Generative AI Trends

All the latest news and trends
What is Generative AI

shared by Naveen Kohli on Friday, July 7, 2023 •

Introduction

Generative AI is something that everybody talks about these days. Everybody is curious about it. Some think of it as a magic lamp. You rub it and a genie is going to appear to grant you wishes. Some think it is some witchcraft that if you use it, it is going to bring misfortune at some point. I think a lot of it has to do with the fact that the technical community responsible for creating this engine has not been able to do a good job describing how it works and what is the purpose of it. There are a lot of misconceptions on either side of the aisle. A lot of it seems to have do with the word “Artificial”. This word translates to “So, it is not human”.

At the end of the day, it is the fact. Generative AI is a machine or engine. But there is one more part of this sentence that a lot of time is forgotten that it is created by humans.

So, what is it?

I am a big fan of AI and automation. I always say, what can be automated must be automated. For me this translates to using a machine to perform the repetitive manual tasks with some intelligence. This does not mean that the machine has some intelligence of its own. No, it does not. A human has trained the machine with all the rules it knows about. A human has provided additional information to the machine to take decisions based on the state and conditions it encounters. The machine must be told about those conditions as well. If the machine is not trained about certain conditions or states, it is not going to be able to make right decisions when it encounters those conditions or states.

When I think of what Generative AI is, it reminds me of the movie History of The World Part 1. There is a conversation between Comicus and an agent at the employment office. Comicus is asked about his profession. I can imagine the same conversation with AI engine like ChatGPT.

B.Arthur: “Occupation?”
ChatGPT: “Generative AI Engine”
B.Arthur: “What?”
ChatGPT: “A Generative AI Engine; I coalesce the vapor of human experience into a viable and logical comprehension.”

Generative AI Engine is a very well-trained machine that has been taught a vast set of topics with a vast knowledge base of human experiences in those topics. It has been trained to understand the questions so it can synthesize the best possible answers from the knowledge it possesses.

As you can see from my definition, AI is based on past human experiences. It does not know anything more than what it has been taught. If you ask it a question about any new topic, it will not be able to answer it. Think of Generative AI as a student that is being trained everyday by humans. If a student fails to answer some questions, the teacher will ask the student to go read the topic and learn it. The technology teams behind these AI engines act as students who perform the same task of learning. If they encounter an incident where the AI engine failed to answer a question or provided wrong answers, they feed more data about the topic to the AI engine.

As humans, we are not able to answer each question. We acquire more knowledge every day and try to achieve perfection. Generative AI engines go through the same process as well.

How do I use it?

There is no better way to form an opinion about anything than having a first-hand experience about it. I have come across people who have made statements like below.

  • “So, it is another search engine.”
  • “What, is that it will tell me different than when I google it.”
  • “It knows what I think?”

I saw it happening in my own home. I showed DALE-E to my kid. At first, she was not sure what she was looking at. Then there was some skepticism. She is into cosplay and art related to avatar creation. I showed her how to generate an avatar based on her current theme and ideas. We created an avatar. She did not like it and said, “it is not what I had in my mind”. I asked her “did you tell it what is in your mind”. Her response was not surprising. She said it is artificial intelligence, it should have figured it out.

So, I said, let’s try one more time. This time describe everything that is in your mind including the smallest detail. Voila, we created an avatar that was close to what she had in mind. Next thing that she said was very interesting. “Now I have something I can work with to create the actual avatar for my character”. Why this statement is critical, I will talk about it later.

To harness the full power of Generative AI engine, it is important to understand how to use it. It is like using any machine. If you do not learn it, you are more likely to generate output that is of bad quality. That will lead to blaming the machine.

In the next few sections, I will list some of the key elements of how best to use the Generative AI engine.

