Now that you have developed an understanding for what goes into creating an AI model it is time to look into the current landscape of existing products and how they can most effectively be used. It is no secret that using AI can supercharge your productivity so in this chapter we will discuss the best uses for LLMs and cover in which situations AI may be counterproductive. LLMs like ChatGPT can be useful for many situations, but since this course is for young entrepreneurs we will focus on the most business oriented uses.
Benefits
- Fast generation of related topics
- Familiar with many concepts and good at linking related points
- Can bounce ideas off you, similar to how you would talk to a person
Limitations
- Sometimes does not understand what you are expecting so you need to be more specific
- Only trained on available data so if it is something proprietary or outside of the dataset you are out of luck.
Benefits
- Usually writes more coherently than most humans, especially in their 2nd language
- Familiar with many concepts and good at linking related points
- Can bounce ideas off you, similar to how you would talk to a person
Limitations
- Sometimes the tone is not relevant for the type of text
- If you are working with someone that does not accept AI content they can use an AI scanner.
- Might take longer to explain the context and expectation than writing it yourself
Benefits
- Spots mistakes that you may overlook
- Faster than you reading your own work, especially for larger texts
- Can suggest improvements once critiques have been identified
Limitations
- Programmed to give a response no matter how forced, resulting in sometimes trivial feedback
- Sometimes much more critical than necessary for the task at hand
- Often uses more words than necessary for the message it is trying to convey unless you instruct it not to
Benefits
- Fast code generation
- Familiar with many languages and frameworks
- Can fix its own code if it doesn’t work first try
- Can explain the code line by line
Limitations
- Unable to generate file systems and complex integrations
- May waste time by generating code that you do not understand
- Sometimes hallucinates false facts
- Limited to training data that has been provided
- Only knows the contextual information that you provide
LLMs work on the basis of predicting the next word that should be generated. While newer models have more advanced systems for citation and providing information, you still want to double check important facts yourself as mistakes can happen.
If the topics that you are trying to get information about or discuss with the AI are not prominent in its training data it will not do a good job in any prompts related to these topics.
If you are, for example, trying to write a text for the website of your company then you need to provide the AI with a whole lot of information about your vision, operations, and so on. If this information is not already in accessible text that you can copy paste, it may take longer for you to write it up than writing the website copy yourself. This is especially true when you need a short text with very precise meaning. In those cases it may be effective to write the initial text and then ask the AI if it sees any room for improvements.
If the topics that you are trying to get information about or discuss with the AI are not prominent in its training data it will not do a good job in any prompts related to these topics.
If you are, for example, trying to write a text for the website of your company then you need to provide the AI with a whole lot of information about your vision, operations, and so on. If this information is not already in accessible text that you can copy paste, it may take longer for you to write it up than writing the website copy yourself. This is especially true when you need a short text with very precise meaning. In those cases it may be effective to write the initial text and then ask the AI if it sees any room for improvements.