Marcus Richards
Generative Aritifical Intelligence chatbots, commonly referred to as large language models, have taken the world by storm since we caught a glimpse of Chat-GPT.
Copywriters, poets, lawyers, students and everyone inbetween have found uses for the platforms, from producing university admission applications to medical papers.
But as the adoption of the technology upticks, so too do the errors and there have been numerous highly-publicised blunders by the software ranging from opinion pieces drawing on false information to coding tests, after initially becoming the darling of, well just about everyone.
The bots have been in development for years, but since ChatGPT3 became a viral phenomenon last December, companies have scrambled to rush better versions out to consumers. Chances are they are here to stay, so in order to stay relevant, AI chatbots need to constantly evolve and improve. Here are a few ways they could do that:
Be more human-like
One of the biggest challenges for AI chatbots is to be able to mimic human conversation in a way that is natural and engaging. This means being able to understand complex questions, respond in a timely manner, and generate text that is grammatically correct and free of errors.
It’s a weird space for the developers to navigate. On the one hand, users demand that the content generator writes like a human for authenticity. On the other, many are concerned with chatbots becoming overly realistic and, thanks to the Black Mirror effect, sentient.
We’re not quite at the point where chatbot and human content are indiscernable; only around three in ten people can’t actually tell the difference.
But if the measure of the language models’ success is the human mimicry, we should expect that number to rise and logically that means the quality of chatbot output should too.
Be more proactive
In addition to being able to answer questions, AI chatbots can also be used to proactively engage with users. This could involve sending out notifications about important updates, offering recommendations, or even starting conversations with users based on their interests.
Adopters of bot technology are between a rock and hard place somewhat with this one, given most of the current widescale usage of the technology consists of chat window helpers on sites where you buy products or services.
Typically, these bots are either passive or reactive meaning they have to rely on a human user first interacting with them.
In the customer service world, predictive analytics can enable chatbots to make personalised recommendations and guide customers through their purchase journey before they ask for the help themselves.
For example, if a shopper is spending extended browsing time on a delivery or returns page, a pre-emptive chatbot would pop up and ask if the customer needs to know more about the returns policy.
They are, in some instances, used to great effect during a shopper’s journey. In an era where so much shopping is done on a mobile device and internet connections are patchy, purchase journeys are often incomplete.
If a bot notices items are left in a bag before checking out, a prompt can help convert a sale that would otherwise disappear.
Be more accurate
One of the biggest criticisms of AI chatbots is that they can sometimes make mistakes. This is because they are trained on large datasets of text and code.
It’s not so much that the datasets are filled with a majority of incorrect information, rather that the data sets being so large naturally contain small amounts of bad information, which can impact the content the bots produce.
Several journalism outlets have used AI to great effect to replace human data-gatherers, including The Washington Post, Reuters and the Associated Press.
However there have been many high-profile examples of outlets using bots to generate articles filled with blatant misinformation, the most high-profile of these being the CNET ‘money team’ running a slew of badly-written, badly-informed AI-generated articles on personal finance.
There are other more ridiculous and downright dangerous examples, like this New Zealand-based supermarket using chatbots to generate recipes such as a chlorine gas cocktail and bleach-infused rice.
AI technology does without doubt have its place in the future of data collection for sectors like journalism and the hope is that as more money is invested in the technology, chatbots will be able to be trained on more accurate and comprehensive datasets, which will lead to fewer mistakes.
Be more secure
AI chatbots collect a lot of data about users, including their personal information, contact details, and purchase history. This data needs to be protected from unauthorised access, and chatbots need to be designed with security in mind. As AI chatbots become more popular, it is important to ensure that they are secure and that user data is protected.
By constantly evolving and improving, AI chatbots can stay relevant and continue to be a valuable tool for businesses and users alike.
All of these teething problems and issues mean that AI chatbots have struggled to command the investment they once initially expected, though relatively speaking they are still attracting interest and capital.
In March, Character.AI, a two-year-old start-up that builds online chatbots, declared that they had raised $150 million in a 2023 funding round that valued the company at $1 billion.
There is a small conglomerate of AI chatbot builders who are working to disrupt and improve the world being built by Microsoft’s ChatGPT and Google’s Bard.
Enhanced security is after all one of the most highly-touted features of the ‘premium’ version of OpenAI’s ChatGPT.
The $20-a-month version was rolled out for customers to trial in May and seems to be a hit with full-time techies and hobbyists combined; it offers the services of a virtual web assistant who can direct your internet journey through voice or typed command.
The joy of watching this technology develop is that it makes the simplistic call and respond structure to Siri and Alexa seem primitive.
Think more, ‘What are the most popular drinks at a nearby bar based on sales?’ or ‘based on the weather forecast, which weekend should I visit Brighton this summer?’
It is the continuous development of this side of technology that will continue to keep chatbots a relevant and exciting new technology for the general public, outside of the thousand possible use cases they will continue to have in the professional world.
