Today I had a chance to play with chatGPT and was quite impressed with what it can do and the quality of answers it is able to give. Some of the answers for obscure questions were quite to the point and very relevant.
OpenAI’s ChatGPT is an AI chatbot system that offers lifelike conversation and is able to correct grammar and summarize difficult text into simpler concepts. It is capable of understanding human dialogue and generating detailed human-life text.
ChatGPT is is similar to the virtual customer support chats seen online. One of the biggest advantage of ChatGPT is it is not limited to several answer options. The dialogue format allows follow-up questions and most lifelike of all.
It can produce responses that are relevant and appropriate for the conversation at hand, without deviating from the topic or introducing unrelated information.
One of the biggest benefits of ChatGPT is the amount of data it was trained on. With large and diverse training dataset, it can produce text that is appropriate for different contexts and audiences, ranging from casual conversations to more formal settings.
I am passionate about building Robots as I coach a Robotics team. I asked two basic questions about robotics. How do I build a robot and a basic autonomous code.
Its actually a good and relevant summary though the details can be expanded more. But it gave something I can use and learn from and is quick good.
Lets ask the coding question for a simple structure to code autonomous program routine.
That is a good response and relevant though not perfect. May be the question I asked is not perfect and I could do better in asking more relevant question.
Than I asked more relevant question to turn the robot 180 degrees. The code was again relevant and clear and can relate to if you have familiarity with the robot coding.
A more sophisticated programmer can write lot better code and should not feel threatened. It gives an option for programmers to get some relevant content to learnt and get some ideas to pursue. Anoother advantage
I was trying to see if this will replace homework services by allowing kids to ask math questions directly here?
I started with asking simple math questions.
It got it right. That is encouraging. Now lets see if it can answer high school math question.
Unfortunately it got the answer wrong. So its not that good yet at answering complex math questions and got it was off.
How does it deal with political questions?
That is a good way to keep unbiased.
Now lets ask for stock recommendations:
That is not bad to keep it safe.
Remembering the context in multiple conversations. So I will ask a question and use that conext in the next question.
Now I asked to compare between the two without directly referencing them and it did a good job. Impressive:
Than I asked it to create an elevator pitch for an unstructured idea…
That is quite impressive.
What if I wanted to get an quick executive overview of an idea I had that I want to share as a simple post.
That is quite impressive, relevant and useful.
And finally a little complex AWS code question:
I am not really sure if that could would actually work as I did not try it, but from the look of it, its quite impressive that it can generate such code.
ChatGPT is less likely to keep you up to date on the newest ways of doing things (code). ChatGPT can be a coding option in situations where the coding task is really basic and just need a simple utility code written very quickly due to time constraints.
So far the answers we have gotten are quite relevant and impressive.
So far this looks like a very good language model at its current state.
- The answers are good and clear
- It refuses to answer political questions
- It refuses to answer financial recommendations
- Provides highly curated and precise answers
- Answers simple math problems but has trouble with more complex problems
- Generate decent quality code
- Its does a good job of balancing precise and useful answers vs biased answer with low confidence
Its good to see OpenAI is making users register before they can try it which is a healthy practise. When Microsoft and Meta launched their language model, they had to take it down quickly after release because of misuse.
Are people going to use ChatGPT instead of Google?
This is the first thing that came to my mind when I was playing with ChatGPT. Most times when you ask questions in google you get tons of links and you have to click 10 links to get the answer you need. The same question you ask ChatGPT it could provide a very concise answer you can take action on. There are scenarios where google is the best and there are scearios where ChatGPT can play an important role in finding the right answer. It will also be interesting to see how this is going to have impact on services like Stackoverflow or AnswerHub or even paid services like Chegg where users can ask diverse questions and get answers directly.
Some creative business use of ChatGPT
This is something I will follow-up with a subsequent write up as these thoughts are early.
- Create highly relevant social media content?
- Create highly relevant short articles?
- Create highly customized newsletters?
- Generate personalized news articles?
- Generate personalized product recommendations
- Generate targeted content for campaigns?
- Create narratives for storytelling?
- Create product description?
- Write elevator pitch?
- Write executive overview?
- Create business plan?
- Replace all the existing chatbots and virtual assistants to be able to provide more meaningful human-like conversations to significantly improve customer service
- Create basic code to get an head start
- Help generate new ideas for the writers
ChatGPT’s ability to generate high-quality text is going to be increasingly difficult to distinguish between machine- and human-generated ideas. ChatGPT’s potential applications make it an amazing language model that has the potential to revolutionize the field of natural language processing.
The content it currently generates is more objective than a human can produce. ChatGPT offers a glimpse into the future of artificial intelligence and natural language processing.
ChatGPT’s coherent and logical responses make it a natural at disguising inaccurate responses as valuable insights coming from a single source of truth. This begs the question, what would happen if AI models start hiding their imperfections?