5 Artificial Intelligence tools you need to start using now to build elearning
For those involved in Education or Learning and Development, it is essential to understand the benefits which Artificial Intelligence can immediately bring to dramatically enhance learning in your organisation. The article showcases several tools that can be implemented immediately to transform one's approach to EdTech. These technologies are valuable to trainers, educators, content creators, and even consumers of edTech content.
Chat GPT serves as the flagship of AI, it has experienced unprecedented adoption rates over the last few months. To put this into perspective, Netflix took 3.5 years to reach one million users, Twitter took 2 years, and Instagram took 2.5 months, by contrast, Chat GPT reached one million users in just five days. The benefits of AI are so obvious that the system had 100 million active users within two months of its launch! Its success is best explained by using it to complete tasks like writing a Shakespearean sonnet, creating a pub quiz, or summarising War and Peace in 200 words. Chat GPT can now even be found embedded into the Bing search engine. It generates answers from scratch each time instead of regurgitating information from a website, making it a unique tool that is not merely a search engine.
So how can this apply to edTech? Ask Chat GPT to construct a curriculum, write a training script or a quiz on any given topic. Chat GPT serves as the base engine for the rapid production of elearning content; although initially there have been some doubts as to the veracity of the data produced by AI, within months the engine has gone from merely passing secondary school tests to acing MBA exams. The results may not always be 100% accurate in all cases, but they can be edited and are generally surprisingly accurate.
Midjourney is one of the most fun AI tools on the market at the moment. Midjourney is an image creation AI engine which has been trained on millions of images in a variety of styles and uses that knowledge to create images from scratch. Images produced by this tool, such as the one below, can be utilised in elearning to create immersive environments in which you can place your training.
There is an issue with AI image generation however, in that the engines have been trained on other people’s work. Midjourney has for instance taken lots of artists' work and broken down their styles and reproduced new images in those styles. The content creators are a little miffed and there are copyright cases around at the moment which are looking for some redress. For this reason, caution should be taken in using Midjourney for commercial elearning production, but restrict its use for internal use where there are no direct course sales.
Other image creation tools look to be on more solid ground, for instance DALL-E from Open AI the people behind Chat GPT - produces images rapidly which are a little less impressive than Midjourney. There are also issues with fingers and writing but the quality is getting better all the time - it seems that many of these issues have been ironed out in the latest GTP 4 release.
Descript is an amazing tool if you want to save time in doing voice overs within your company. The idea is that you train the AI to learn a voice and then when a voiceover task is needed you simply type in the words and a studio quality voiceover is created.
Content creators will be familiar with a scenario where a voiceover is needed for a piece of learning but the managing director is tied up in meetings for the foreseeable future or the actor you used earlier is in a touring production of King Lear and won't be back until spring. Pre-AI content creators would have had to re-record all the audio etcetera in a different voice, but now you can fake new audio using Descript.
The resultant voiceover may lack a little emotion and have issues with the odd word, but that can be trained out. Generally the result is acceptable in most cases and a great addition to the elearning creator’s toolkit.
D-id does a similar job to Discord but this time with video presenters which can be created to vocalise written text or animate an uploaded voice. Editors can upload the face of anyone and it will create a video character.
The videos it creates are convincing, using lip synching technology with any pre-selected voice. The video can be downloaded and used in elearning or as stand-alone resources to add personalisation to learning content.
It would be great if the video presenters could do something to make their characters walk and talk. The technology is there to do this - currently under development in the games industry - so expect to see much more dynamic iterations of this technology later this year.
Openelms.ai combines technology from the aforementioned tools into a single product focussed purely on elearning development. The software will produce an entire elearning course with a quiz from a single line of text, indeed it is currently being expanded to produce an entire curriculum of learning by using the same approach.
The elearning is ‘designed’ by the AI, featuring imagery, animation and video. My company, Open eLMS, offers a premium version which will allow users to produce one new elearning course every month and will be released in April 2023, but can currently be previewed at www.openelms.ai.
AI is a rapidly evolving technology. I remember giving a presentation of some elearning to customers ten years ago where we used green screen actors in an elearning course (a common feature in my company’s work.) The client asked whether, once we’d completed filming, they could change the text and have the pre-recorded actors change what they were saying. We suppressed a few giggles at the suggestion that we could change a recording after the fact. Well, the joke is on me because that is now very possible and this is only the start.
AI improves at the rate it does because it is a self-learning technology, backed up with immense processing power. What amazes us now, will seem stilted and awkward very soon. As an example we have included a picture of a ‘blue bicycle’ unrecognisable as such in MidJourney 1 compared to the photorealistic version in MidJourney 5. This improvement has taken several months and is an excellent example of AI’s ability to train itself to get exponentially better over time.
In a year’s time the images will be moving around in video form, the prospect of creating an entire movie from a single prompt which (for the more megalomaniacal of you) you can star in is a realistic possibility in a few years time. Movie studios will soon be investing in AI engines rather than actual films.
This future is coming very soon. Online learning is changing – that is the lesson AI has to teach us all.