:
1.1. writing texts on "evergreen learning contents
Current AI tools are still largely dependent on data sources from the Internet or learning processes in interaction with humans or internal data.
If there is nothing or too little on the Internet about a specific topic and the human trainers of the AI tools do not know about it, an A.I. - as of today - can still contribute little of high quality in the creation of learning content.
Provided creativity consists of recombining existing information - and it usually does - artificial intelligence can be breathtakingly creative.
The current state of AI development is not yet far enough for completely new ideas and inventions.
Some experts already expect in the medium term that artificial intelligence will be able to carry out research & development on its own - with radical consequences for the whole of humanity, as James Barret, the author of "Our Final Invention" predicted several years ago.
Today, there is also a risk that false content will be created by AI tools based on false information on the Internet:
The problem of distinguishing between facts, Fake News, and the even more dangerous garbled half-truths - too New German: Bullshit - is still a given, but the major platforms are increasingly combating it with artificial intelligence and veritable legions of human reviewers.
At MaxBrain, we call such topics that are part of general knowledge "evergreen topics".
Our assessment of the individual education levels today is as follows:
Texts on evergreen topics can now be created frantically with artificial intelligence at a top 10% level or better.
In contrast, artificial intelligence today can be used to specific topicsIn contrast, artificial intelligence today can hardly generate valuable content on specific topics that concern only one company, unpublished research results, or a small group of people in general, and on which nothing or little is available on the Internet without additional "training.
In a business context, specific internal expertise, values or training on specific internal processes are usually among the topics that an AI tool can only master with training efforts by humans or internal data .
Digression: The ChatGPT Prompts
All the previous remarks only concern the question of how artificial intelligence generates entire texts without any preceding basis.
This is done via the so-called "prompts".
A prompt is a request to the AI tool to execute certain commands.
In the next few years, one of the things we humans will have to learn is how to write good, meaningful, targeted prompts.
The better the prompt, the more effective and productive our use of the new artificial intelligence. You can see the first tips on this in the following video:
For the search query "The best prompts for ChatGPT" there are already over 6 million search results on Google.
There are also already websites that list prompts for ChatGPT for each business area, e.g. Learnprompt.org . - Prompts specifically for teachers are offered free of charge.
We will briefly outline some of such very useful prompts for teachers, coaches or content producers:
1.2 Orthography: Rapid corrections
Proofreading texts can be time-consuming.
Quality-conscious people write texts, e-mails or even learning content and then spend most of their time - often unnoticed - correcting their texts.
AI tools can speed up this process a hundredfold: instead of taking 30 minutes to correct three A4 pages, ChatGPT can do it in 15 seconds or less.
The perfection of artificial intelligence also promises fewer errors than the usually less precise human eye.
A very simple prompt for this useful function is:
Please correct the text: [...]
A small example:
1.3. translations of learning content via A.I.
Translations of learning content are also becoming easier and easier and are usually possible very quickly and in impressive quality with artificial intelligence. For teachers, the translations of AI tools offer at least a very good starting position that saves an enormous amount of time.
In addition, there are plugins for certain software providers that translate components automatically.
An example would be the WordPress plugin Weglot, by means of which administrators of a WordPress website can switch on a language on the entire website within seconds.
The MaxBrain website is a good example: just change the language at the bottom of the footer of this website and the Weglot plugin will translate this article for you in a few seconds.
One formulation for a translation prompt might be:
Translate the following sentence to [language]: [...]
A concrete example:
1.4. summaries made quickly
As mentioned in our article "Creating learning content for your LMS in the best possible way [Checklist 2023]", summaries are among the most important elements of a successful learning program.
Summaries can also be used later in other formats, such as short listening units or videos.
The idea of combining content from one website is also exciting.
Prompts for summaries could be something like:
Summarize the following text to [number of lines]: [...]
Summarize the following website article to a maximum of [number of lines]: [link]
Example from ChatGPT:
1.5. simplify the language by means of artificial intelligence or change the tone
Sometimes it pays to simplify language or adapt the tone to the target group you want to reach with a learning content. I speak from experience: I regularly write something that is too tedious, not very understandable or not pointed enough for others.
