I have been tinkering with ChatGPT for the last 2 days and I love it! I can see it saving me so much time! ChatGPT is so much better than google. Google searches are purely based on keyword matching. When I have a question and I do not know the keywords for the answer, I can only hope that my question was already asked somewhere on websites like quora, reddit, or stack overflow. The good thing about ChatGPT is that it can be trained on anything where there might not be any instance of my question. It can use that information to provide the answer I'm looking for. I see how ChatGPT is a threat to google and I'm excited to see how this can revolutionize search!

I see two problems/opportunities:

First - It should be able to easily index new data. ChatGPT currently cannot do that automatically. It has to be manually trained with the new data and the new model has to be released. Just solving this can make it a bigger threat to google. The only other step would be to build the data pipelines to make this re-indexing work. After that ChatGPT can almost completely replace google.

Second - ChatGPT is trained on public data. If we want a general-purpose search software, it should be able to search through anything. It should easily be re-trainable on any data. Currently, such re-training doesn't seem straightforward (at least to someone who doesn't have any knowledge of ML). If this is solved, it can become a much bigger success! It can be trained on "law books" making it an unmatchable assistant to any lawyer. It can be trained on internal organization data - where searching for the right information/documentation is a huge problem. It can also be trained to serve as an amazing chatbot.

Try out ChatGPT: https://chat.openai.com/