What skills do I need to use ChatGPT?
Steven Forth is co-founder and managing partner at Ibbaka. See his skill profile here.
ChatGPT from Open.ai has attracted a lot of attention since its introduction on November 30. If you have not had a chance to try it, take some time to do so.
ChatGTP is an example of an artificial intelligence that people are going to need to work with in the future. Increasingly our teams will include AIs and we will need to get better at working with them. I am collaborating with one of my granddaughters on a novel, The Enchanted Laboratory, and we have started to use ChatGPT as a way to generate some of the dialog and to get a different perspective. We then edit the resulting text. This sort of collaboration with AIs is going to become common across many business functions. ChatGPT is already being used in the real world for high value tasks like writing content, debugging and even generating code.
Given this, what skills are needed to work effectively with ChatGPT and similar tools?
TL:DR A simple role and skill model for using ChatGPT
Working with AIs like ChatGPT will become a daily activity for most knowledge workers. Here are some of the skills we will need to cultivate.
Structured Questioning - skill in asking a sequence of questions that guide a conversation or enquiry, open up new directions for questioning, while testing assumptions and verifying assertions
Generative Thinking - skill in generating new ideas and solutions from existing ideas and solutions
Critical Thinking - skill in evaluating ideas and solutions, testing assumptions and verifying assertions
Curiosity - a behavior or character attribute rather than a skill, interest in exploring new ideas and domains an openness to learning new things
Domain Knowledge - an understanding of the domain on enquiry is critical for directing questions and evaluating responses
For those of us who work directly with the models there are some additional skills needed.
Natural Language Processing - a set of technical skills and a knowledge base on how language (syntax and semantics) can be represented and manipulated as data structures
Probability Theory - the foundation for working in AI and Deep Learning
GPT or Generative Pre-Trained Transformer (currently 3.5, 4 coming soon) - the Open.ai language model that ChatGPT is built on, Open.ai provides APIs for working with this model
TensorFlow, SageMaker, AutoML, Pytorch and other languages used to build machine learning models
Keras - a high-level, deep learning API developed by Google for implementing neural networks. It is written in Python and is used to make the implementation of neural networks easy. It also supports multiple backend neural network computations
Data Analysis Language (Python, R, Julia) - these languages are powerful ways to define and manipulate the data structures used in AI
Deep Learning - the approach to artificial intelligence getting the most traction, uses multilayer neural networks (hence ‘deep’)
Generative Adversarial Networks - a way to structure two deep learning models so that they compete with each other, for example one trying to categorize a set of images, the other trying to generate a set of images that will fool the first model
A conversation with ChatGPT about the skills needed to work with ChatGPT
Let’s start by asking ChatGPT what skills are needed.
Who can make best use of ChatGPT?
ChatGPT is a chatbot developed by OpenAI that uses a large language model trained on a vast amount of text data to generate human-like responses to user input. It can be used by anyone who wants to have a conversation with a chatbot, whether for entertainment, information, or to practice language skills. For example, someone who is learning a new language could use ChatGPT to help them practice their conversational skills, or someone who is looking for a fun way to pass the time could use it to have a conversation with a virtual companion. It can also be used by businesses or organizations as a customer service or support tool.
What skills are needed to use ChatGPT effectively?
To use ChatGPT effectively, you don't need any particular skills or knowledge. All you need is the ability to type and read, and an interest in having a conversation with a chatbot. ChatGPT is designed to be easy to use and intuitive, so you can start using it right away without any prior experience.
Does knowledge of NLP or programming help one use ChatGTP more effectively?
While knowledge of natural language processing (NLP) or programming may help you understand how ChatGPT works and how it generates its responses, it is not necessary to use the chatbot effectively. ChatGPT is designed to be used by anyone, regardless of their background or expertise. You do not need to have any knowledge of NLP or programming to have a conversation with ChatGPT and get value from it.
What skills will help me evaluate ChatGTP responses?
