My Predictions On What I Think Will Be Successful ChatGPT Plugins...

My Predictions On What I Think Will Be Successful ChatGPT Plugins...

Plugins on ChatGPT are software that is independently developed and installed on ChatGPT to help ChatGPT access up-to-date information, run computations, or use third-party services.


Those last three things are essential to note…


What this means is we tech nerds can code apps and install them officially on OpenAIs ChatGPT, and then users can choose to use them to do those things that ChatGPT cannot.


There is still a waitlist for both users of and developers of Plugins (I’ve been on the waitlist for at least 3 months now and still do not have access).


So, why am I so keen?


100,000,000 monthly active users and not many Plugins will be a gold rush for brands looking to get access to new attention and awareness and leverage it within ChatGPT for generating leads and sales.


As I cannot develop (and more importantly test Plugins), I wanted to write my thoughts on what I think will constitute successful ChatGPT Plugins…


There are two things to think about:


1. ChatGPT core features (and limitations); and

2. ChatGPT user experience.


On the one hand, ChatGPT core features (including the GPT4 model), are trained up until September 2021.


GPT4 is including a web-browsing model which allows users to scrape a website (I have written about how to avoid that before) and have Artificial Intelligence do something with that information – computations (broadly speaking) such as analyse, interpret, summarise etc.


On the other hand, as a user, to use ChatGPTs web-browsing model, the following steps are required:


  1. Open an internet browser
  2. Navigate to https://chat.openai.com/
  3. Log in
  4. Start a new chat
  5. Select Plugin as the model
  6. Select the web-browsing Plugin
  7. Enter a prompt
  8. Submit the prompt


As a user, to use ChatGPT *Plugins* the following steps are required:


  1. Open an internet browser
  2. Navigate to https://chat.openai.com/
  3. Log in
  4. Start a new chat
  5. Select Plugin as the model
  6. Select the specific Plugin
  7. Enter a prompt
  8. Submit the prompt


Both require 8 steps.


If a user wanted to get information from a website without the “computations”, all they need to do is:


  1. Open an internet browser;
  2. Navigate to the URL they want; and
  3. Peruse or enter the URL for the information they need.


5 fewer steps than ChatGPT.


A rule of thumb for user experience is the less friction the better the user experience is.


5 more steps are a lot of friction to get a user to choose ChatGPT Plugins or ChatGPT web browsing over a URL, unless it is worth their time or effort, even if it’s micro friction.


Now, ChatGPT has stated in the developer documents that ChatGPT would only choose the independently developed Plugin *IF*it seems relevant.


In other words, if the Plugin is simply replicating what ChatGPT can already do, or ChatGPT can complete the user prompt without the Plugin, it will.


So, the first prediction:


ChatGPT Plugins need to provide additional functionality to ChatGPT.


This means providing data after September 2021 or anything in real-time that cannot be accessed from a URL.


The second prediction:


ChatGPT Plugins need to run computations that ChatGPT cannot.


One obvious one is complex maths – it’s been proven (https://www.pcguide.com/apps/chat-gpt-cant-do-math/) that ChatGPT is really bad at this, and the Wolfram Plugin has already been developed to fill this hole.


The third prediction:


ChatGPT Plugins need to integrate with third-party web services - that being webforms, Application Programming Interfaces, etc


One obvious one is ordering – ChatGPT cannot do this, and the Instacart Plugin has already been developed to fill this hole.


Now, although both example Plugins cited arguably fill gaps in ChatGPT core features, run computations or integrate with third-party web services, it’s questionable that the added 5 extra steps to access them via ChatGPT compared to natively via their URLs, it’s too be proven when 100,000,000 users get access to them.


Only time will tell with this.


But to summarise, successful ChatGPT apps will have the following features:


  1. Access to data post-September 2021 and ideally real-time; and
  2. Do something with that information ChatGPT cannot do out-of-the-box.


Let’s see if I am right…

Muhammad Tayyab

Helping Job Seekers!

1 年

Wow, Orren Prunckun! Your predictions on ChatGPT plugins are spot-on! As a fellow biz dev expert, I appreciate your insights on how these plugins can help ChatGPT provide more efficient and accurate information. Keep up the great work in MarTech and design thinking! Looking forward to more of your content. #chatgpt #marTech #bizdev #designthinking #upskillrevolution #linkedinexpert Also you can check all the 88 plugins here: https://www.dhirubhai.net/pulse/unleashing-power-ai-comprehensive-guide-chatgpt-plugin-tayyab-1f/?trackingId=ff%2BLAn5XQtCVuonecyD8cg%3D%3D

回复

Let's see if they deliver on their promise this time: https://twitter.com/OpenAI/status/1657128759659745280

回复

要查看或添加评论,请登录

Orren Prunckun的更多文章

社区洞察

其他会员也浏览了