In the case of learning content, in any case, it can be fatal not to consider the target group:
Learning units that are too mundane or too difficult are guaranteed to defeat their purpose.
Impressive and funny are also prompts to change the tonality. For example, they read as follows:
Write the following text for [description target group]: [...]
Examples from ChatGPT:
Our project leader Ziad Fathallah had the following suggestion to show how multifaceted ChatGPT can imitate the tonality of well-known and even virtual personalities:
Of course, in our fun MaxBrain office, this exercise immediately led to dozens more tests and ideas.
Badass with Rambo to the point:
Excursus: Integration of A.I. Text Tools in Editors and Learning Platforms
Some editors on the market already offer ready integration with ChatGPT.
An example that I use on this website: The Elementor Editor for WordPress. This is what an automatic translation into English looks like:
MaxBrain's authoring tools, most notably Articulate Rise, are also in the process of integrating the handy AI tools for text creation, as mentioned a few months ago in the blog article ChatGPT AI - A new starting point for content creation?
At MaxBrain, we are working in many other places to integrate artificial intelligence to create faster, easier, and better learning content.
2. converting text to images using AI
Texts are, of course, only one of many formats that learning platforms can use.
Artificial intelligence can convert texts into a variety of other media formats, such as:
The possibilities are continuously being expanded.
In this article, we will focus only on the three most relevant formats for education: Images, Audio and Video.
2.1 Creating new images with Midjourney & Co.
Anyone who creates content knows the problem of finding suitable images or creating them themselves: It can be time-consuming to find an image that fits a specific topic.
Pictures without reference to the learning content are usually rather confusing.
Today, there are more and more artificial intelligence providers to generate images from texts. The following AI tools are currently receiving the most media attention:
By far the most advanced of these AI tools as of today (June 2023) is certainly Midjourney.
It requires some training time and, unlike ChatGPT, no longer offers a free version.
What Midjourney is superior to the others in, according to my tests so far, is the creation of photorealistic images of people. All the others struggle with this and regularly produce monster versions of faces.
While the quality of images generated via midjourney is impressive in terms of composition and level of detail.
However, I personally still find it difficult today to use AI tools to create images that fit a particular style or imagery in a reasonable amount of time.
So for this blog post, I tried to generate a cover image with Midjourney.
With the certainly somewhat simple prompt "an image for the use of AI in education", Midjourney generates the following image suggestions:
Perhaps with the exception of the first proposal, these are all successful, impressive images.
Only they don't match the style of the rest of MaxBrain's website and leave a rather gloomy impression for my taste.
Artificial intelligence to create images from a text input needs to evolve significantly and become simpler as of today (June 2023) to become an integral part of work processes, as ChatGPT is already succeeding in doing for text processing.
However, artificial intelligence is expected to meet these requirements very soon: In addition to fast-growing startups like Midjourney, global market leaders like Adobe are working hard on the topic.
2.2 Image processing with artificial intelligence
Similar to text processing, artificial intelligence today is not yet perfect for creating images out of thin air without a specification.
Examples that I have been using frequently for some time now:
We can be curious about what will be possible in terms of image processing in the near future.
In any case, the education industry can look forward to more efficient and perhaps even more creative images for its learning content.
3. converting text to audio formats
Audio formats - especially podcasts and audio books - are becoming increasingly popular.
Once you discover the possibility of using podcasts to turn the boring time spent stuck in traffic on the way to work into valuable learning time, or to listen to exciting audio books while going for a walk, you will permanently appreciate the audio format and be happy to get used to it.
In some cases - such as jobs that require a lot of travel time - audio formats are often the only way to consume learning content at all while working.
Audio formats can also play an important role in the internalization of learning units; apart from the fact that many people clearly prefer to have a text read to them rather than having to read it themselves.
Exciting for teachers, coaches, and learning platforms:
The biggest challenge that complex algorithms and learning processes demand from AI tools is intonation.
Intonation and pauses between words determine whether a spoken text comes across as monotonous or exciting. Especially with longer texts or emotionally charged stories, very few people can stand monotonous speech.