To evaluate ChatGPT's responses effectively, you should pay attention to how well the chatbot understands your input and whether its responses are relevant and coherent. You can also evaluate the quality of the chatbot's responses by comparing them to what a human might say in the same situation. For example, if you ask ChatGPT a question and its response is not related to the question or is difficult to understand, it may not be a good response.
Does skill in asking questions help me use ChatGTP?
Skill in asking questions can help you get better responses from ChatGPT, as it will enable you to ask the chatbot more focused and specific questions. This can help the chatbot understand your intentions better and generate more relevant and accurate responses. For example, instead of asking a general question like "What do you think about the weather?" you could ask a more specific question like "What is the current temperature in New York City?" This will give ChatGPT a clearer idea of what you are looking for and help it provide a more useful response.
How do I get better at asking questions of ChatGPT?
To get better at asking questions of ChatGPT, you can practice framing your questions in a way that is clear and specific. This will help the chatbot understand your intentions better and generate more relevant and accurate responses. Here are some tips for framing effective questions:
Start with a clear and specific topic or question. For example, instead of asking "What do you think about the world?" you could ask "What are the biggest challenges facing the world today?"
Use appropriate vocabulary and grammar. Avoid using slang or colloquial language, and try to use proper sentence structure and punctuation.
Be concise and to the point. Avoid asking long, rambling questions that are difficult to understand or answer.
Be open-minded and curious. Ask questions that you are genuinely interested in, and be willing to explore different perspectives and ideas.
Is curiosity important when using ChatGPT?
Yes, curiosity is an important quality to have when using ChatGPT. ChatGPT is a chatbot that is designed to engage in open-ended conversations, and it can generate a wide range of responses depending on the input it receives. By being curious and open-minded, you can encourage ChatGPT to explore new ideas and topics, and you may be surprised by the responses it generates. This can make your conversations with the chatbot more interesting and engaging. Additionally, being curious can also help you learn more about different topics and broaden your knowledge.
When using ChatGPT, should I begin with a general question and then narrow the focus or begin with a specific question and then explore different aspects?
When using ChatGPT, it can be helpful to begin with a general question and then narrow the focus as the conversation progresses. This allows the chatbot to generate a range of responses and gives you the opportunity to explore different aspects of the topic. For example, you could start by asking a general question like "What do you think about the future of technology?" and then follow up with more specific questions like "What do you think will be the biggest breakthroughs in technology over the next decade?" or "How do you think technology will change the way we live and work?"
How much domain knowledge do I need to use ChatGPT effectively?
You do not need to have any particular domain knowledge to use ChatGPT effectively. ChatGPT is a chatbot that uses a large language model trained on a vast amount of text data, so it can generate responses on a wide range of topics. You can have a conversation with ChatGPT about any topic you are interested in, and the chatbot will do its best to provide relevant and coherent responses. However, if you have some knowledge or expertise on a particular topic, you may be able to have more in-depth and informative conversations with ChatGPT about that topic.
How does ChatGPT expand its language model? Is this happening all the time or are their periodic updates?
I'm sorry, but I am not able to browse the internet to provide an answer to your question. I am a large language model trained by OpenAI, and I can provide information and answer questions based on my training, but I don't have access to real-time information or the ability to browse the internet. My knowledge is based on the data that was used to train me, and my responses are generated based on that knowledge.
How often is the language model updated?
As a large language model trained by OpenAI, I am not updated or changed after I am trained. I am a static model that can generate responses based on the data that was used to train me, but I do not have the ability to learn or adapt over time. My responses are generated based on the knowledge and information I have been trained on, and I cannot update or expand that knowledge.
The last two questions are not really concerned with skills of course, but it is interesting to know how ChatGPT describes itself.
What skills are really needed to use ChatGPT?
I took the above text and ran it through Ibbaka Talio’s skill extraction function. The following skills were discovered.
