Future of the Knowledge Economy: GPT Based Businesses

It's no secret that the release of Chat-GPT in November of 2022 was met with widespread enthusiasm. Within just five days of its release, the chatbot had garnered over one million users, a testament to its captivating abilities and the allure of large language models.

But what does this mean for the future of the knowledge economy? As discussion continues to swirl around the true openness of large language models like Chat-GPT, it's worth considering how these technologies will shape and reshape the way we do business.

I wanted to build some projects to get first hand experience with the types of applications that can be built on top of the GPT engine. So I started with learning the API.

The following code snippet shows how you can send a message to the API and receive a response using the fetch function in a next.js API route:

import { Configuration, OpenAIApi } from 'openai';
import dotenv from 'dotenv';

export default async function handler(req, res) {
  const { message, chatHistory } = req.body;

const configuration = new Configuration({
apiKey : process.env.OPENAI_API_KEY,
organizationId : process.env.ORG_KEY,
});

const openai = new OpenAIApi(configuration);
try {
const response = await openai.createCompletion({
    model: "text-davinci-003",
    prompt: {message},
    max_tokens: 3000,
    temperature: 0.7,
    presence_penalty: 0.5,
    frequency_penalty: 0.5,
    
  });

res.json({ 
    message: response.data.choices[0].text,
    status: response.status,
 });

Given the models extensive knowledge in a diverse set of domains. I decided to test it's knowledge on personality typing (specifically MBTI). To my amusement the model did a great job at understanding the differences between the different types and personalizing the conversation if you provide the right prompt.

Even though the scientific validity of MBTI is up for debate I thought this would be an interesting use case where people could receive customized advice that is based on their personality type. I created a beta project called AI Personality Coach. Eventually I believe that combining it's ability to understand personality with tracking goals and keeping long term memory and summarized context stored in a database could be incredibly valuable. This could be a way to receive effective business and personal coaching for the fraction of the price of a real coach.

The other use case that came to mind was customer service. With my project Grapevyyn we send automated review requests once customers complete their visit at a business. One of the features is that we try to gauge their enthusiasm before asking them to leave a review. This process is done with hard coding, however utilizing GPT's ability for human like communication this could be taken to the next level.

I think integrating that as the highest tier of the pricing model could be a valuable service. This is just the beginning of the potential for customer service application and I know other companies such as Intercom which use the GPT engine to generate helpful responses to customer support tickets.

One concern that I have seen brought up multiple times is about the true "openness" of the GPT-3 models. Although OpenAi does make the API available it is still not the same degree of openenss as releasing the model so people could run it on their own machines or in the cloud - assuming of course you have the compute power and the estimated 600 GB of VRAM to run the inferences. I think it is important to distribute this technology equitably and that is why I believe organizations like EleutherAI and Hugging Face are playing an important role in making these LLMs open source.

I believe the knowledge economy will continue to change at a rapid rate and the economic implications will be exciting to experience.

"My model for the next decade is that the marginal cost of intelligence and energy will rapidly trend to zero." -- Sam Altman

Ultimately, it's hard to say exactly which business models with be able to thrive the most (other than the base layers themselves like GPT and Palm). However I believe the change is hard to underestimate and businesses will need to adapt to the new environment.