Not a Search Engine

I personally think of tools like ChatGPT as conversation engines. You have some ideas, and you want to run it past by some person who has knowledge about it. The other person provides some answers. You look at the answers. You have some follow-up questions or want to tell the person that some of the answers are not relevant. The other person listens to you and provides additional information or removes the irrelevant information from the conversation. Now you have some answers that look viable, and you want the other person to elaborate on those answers. You carry on this conversation for some time until you reach the point where you have sufficient information to start working on materializing your ideas.

Sometime your conversation may not materialize because the other person does not possess the knowledge in the subject matter. Well, tools like ChatGPT have the same problem as well. As I mentioned earlier, it only knows about what it has been taught.

Good Questions, Good Answers

I use a phrase, Garbage in Garbage out. It simply means that if you do not ask good questions, you are likely to get bad answers. There are two parts to the training of AI engines. One part is to train the engine to decompose the input (questions) into the pathways that it can travel to reach the correct destination to extract the information. Think of it as asking GPS for directions. The more precise your description of the destination, the more precise the directions are going to be. If you leave any gaps in your input, the machine will try to fill those gaps with the best possible guess and assumptions. That leads to another phrase I use, if you assumed something, you assumed it wrong.

In the Generative AI world, it is called Prompting. A new skill set called Prompt Engineering has started based on this term. You must prompt the engine with good context and provide as much information as possible. Lack of good prompts will lead to inaccurate or incomplete information. Some time the engine may completely fail to provide any answer.

I experimented with it with my 9 year old kid. I asked him to prompt ChatGPT. All he said was “pet elf”. ChatGPT completely failed on this prompt. It did not provide any answer. It apologized that it does not have any information on this topic as of September 2021. He was very surprised that ChatGPT did not provide any information. I realized what was wrong. I was having a conversation with ChatGPT about a broken washing machine. Suddenly, I prompted it with something that was so out of context. It tried to find the answers within the earlier context. I am not sure if there is any known washing machine that has “pet elf” theme or Santa is planning to set up some washing machine factory.

I started a new conversation and prompted with same phrase “pet elf”. This time it gave information about the related toys etc.

It is all about good prompts and questions. Do not treat conversation Generative AI engine anything differently than having a conversation with any other person. There is a tip that I can share about good prompts. Use small but meaningful sentences. It helps Generative AI engine in decomposing the input in an efficient manner. I am not saying this because I know how ChatGPT works. I am telling you based on my personal experience. At times it may feel that it is naïve to write a long question in the form of small and simple questions. Again, think of this in the context of a conversation with a real person. If you keep your questions short and simple, it takes less time for the person to understand the actual intent behind the questions. This leads the person to provide a good answer.

Practice, practice and practice. There is no other way to figure out what will be a good prompt. One solution does not fit all. What is a good prompt in one subject matter, may not work very well in another subject matter. Do not be discouraged by irrelevant answers in the early days. Do not form your opinion with a few bad answers. Fewer data points are not indicative of the actual trends.

Creativity and Human Factor

Earlier I gave an example of my conversation with my kid about generating avatars. I am going to repeat that sentence she used.

“Now I have something I can work with to create the actual avatar for my character”.

The key here is that Generative AI engine provided ideas that she could work with. It is a machine. It can process information at a faster speed and produce output at a faster speed. If a human must generate a lot of ideas for an artwork, it can take a very long time. Some of the ideas you keep and some you throw away. Out of the ideas we keep, then we take the best parts of the ideas and produce the final master work that we are happy with.

I use Generative AI sometime to create test cases in software development. It is not that I do not know how to generate test cases. I use ChatGPT to quickly assemble the outline of the areas that are possible candidates for functional and security testing. If I do this work manually, I will reach the same answers as well. It may take some iterations. I may come up with 5 items and then a few hours later while the testing is taking place, we will realize that we forgot to include 3 other items. Since AI engine has been trained with some common use cases, it synthesized all that information for me without having the actual context. There will a good chance that some of the items are not relevant for my product and I will remove those items from the idea list.