One needs to differentiate between using ChatGPT and using the underlying GPT model. For users here is our understanding of the key skills. This model will be developed further as new use cases for ChatGPT emerge.
Key Skills for Using ChatGPT
Structured Questioning - skill in asking a sequence of questions that guide a conversation or enquiry, open up new directions for questioning, while testing assumptions and verifying assertions
Generative Thinking - skill in generating new ideas and solutions from existing ideas and solutions
Critical Thinking - skill in evaluating ideas and solutions, testing assumptions and verifying assertions
Curiosity - a behavior or character attribute rather than a skill, interest in exploring new ideas and domains an openness to learning new things
Domain Knowledge - an understanding of the domain on enquiry is critical for directing questions and evaluating responses
Key Skills for Working with the GPT Model
For those of us who work directly with the models there are some additional skills needed.
Natural Language Processing - a set of technical skills and a knowledge base on how language (syntax and semantics) can be represented and manipulated as data structures
Probability Theory - the foundation for working in AI and Deep Learning
GPT or Generative Pre-Trained Transformer (currently 3.5, 4 coming soon) - the Open.ai language model that ChatGPT is built on, Open.ai provides APIs for working with this model
TensorFlow, SageMaker, AutoML, Pytorch and other languages used to build machine learning models
Keras - a high-level, deep learning API developed by Google for implementing neural networks. It is written in Python and is used to make the implementation of neural networks easy. It also supports multiple backend neural network computation.
Data Analysis Language (Python, R, Julia) - these languages are powerful ways to define and manipulate the data structures used in AI
Deep Learning - the approach to artificial intelligence getting the most traction, uses multilayer neural networks (hence ‘deep’)
Generative Adversarial Networks - a way to structure two deep learning models so that they compete with each other, for example one trying to categorize a set of images, the other trying to generate a set of images that will fool the first model
Emerging Use Cases for ChatGPT and Similar Technologies
New ways to use ChatGPT are emerging everyday. These different use cases will drive the skill requirements. Open.ai released ChatGPT in part to discover these use cases and make sure that GPT 4 can support these many different uses. You can read more about GPT 4 here.
It helps to have a way to organize the different use cases. At Ibbaka, we are exploring the following framework.
There are four modes for using ChatGPT and other similar AIs.
Exploration - Explore ideas, find new ideas, build the graph of how ideas are connected. This post is an example of this.
Generation - ChatGPT can be used to create new things - it can generate code, even develop whole new languages. Its sister application, Dall-E generates images. What will happen when ChatGPT and Dall-E are cross pollinated?
Validation - As has been demonstrated, in certain fields ChatGPT is able to debug and correct errors. Not a coder? Try using ChatGPT to correct grammar or simplify statements.
Exposition - ChatGPT can generate pretty good expository prose and in a variety of styles. Some people worry that this will make writing ourselves redundant. I doubt that. But anything that reduces rote work is a good thing.
These modes cross with different applications.
Learning - One can learn a great deal about many topics by exploring ChatGPT and asking questions. This is a far better learning experience than reading or listening to canned material. It supports exploratory and conversational learning.
Documentation - ChatGPT is a great complement to Wikipedia in documenting what we as a society know. It does not really generate new knowledge. And how it will handle error correction is not clear (there is a growing body of work documenting errors in ChatGPT). But this is part of the store of knowledge that all of us can now draw on.
Persuasion - Can ChatGPT be used to generate content meant to persuade rather than explain or explore? Yes. That is already happening. I know several copywriters that are using ChatGPT, as
A brave new world
This is a ‘Brave New World’ in both Shakespeare’s and Huxley’s use of the phrase.
“Oh, wonder!
How many goodly creatures are there here!
How beauteous mankind is!
O brave new world, that has such people in ‘t!”
The Tempest is a story of forgiveness, reconciliation and a recognition of our limitations.
Aldous Huxley’s Brave New World is a dystopian story of social engineering and alienation. Which direction will AIs push us in?