Learning System Trends 2024

0
86
Learning System Trends 2024
  • Any learning system or learning technology or authoring tool – must have a voice narration capability – not just a text-to-audio for the content, but also what I see as a “do this” is the ability for an end-user to move around the system by using their voice. “Open course X.” “Start the recommended playlist’s first course.” “Scroll down and click on the calendar. Click on Tuesday. Click on available time.” “Register me.” No system does this. Sure the majority are ADA508 compliant or similar (I always forget what that is). However, not all of them are, and secondly, when it comes to the voice moving around and taking content within the system – navigation and so on – it’s not there. It should be. It can be if a vendor either built it into their system OR tapped into LAM (which is what takes a gen AI further – which you are unaware of, but it exists and is another AI option. OR tap into machine learning (again AI).
  • When seeing the prompt window on the learner side or admin side, provide your clients with “recommended,” or “top 50 prompt modifiers,” and so on. On top of that, allow someone to say it, and then it is typed in, and they say “click,” and the response shows up. Today, it is all click, click by using your hand.
  • When I hear the vendor say they have AI – I ask them immediately if it is machine learning, gen AI or a combo of both (it can be)
  • If the vendor says combo or they may say it depends on what the capability is – then find out. I can easily spot something and go that machine learning because it looks like it is – so far, I’m 100% accurate (knocks on wood table). I wouldn’t expect you to be unless you see a lot of these or know AI and can spot the difference between the two.
  • I want to know what LLM they are using, or if they say we use multiple, which ones. Is it relevant to you?
  • GPT Turbo, aka GPT 4.5, is the latest version of GPT. It is fee-based. Folks on enterprise with OpenAI, you are using Turbo. There are vendors using 4.5, which is good. The downside? Token fees. Sure, they are low now because, overwhelmingly, the vendors using Gen AI have it on the admin side—create courses/content, identify skills, and create assessments with AI.
  • GPT 3.5 did far better than 3, the freebie of ChatGPT. Folks who were on 3.5 are tapping into 4 (whichever version still has a lower token fee), and others are going into 4.5.
  • OpenAI – the makers of the above LLMs – dominates. It is not even close in our industry. Number two is Azure’s LLM which is built on Open AI’s version.
  • Llama 2 – 100% free open source, which means no token fees. Llama 3 just rolled it and Meta claims it is as good as GPT Turbo. It’s also 100% free open source. Vendors that go this route, take this LLM and do a lot with it. I do know of vendors that tapped into LLama.
  • Google Cloud—Model Garden requires you to use Vertex ML, and then you can select from a variety of Gen AI LLMs.
  • Foundation model – Pretrained multitask large models that can be tuned or customized for specific tasks using Vertex AI Studio, Vertex AI API, and the Vertex AI SDK for Python.
  • Fine Tuneable model – Models that you can fine-tune using a custom notebook or pipeline.
  • Task specific solutions – Most of these prebuilt models are ready to use. Many can be customized using your own data.
  • It requires a lot of computing power – which is why you are seeing small LLMs that can go from a mobile device, with minimal computer power. For the Open AI and others, though, those computers – and the power it is creating a massive carbon footprint – are higher than anything on the cryptocurrency side.
  • To cool the machines down, you use water—a lot of it. Since water is crucial nowadays more so than ever before, yeah, I’d say this is relevant.
  • Create courses/content (I refer to it as a content creator tool, vendors will say authoring tool) – Only one I’ve seen today is robust, whereas an ID person could use it to a point. With synthetic audio, materials too. The rest is headers (I prefer Chapter titles), then pages (they call it something else), and in some cases, content within those items – more often than not, actually. A few have the WYSIWYG window, and in quite a few cases, you can change the text by choosing (professional, witty, serious, and so on). Other vendors use the emoji angle. I wish I could remember what one vendor had as an option with the text change. It was so demeaning, and offensive that I called them out and stated such. I need to go back to my notes and find the vendor, then see if they fixed it. My guess? Nope.
  • Create an assessment with Gen AI. I’ve Seen it a lot. This is the one I hear. You haven’t seen this before, spin. I see a lot of systems, so yeah, I have. Some are better than others.
  • Identify skills that can be presented on the learner side with AI. These skills can be tied to systems that allow the opportunities aspect in their system, or someone uploads their resume, and it takes out key items and lists the skills. I am starting to see vendors tie skills directly with content initially using Gen AI. Equally, I see machine learning with the recommended content/course angle, which looks skewed more often than not due to the algorithm.
  • A couple of vendors are launching or have launched a prompt window on the learner side. The learner can ask questions or statements and the window responds with the answer or supposedly answer. Yes, it can be wrong – hence, always verify before just thinking, “Yeah, this is right, ” and roll on.
  • There is no prompt window either on the admin side or the learner side, whereas if the information that was outputted was wrong, the person can state so and put in the correct retort. AI learns from itself, so if you are not saying this is wrong, then surprise – it assumes it is right. I continue to be stunned that zero vendors are doing this. Every LLM you test out, heck go to the freebie one, has this window.
  • Minimal vendors add the text that the system may output mistakes (a term I see in vendor land) or fake or false information, so review. I’ve seen this text in only a few vendors. As noted, a couple told me they had clients who wanted it removed. That is like saying, “jump out of the airplane, and what? We didn’t teach you how to open it and when. Don’t worry, nothing will happen to you.”
  • The content: The client selects what they want and by what publisher/provider and identifies the number of seats they need (licenses) for that specific content. The seats, by the way, do not have to match all your user base—this content is for this group and only for these folks.  
  • Depending on the vendor, you will see many different third-party providers, minimal, or, yes, I can’t believe I am saying this in 2024: zero.  
  • The plus for you is that you are not required to go to the internet and type in some words, hoping to see what content is ideal for you. Nor should you type this into Gemini (formally known as Bard) or Copilot and similar for search engines. They all scour the web and pull down quite a bit of those worthless top 10s written by someone who has no background in the industry but loves van travel. All of these, those top 10, are affiliates of the vendor. Yes, they make money off you. You will see this with the “we may receive a small commission.” 
  • The benefit is to go to the content marketplace with your vendor and see what content is available based on your use or use cases.   
  • Vendors should consider adding more content providers, especially those that are reliable and can filter out the junk that’s prevalent on the internet. This way, if a client already has a trusted content provider, the integration is already in place, eliminating the need for additional work that may or may not be successful. There are plenty of learning systems out there that have content/course aggregators, such as GO1 and OpenSesame. I often hear that both are the same and have the same publishers. That isn’t the case. There are pros and cons to each of the above. I like OpenSesame (yes, that is the vendor with the built-in AI, and OS will work directly with your client to identify their use cases, and what content is best suited for that). Thus, the idea that you get all this content with a subscription plan – think all you can-eat aspect – is eliminated. OS is one of many vendors that do this. Content Anytime from Cornerstone does, too – but CA is only for Cornerstone clients. OS and GO1 can be used by anybody, regardless of the system. 

I’ll stop at Trend Six. I could easily go into Trend 20. I am holding off because typing with a bruised bone in my wrist is no longer attainable without pain. I guess that is a trend of mine.