Taking the Chat out of ChatGPT
A first impression is stickier than a second.
It is no surprise then, that the unprecedented success of ChatGPT has put the focus on chatbots as THE application of generative AI.
When electricity was first brought to the public by Thomas Edison, it was mainly used to power light bulbs. Lighting was seen as the primary application of electricity.
Lighting, of course, only scratches the surface of what electricity could do. The same applies to chatbots and generative AI.
Look beyond chatbots and generative AI gets a lot bigger. There is a lot more opportunity to create value with AI for your business and your customers than most people realize.
General-Purpose versus Specific-Purpose
To understand what we can do with generative AI, it's helpful to categorize applications into two broad groups: general-purpose and specific-purpose applications.
General-Purpose
General-purpose AI applications such as ChatGPT are designed to handle a wide range of tasks. While they are useful for many things, they often lack depth in specific domains.
General-purpose applications often come in the form of chatbots but it is crucial to remember that this is just one possible user interface, not the AI itself.
While general-purpose AI applications are useful, their true promise won't be reached until we achieve AGI. Many applications going down this route are doomed to stay "cool" demos, rather than useful applications.
Specific-Purpose
Specific-purpose AI applications are trained to be really good at certain tasks. They often work with existing systems and get a lot of context about their domain.
We don't know when or if we'll have AI that can do everything one day. So, for now, these specific-purpose applications are the most useful.
We are incredibly excited about the potential of specific-purpose applications at Pretrain.
Right now, the most common use is a chatbot made for a specific company or industry. It uses a knowledge base and some rules to answer questions and do tasks in that area.
But there's more than just chatbots. At Pretrain, we call these other uses "workflows".
Workflows
Workflows are about making an AI really good at one specific job. There's no chatting here. It's just about taking one or more inputs and transforming them into a different kind of output in a set way.
There is some irony in that workflows are probably the most useful application of generative AI today. Workflows only work when you limit a general-purpose AI enough to be good at a specific task.
Some exciting workflows include:
- • Product description writing for e-commerce
- • Email response suggestions
- • Social media post creation
- • Code snippet generation
- • Meeting summarization
- • Personalized marketing copy
- • SEO-optimized blog post outlines
- • Legal document drafting assistance
- • Technical documentation generation
Workflows help automate tasks that are repetitive and time-consuming.
For example, an AI trained to create social media posts could take a topic (input) and write a 280-character tweet (output) that fits what the user wants.
Making great workflows isn't easy. You need to decide what makes a good output. What makes a tweet good? It's tricky, and it might be different for each user.
And that's why we built Pretrain.
There are many useful workflows waiting to be built. But teaching an AI to do a workflow well, and to make it work for each user, is hard.
So, if you want to build a workflow, but don't know where to start, we'd love to help.
Schedule a demo. Let's use AI to make your business even better.