Intonation and pauses are directly related to the context of a content: One must understand a text in order to identify the exact words and syllables that need to be stressed or require a pause in speech.
Sometimes intonation is conveyed by commas or pauses, which make significant content distinctions but are not explicitly stated:
Artificial Intelligence Provider for Text-to-Speech (TTS)
There are now quite a few speech-to-text providers that manage to get very close to human speech via artificial intelligence and imitation.
The most conspicuous providers on the market today are:
In our opinion, Murf.ai is the best TTS (text-to-speech) tool on the market today (June 2023); among other things, because the intonation ("emphasis") of a word, as well as the addition of pauses of different lengths, can be corrected afterwards in the editor of Murf Studio:
As you can hear from the following audio file generated by Murf.ai, the cooperation of man and machine can already deliver relatively good results:
Another provider that has been well-known on the market for some time is Speechify. For English-language videos, Speechify offers, among other things, voices of well-known personalities with which you can have your texts recorded. However, the current "standard voices" in German are not pleasant enough in our opinion.
Descript Overdub is also interesting: with this tool you can record your own voice. The artificial intelligence learns your pitch, accentuation, speaking style and can convert entirely new texts into recordings with your own voice. Provided you - unlike me - like your own voice, this is an exciting option.
4.convert text to video formats
Video is the most convenient and therefore preferred format to consume content on the Internet.
For education and learning platforms, this actually means that online content should preferably be offered in video format.
The problem with videos:
For most content creators, creating video is time-consuming, usually expensive, and requires knowledge to learn.
Artificial intelligence will help: After all, videos are ultimately nothing more than the combination of content (texts), audio and image sequences.
A whole range of providers are therefore already trying their hand on the market today, including:
We have tested these AI tools.
1. at the present time (June 2023), all text-to-video providers leave a rather experimental, immature impression.
2. the quality of AI tools and adaptability to the needs of the education industry are still too limited.
3. professionally created, "man-made" videos are still today by dimensions more beautiful, more creative, more exciting and more involving.
This conclusion does not surprise us.
Video is the supreme discipline:
To the extent that artificial intelligence understands text, audio, image creation, and additionally image sequencing, it can master video creation.
All these puzzle pieces are still in their infancy. Their synthesis into a high-quality video is not yet practical for higher quality demands.
However, it is to be expected that our rather cautious assessment of text-to-video generators will be much more positive in the future.
Do we now have to wait a few years until artificial intelligence is usable for video production?
It depends on what demands we have on the quality of videos.
For many purposes in the education industry, the above AI video tools may be sufficient.
In addition, there are other solutions around more efficient video production that are exciting for education:
For example, the aforementioned vendor Murf.ai is in the process of developing a system that combines Google Slides presentations with audio AI to generate videos.
The idea: Create explainer videos from presentations.
We have tested the Murf add-on for Google Slides.
The result after about 1 hour of work is not perfect. However, the potential of combining Google Slides with an AI voice-over tool cannot be dismissed:
We will now continue our review of artificial intelligence applications in education in this blog article.
5.1 The problem with the data: Insufficient utilization
Predictive analytics involves data analysis to predict and exploit future events or outcomes.
Not a new idea in itself:
Many tech companies and SMEs claim to be "data-driven"; after all, this sells a bit more objectivity when it comes to added value, customer needs, trends or optimization potential.
This wishful thinking usually does not correspond to reality:
The big challenge:
Education is no different:
Learning programs are hardly ever offered personalized.
In elementary school, every child is confronted with the exact same learning content, some of which dates back to the century before last - the Prussian educational model.
In higher education and further education, learners tend to select the subjects and topics that interest them, but in terms of learning content within this selection, they are usually uniformly locked into the same learning program.
Providers of modern learning platforms, on the other hand, are slowly but surely recognizing the potential of personalized learning experiences.
In our blog article "LMS vs. LXP: A Change of Perspective in Education" we have already gone into detail about the difference between a Learning Management System (LMS) focused on administration of learning programs versus a platform (LXP) focused on the ideal learning experience.
