Should Rasa be worried about GPT-3?

Hello everyone, I came across an article by Mark Ryan in Medium that got my attention, where he compared GPT-3 vs. Rasa chatbots (this is the link for anyone interested in to read https://towardsdatascience.com/gpt-3-vs-rasa-chatbots-8b3041daf91d).

In the article, he stated that while Rasa chatbots were more accurate than GPT-3, they (ie Rasa chatbots) took longer to build than GPT-3 chatbots. Also, he was of the opinion that GPT-3 success in this area of speed could make it revolutionize chatbots creation methodology in the future.

After reading his article, I will love to recommend that the Rasa team take this potential threat seriously even if GPT-3 is still flawed as was stated in this Rasa blog (GPT-3: Careful First Impressions) and look for alternatives to give Rasa a strategic advantage over GPT-3 and other products out there, that for now we are have no idea exist such as:

  1. coming up with an upgraded algorithm or feature that will make it equal or defeat GPT-3, but maintain Rasa’s quality which for now is considered a gold standard (eg Apple vs IBM);
  2. looking for a way to incorporate GPT-3 into Rasa (or vice versa), so that Rasa will ride on the back of GPT-3 success to greatness (eg Microsoft riding on the backs of Apple and IBM to success); or
  3. trying a Blue Ocean Strategy where a unique solution is created that will give Rasa a niche that will remain uncontested for the foreseeable future (eg Apple creation of the iPod, and iTunes).

The reason for my message is because I have a feeling we might be witnessing something similar to Nuclear Arms Race (USA vs USSR), or a Search Engine Arms Race (Google vs Yahoo/Excite), or a Social Media Arms Race (Facebook vs MySpace), whereby the winner might just take all. However, I pray I am just paranoid, that Rasa will always triumph and be fine, as well as, doesn’t need to worry over other competitors.

That will be absolutely great.

Nonetheless, we must show humility from history and remember that winners like Google, Facebook, Steve Jobs, Bill Gates, and Jeff Bezos, made it by:

  1. Being obsessed with giving the final consumer more or unique value: and,
  2. Obeyed Charles Darwin’s Law of Adaptation.

I hope my message will be helpful to all of you. Also, I want to commend you all for the great work you all have done especially with the forums, blogs, and YouTube tutorials. Thank you for helping us all, and I wish you all success.

2 Likes

Thanks for the thoughtful post. I am a huge fan of Rasa for its versatility, openness and pragmatic architecture. I invested a lot of time to learn about creating chatbots in Rasa and I have been impressed with the generosity and creativity of the Rasa community. For all that, I feel that GPT-3 (or rather, its successors) represents a threat to Rasa as a general-purpose chatbot framework. I suppose time will tell. In the mean time, I’ve posted a video demo of the experiment described in the article you cite: GPT-3 vs. a Rasa chatbot - YouTube