Creativity is a human trait. In using Generative AI, you are using your creativity at multiple levels. You must be creative with prompts to generate content that is relevant to your subject matter. You must be creative to pick and choose the content that is genuine and trustworthy. You must be creative to synthesize the ideas into a work that reflective of quality of your and your organization’s standards.

Ethics and Legality

Generative AI has opened pandora’s box that organizations were not fully prepared for. It is raising lot of concerns around topics like:

  • Intellectual Property
  • Ownership
  • Plagiarism
  • Disclosure
  • Transparency
  • Authenticity of Information
  • Quality of Work

Over the next few years there will be a lot of legislation about these topics around the world. There will be a lot of legal opinions and legal actions on how people used and how they should use the information generated by these engines.

There are a few facts that I use for my personal use of Generative AI.

  1. I did not create the content myself. I cannot claim ownership of the material. I cannot claim any type of Copyright on the content.
  2. I did not create the content myself. I cannot claim any type of intellectual property rights on it.
  3. I did not create the content myself. The content was synthesized from the public information available from various sources. The AI engine has copied a lot of information as is from the sources. All the content belongs to someone else. Since I may not know the source, I will treat this content as somebody else’s work and will not publish it under my name.
  4. I did not create the content myself. If I publish any part of the content as is, I will disclose that fact to the audience. I will use appropriate disclaimer about it as well.
  5. I am the best judge of my subject matter. I will use generated information as another source of information and not the final authority.
  6. If in doubt, seek legal advice from your organization’s counsel.

There are certain topics that I personally think should be kept out of limits from use of Generative AI. If you are asking questions like “What do you think?” or “What is your opinion on the topic?”, you are eventually going to find yourself in a bad spot. Keep in mind that Generative AI engine does not know anything about the information it is providing you. It is synthesizing past information into an output that matches with patterns of the input you provided in your questions. There is no personal thinking involved in this process. At the end of the day, a human must form an opinion or provide judgement. Use Generative AI as an intelligent assistant to quickly gather information you need for your final output. Do not ask the engine to form the final output for you.

Do not ask the engine about topics that could jeopardize the physical safety of humans.

Do not ask the engine about topics leading to forming any legal opinions.

Do not copy paste any source code directly into your applications as is. Keep in mind that Generative AI has provided suggestions based on the previous work by other people. It does not know the actual context of your application. It does not know anything about your security policies. It does not know anything about coding and QA standards used by your organization. Copying the source code as is will lead you in the direction of plagiarism and copyright violations as well. If you find yourself in this situation, check with legal counsel. I am not suggesting that you cannot use Generative AI to provide you with some suggestions. It will help in shortening your research time on the topic. Write the actual code how you will write to reflect quality of your and your organization’s standards.

Stay Current

This is an emerging field. The journey has just begun. There are going to be twists and turns in the future. Rules of engagement will change over time. If you are going to make Generative AI a big part of your work, then you will need to keep yourself current. I am not just talking about the use of technology. Keep yourself informed about legal and ethical issues. Designate a person in your group who keeps the rest of the members informed about the latest developments.

Power Comes with Responsibility

I will conclude the topic with a famous quote.

“With great power comes great responsibility.”

Generative AI is an emerging field that has potential to provide a lot of power to increase productivity and creativity. It is up to you to decide how you want to wield that power. Use it wisely and ethically. Always keep in mind that it is a tool created by humans for humans. Human factors cannot be separated from it. As humans you have to make a judgment regarding the use of technology.


Disclaimer: The views expressed in this article are of the author. They do not reflect the opinion or views of the current employer.

Tags: ai chatgpt dale-e generative ai

Share this post
Recent Posts
July9
Limitations and risks associated with ChatGPT

Shared by Naveen Kohli

Time: 8:34 AM

July8
July8
July7
What is text to image Generative AI

Shared by Naveen Kohli

Time: 10:41 AM

Popular AI Tools