5.2 Artificial Intelligence: A New Hope for Predictive Analytics
Hope summarized in one sentence:
Almost seven months after the release of ChatGPT, much is still speculative here as well.
The following use cases are conceivable thanks to "smarter" or even real data utilization:
Providers of modern learning platforms such as MaxBrain are developing more and more features and functions for these purposes through close, hands-on collaboration with corporate customers.
We are at the beginning of a development here that will fundamentally change education.
6. the personal AI coach
Another groundbreaking idea for education is that of AI coaches who provide individualized guidance to learners.
As in other areas - such as medical, psychological or legal care - there is a heated debate among experts today about whether every person should have their own own coach in order to learn faster, more easily and in a more targeted manner.
As for all other topics in this article, we asked ChatGPT.
The answers were particularly interesting at this point:
How can artificial intelligence coach learners?
Artificial Intelligence (AI) can coach and support learners in a variety of ways. Here are a few ways:
1. personalized learning content: AI can analyze learners' learning behaviors and needs and provide personalized learning content. Based on individual strengths, weaknesses, and interests, AI-driven systems can customize learning materials and exercises to provide a tailored learning experience.
2. feedback and assessment: AI can provide learners with feedback on their performance and assess their progress. By analyzing answers and solutions, AI can provide detailed and immediate feedback that helps learners improve and boosts their motivation.
3. learning paths and progress tracking: AI can monitor individual learning progress and suggest appropriate learning paths. Based on a learner's strengths and weaknesses, AI can recommend the order and pace of learning to ensure an optimal learning curve.
4. speech and language exercises: AI-driven language assistants can help learners improve their language skills. They can simulate conversations, offer corrections, and help with pronunciation to develop language comprehension and communication skills.
5. virtual tutors: AI can act as a virtual tutor and provide individual learning support. Learners can ask questions, discuss problems, and receive additional explanations without relying on the presence of a human tutor.
6. time management and learning planning: AI can help learners organize their time effectively and create learning plans. Based on learning objectives and available time frame, AI can set reminders, prioritize tasks, and optimize learning schedules to ensure efficient use of available time.
7. emotional support: AI-driven systems can also provide emotional support by giving learners encouraging and motivating messages. By integrating AI into learning environments, learners can feel supported and cared for.
It is important to note that AI should be considered as a complement to human instructors when coaching learners. The combination of AI and human empathy can ensure an optimal learning experience.
Reading ChatGPT's response sounds like futuristic utopia:
Something that only a few cranks in the tech industry would like to have, and which - if at all - is still many years away. No need to worry, dear education system!
This mindset is dangerously short-sighted and ignores successful implementations of this idea of an AI learning coach today.
For example, as early as May 2023, the founder and CEO of Khan Academy, which is very successful in the US, Sal Khan, presented the implementation of an AI coach for basic education at a TED talk:
Given this very concrete example of an AI coach, the next obvious question to ChatGPT was:
What can human teachers do better than artificial intelligence for teaching?
Human teachers bring certain skills and qualities that artificial intelligence (AI) has yet to fully replicate. Here are some aspects in which human teachers have advantages in the classroom:
1. emotional intelligence and empathy: Human teachers can perceive emotions, respond empathetically, and establish a supportive relationship with learners. They can identify individual needs and challenges and respond appropriately. Empathy and interpersonal connections play an important role in learner motivation and well-being.
2. adaptability and flexibility: teachers can quickly adjust to changes or unforeseen situations in the classroom. They can adapt their teaching methods and strategies to meet the individual needs of learners. The ability to respond spontaneously to questions and discussions is also a strength of teachers.
3. social interaction and group work: Teachers encourage social interaction and collaboration among learners. They create a learning environment in which students can communicate, share ideas, and learn from each other. Fostering teamwork, communication skills, and social engagement is an important role of human teachers.
4. teaching values and character building: In addition to teaching pure knowledge, teachers can also teach values and ethical principles. They can discuss moral dilemmas, encourage critical thinking, and support the development of ethical behavior in learners.
