Do we strike you as late adopters? The Education Marketing consultancy who are helping institutions adapt to new technologies, new channels of engagement and the arrival of Generation Alpha? Surely not.
But kind of. Do you ever see lots of hype around a movie or TV show, and it actually puts you off watching it? It feels like, if you watch that movie when the noise is at its peak, all you can hear is the voices of others, and you can't find your own.
We've been in on ChatGPT since it first opened its robot arms to the public, but we just haven't felt like talking about it yet. So much hype has surrounded it; from the "end of education" to the "collapse of the human creative process".
It is neither of those things. If anything, it is the death of Google and the final death blow to old-fashioned essay-based assessment methods. However, in terms of creativity, it is simply another stimulus in the way we employ human intelligence, and AI is nothing without us.
We feel like people are starting to find a more sober analysis of ChatGPT on offer now, as we start to see the possibilities of it all, and also to simply embrace that this change is already here. We will be dipping into ChatGPT from a few angles over the coming months, as well as looking more broadly at OpenAI, the company behind it, and AI in general for a broader overview of course.
So if you were looking for a short article on ChatGPT that doesn't use the word "apocalypse", then you've come to the right place.
So there's this web-based tool, which is actually called ChatGPT-3 (yes, the number means there will be more of them, and it's just getting started.).
ChatGPT-3 generates natural sounding language based on the prompts we give it. That sounds simplistic, but consider that the tech draws on 175 billion parameters, which are components that the tool itself can modify as it learns, and helps it make better decisions. If you've had your coffee this morning, you can learn more about difference between a parameter and hyperpamarameter.
This creates a huge "neural network" and perhaps that terminology is part of what scares people. This thing can learn, and when it is learning how to do better, it can use self attention mechanisms to focus in on the bit that is in doubt, such as what the right word to use is in the specific context.
Researchers tested ChatGPT-3 on Natural Language Processing benchmarks, which they have handily shortened to NLP to annoy the Neuro Linguistic Programming community. These researchers found that it could pass many standardized tests, and even generate samples of newspaper articles which experts had difficulty in identifying as AI-produced. This thing is amazing.
This tool was developed by OpenAI a San Francisco based research Laboratory. The company is a "capped profit" model, and you can learn more about its ownership structure by watching this short video here. This is important because its business model is very much based around strategic partnerships and, considering the power and potential of this tool, we will keep a close eye on the companies who partner with them and what it might mean.
Yes, you guessed it, there is a lot more to this than simple "text generation" might imply.
Firstly, this isn't just any text. This thing can write Haikus. We asked it to "write a haiku about an education marketing consultancy called geNEOus (formerly NEO Academy), who are writing about ChatGPT" and this is what it wrote:
geNEOus guides minds to learn,
ChatGPT aids the way to teach,
Now, that is not a bad haiku, but it also neatly makes the next point for us. This thing does not just generate simple text, but it actually generates ideas. We have used it to
brainstorm ideas for a marketing strategy
find better ways to explain complex information to non-specialists
generate discussion prompts for workshops
write the first draft of licensing agreements for partnerships
check code we've written for potential errors or security weaknesses
That is just scratching the surface. There is so much more here, and we will take these things in more detail in subsequent articles. The point is this. We humans are the ones on the front lines. We tailor the prompts to get the information. We evaluate that information for accuracy (yes, ChatGPT makes errors all the time, but sounds so plausible when it does. Be careful!).
AI is nothing without HI (is that even a term?). By using AI to do some of the heavy lifting, we have already found that it frees up more time for the creative stuff a machine cannot do. This is much the same as happens when we automate CRM processes for a client, and they suddenly have more time to do actual strategic thinking, and maybe even have a cup of tea with colleagues and find out how they are really doing. We guide, ChatGPT aids, and education blooms.
Look at what ChatGPT-3 can do already, and it isn't even connected to the internet. That's right; it was taken offline in 2021, so it still thinks Yeezys are pretty cool sneakers. ChatGPT-4 is already on the way with 100 trillion parameters as opposed to the current 175 billion.
This is still not a multimodal tool, so will be text only, but soon this thing will be able to design a clothing range and the marketing assets to support it, mark our words. When it comes online, watch it build a website for you in real time, and run your social media campaigns. Seriously. In fact, you only have to wait until March 2023 to see Microsoft integrate it into Bing search engine to see how things will change. Uncle Bing is about to get a new lease of life.
The initial fear-based hype is subsiding, and we think it's a good time to delve in deeper about the real opportunities and consequences. Keep an eye on our blog for how AI can be used for education marketers, the impact it might have on certain professions, the opportunities it opens for forward thinkers, how education institutions might use it, what it means for Google, and so much more.