ChatGPT is an advanced language model developed by OpenAI. It is based on the GPT-3.5 architecture, which stands for “Generative Pre-trained Transformer 3.5”, one of the most powerful versions of the GPT series.
Now renowned globally, ChatGPT is designed to engage in naturally flowing conversations with humans, providing responses and generating text based on the context of the conversation. It has been trained on a diverse range of internet text, allowing it to understand and generate human-like responses across various topics.
You could ask it to do things by simply typing them out, like drafting an email, writing blogs, creating social posts, or anything that involves text. This is because the ChatGPT model is pre-trained with a lot of data, which includes books, articles, websites, and other sources of text from the internet which lets it generate coherent and contextually relevant responses.
How Do LLMs Work
ChatGPT is a “Large Language Model” which is frequently abbreviated to LLM, other examples of LLMs are Google’s Bard, and Meta’s LLM called LLaMA.
It is important to understand that LLMs are trained on language patterns that they reproduce later on. Quite similar to the autocomplete in your smartphone. They do not have any concept of “learning” or “understanding”, they’re simply trained on large amounts of language data which they later reproduce.
Thus it is always good practice to double check any information that has been provided by an LLM, since they simply follow language patterns they are prone to making up information.
Now let us understand how LLMs are trained.
Pre-training: During pre-training, an LLM is exposed to text data from the internet. It learns to predict the next word in a sentence by considering the context of the preceding words. This process helps the model to capture patterns, grammar, and sentence formation.
Fine-tuning: After pre-training, the LLM undergoes a process called fine-tuning where the model is trained on a more specific dataset with the help of human reviewers. The reviewers follow guidelines and provide feedback on responses. This feedback loop helps the LLM to learn and improve its responses over time and hence keeps up-to-date on the day-to-day lingo as well.
ChatGPT for businesses:
Now that we know how it works, let’s look at how it can help businesses:
- Blogs, SEO and write ups: Create all of your written content from ChatGPT! Just let it know what information you’d like to include and the format you’d like. For example, create a blog on the Salesforce emailing app. You can also ask it to add specific keywords for SEO purposes, but you’d have to mention these details.
- Summarise content: ChatGPT is extremely helpful in summarising large quantities of text. It can analyse contents of a large body of text, and generate a condensed version of it, while ensuring the main information of text is preserved. You can also ask follow-up questions to clarify any doubts you may have after reading the text summary. This is extremely useful for prospecting new customers or researching new technologies.
- Social media: Ask ChatGPT to write tweets, LinkedIn posts, or Instagram captions. Just let ChatGTP know what your strategy is, what information is required, word limits if any. Pro tip: You can ask for 10-15 options at once and choose the best one.
- Sales and client content: It can help you come up with persuasive content which gets clicks and attention or can help your sales team hit the nail on the head. You can ask ChatGPT to create a sales cold email or ask it for topics to gain more clients.
- Customer Service via Chatbots – Additionally ChatGPT can make chatbots more human-like and effective, increasing the quality of customer service, and allowing your customer service reps to get through a higher number of customers than they’d ordinarily be able to.
- Correcting grammar and typos: If you’re not a native English’s speaker, ChatGPT can act as a great tool for improving the quality of your communications, be it emails or messages. Making sure all communication is free of typos and uses the correct grammar, while ensuring your communication is cohesive and coherent.
EinsteinGPT – A collaboration between Salesforce and OpenAI
If all the above use cases and information are overwhelming for you, there is good news!
Salesforce has announced a collaboration with OpenAI, the creators of ChatGPT to launch a product for their Salesforce CRM platform called Einstein GPT. While the Einstein AI is more focused on predictive analysis, Einstein GPT would be focused on generating natural language content based on said analysis. Example:
- EinsteinGPT would combine public and private AI models with your CRM data, so that users can use natural language prompts to answer any query they may have, directly within the Salesforce CRM.
- EinsteinGPT will be trained on Salesforce’s private AI models, i.e. those developed by Salesforce, which users presently have access to from the Einstein technology layer. In addition to this you can also bring in your own models. Since Einstein GPT is open and extensible, you can even finetune it specifically for the industry and sector your business operates in!
Currently Einstein GPT is in closed pilot, with the expected release data to be in the summer of 2023.
But based on announcements made by Salesforce here are the modules that will be available under EinsteinGPT:
- EinsteinGPT for Service: to automate tasks manually done by a service agent. EinsteinGPT will be able to auto generate personalised chat replies, summarise knowledge articles from past case notes and auto generate summaries during case wrap up
- EinsteinGPT for Sales – to auto generate composing emails, scheduling meetings and moving the client along the sales funnel. The sales module will also be able to surface external news for prospecting research along with generating additional collaborating channels.
- EinsteinGPT for Marketing – will be able to dynamically generate personalised content to engage existing and prospective customers across email, mobile, web and advertising.
We hope we have demystified ChatGPT for the reader, and highlighted different ways with which it can supercharge operations for Salesforce users. At BIGWORKS we look forward to this new era of AI with anticipation and optimism.