5. mentoring and individual support: teachers can take on a mentoring role and provide individual support. They can help learners define their goals, overcome obstacles, and reach their full potential. Personal guidance and mentoring are important elements of a holistic education.
6. creativity and inspiration: Teachers bring their own creativity and passion to the classroom. They can develop innovative learning methods, activities, and materials to spark learners' interest and curiosity. Through their own enthusiasm for the subject area, they can foster student motivation and engagement.
These human skills complement and extend the potential of artificial intelligence in the classroom. The combination of human teachers with AI-enabled technologies can provide an effective and enriching learning experience.
You may have wondered during the course of this article how much has been written by me, the author Andrea Anderheggen, and how much by ChatGPT.
The somewhat sobering answer is:
ChatGPT did help me to collect ideas and correct texts, yes!
However, writing this article took about seven net working days; about twenty thousand times longer than the 10 seconds it would have taken me to write an article using only ChatGPT.
Many strands of thought in this article are not what you will quickly find in a Google search. Because my motto for blog articles is:
"More exciting, detailed, and practical than what readers will find on the same topic on the Internet. Or, leave it alone."
Accordingly, ChatGPT couldn't write a blog article like this one; at least not without time-consuming training on my part that probably would have taken longer than 7 business days.
From this personal experience, I deduce the following as of today (June 2023):
And this also applies to teachers, coaches and mentors, who have a significant influence on education and thus on the knowledge of mankind.
Still, artificial intelligence is changing education forever:
Firstly, via the applications already mentioned above.
On the other hand, teachers, coaches and learning platform providers have now been given new and exciting homework:
Artificial intelligence is not only changing our everyday lives, but also the demands placed on us humans.
Artificial intelligence not only changes our learning content and learning processes, but also requires new skills and competencies in dealing with AI.
Many of the existing human competencies will be completely taken over by artificial intelligence. Such competencies will become obsolete for ourselves.
What this means in concrete terms is largely open today (June 2023):
Exactly what we need to learn or forget in order to deal with AI is related to the evolution of applications and tools around artificial intelligence.
However, some important skills that we will need to relearn or improve in order to use AI tools are foreseeable today.
These include the following:
Artificial intelligence is comparable to a genieThe more precisely we formulate our wishes, the more likely the AI tools can help us.
Easier said than done:
Often we don't even know what we want because we don't know the basics and consequences of what we want or we haven't thought the problem through enough.
As one software developer recently put it on Twitter:
And that's just the beginning:
The responsibility of questioning information, texts, images or videos of an artificial intelligence is followed by another, much more difficult one:
This new requirement is probably the most difficult and at some point possibly the most important of all:
And this becomes a pressing practical problem as soon as you start talking about artificial intelligence or worse "Superintelligence" is being thought about:
For education, this global lack of universally accepted ethical principles and rules derived from them raises many practical issues:
Segmentation vs. discrimination: How exactly does one ensure that personalization of learning programs via data-driven segmentation of learners does not lead to individual learners being discriminated against?
Freedom to learn vs. course recommendations: To what extent is there freedom to learn what one wants if course recommendations only display learning content at all that an AI's algorithm particularly calculates to be appropriate for individual learners?
Data protection: How is learner data used and protected in a way that does not violate their personal rights?
Creativity: How do you ensure that people still remain creative when an artificial intelligence provides you with the framework of knowledge and creativity?
We will have to answer these and other difficult questions if we do not want to risk the applications of artificial intelligence degenerating into a mere instrument of power for a few interest groups.
"Knowledge is power." as Francis Bacon clearly put it some 500 years ago.
And learning determines our knowledge.
Once again, I personally see no other option than to simply trust the leading developers of powerful tools such as artificial intelligence and occasionally steer them in an acceptable direction via boycotts or laws; as we have already had to do more or less successfully with the inventors of the atomic bomb, genetics, the computer, the Internet or the iPhone.
Instead of fighting the unstoppable development of artificial intelligence, it makes much more sense to minimize its risks, understand its potential and use it for the benefit of others.
Did you like this article?
If yes, please share this article on your social media channels. Thank you very much!