ChatGPT has changed the world. Not necessarily because it’s new technology or something that the wide world of AI has not seen before, but because OpenAI has done such a stellar job of marketing their product that it has become the default term for artificial intelligence.
When the company was founded in 2015 – with backing from a consortium that included Elon Musk, Sam Altman, and Amazon Web Services – the goal was to be a refuge for human-friendly AI, staffed by the world’s leading machine learning minds.
With the release of ChatGPT (and the GPT-4 language model that followed), the OpenAI team delivered on that promise, and humans are finding new ways to use ChatGPT every single day. And it deserves every inch of the coverage it’s getting because it has moved AI out of the abstract, into an understandable, achievable, and relatable realm.
People are excited about AI in a way that they couldn’t be before. This is an inflection point that is both scary and invigorating. It’s the starting point of a gold rush that will change the way we interact with technology.
ChatGPT has made it easier to make a case for AI integration, especially with clients who didn’t necessarily understand what natural language processing, large language models, and computer vision were. Now, companies with AI products can work with clients to move humans, employees and customers into a world where simple tasks can be automated in order for them to focus on higher order tasks. There is the concern that we will lose customer service jobs in the process, but we will still need people in place when AI is unable to assist, or in the many, many instances in which customers need the human touch.
Automation can therefore unlock more complex tasks which will now move into the human domain. This will, of course, mean that the end user will have greater expectations for every brand’s customer experience – the same way they felt when Uber revolutionised transport, AirBnB transformed travel accommodation, and Takealot changed the face of online shopping in South Africa.
Businesses will need to evolve or risk losing customers, who are now quickly becoming accustomed to experiences that are far richer and more personal.
This chaotic moment we’re in right now will reshape how we do business, but on the upside, there will be more convenience and more opportunity for innovation. One of those opportunities blends the human touch and technical knowledge of natural language models, and how to apply it to specialised domains.
There are three ways organisations can maximise the impact of AI in their business:
UNDERSTAND THAT ONE SIZE DOES NOT FIT ALL
Helm, and companies like us, don’t see Large Language Models like GPT-4 as a threat because API, Natural Language Understanding, and Machine Learning integration are large parts of our business. Yes, we have Helm Engine that underpins our technologies, but we have also helped many clients with custom builds that utilise language models, open-source tools (where relevant), and experience design thinking to build out these next generation customer experiences.
Integrating models like GPT-4 and Stable Diffusion is not as simple as just integrating an API that’s been made available. It requires an understanding of the experience one is looking to create, designing it, leveraging data sources available, integrating into back-end systems, building out language models for the specific use case, training the models constantly, guarding against made-up facts aka “hallucinations” – and so on.
The experience we have with ChatGPT now has taken many years to develop, an insane amount of humans to curate and train the models, and massive amounts of cash to make it work. Whilst the core can be leveraged, making it work in an organisation requires similar effort – the scale will differ though.
INVEST IN SKILLS AND ABILITIES AT HOME
The easiest way to bring down the cost and increase the local relevance of AI is to invest in the human capital within our borders. Helm Engine was built from scratch using homegrown talent to solve African problems.
Our work optimising chat systems to help automate mundane tasks – like helping customers diagnose error codes on their decoders – has built a foundation of customer interaction that we can feed into more complex machine learning models.
This has helped us, and our clients, keep our costs down and enhance our domain specificity.
KNOW WHAT GOOD LOOKS LIKE
In the early days of the internet (and even now in some cases) everyone had a website but not everyone had a good website. ChatGPT and even the Bing integration of GPT is popularising the text box, but it can be clunky – and even dangerous – if taken out of those specific contexts.
The biggest challenge in the industry now is helping people understand what good AI integration looks like, what good customer experience looks like, or what a good ‘chatbot’ should be able to do. There is a lot of AI being shoved into services for the sake of it, with little consideration for the user experience.
As an analogy, the iPhone wasn’t the first smartphone, but it was the first one that felt like it considered the user interaction first, before function.
ChatGPT has made AI accessible to everyone and is inspiring ideas of unique integrations. It isn’t the only solution on the market, and all AI boats – ours included – will rise on this tide of excitement.
This did not happen overnight, and success has been a long time coming. The Helm team is covered in scars and bruises from many years of trying, testing, breaking, and recreating automated solutions. But here we are on the other side, ready to ride the wave of the AI revolution.
Article Written by Dawood Patel, CEO of Helm