← Back to blogblog

OpenAI GPTs vs Chat Thing



OpenAI's rapid innovation in AI technology is a double-edged sword for startups using their APIs. Watching their developer events is a nerve racking experience! On one hand you are excited for amazing new features that can help us make our products even more powerful, but on the other hand OpenAI can release competing features that could make you redundant.

After OpenAI released GPTs, we had a number of people reach out and ask...

"Oh no, have OpenAI just killed Chat Thing?".

Well thankfully the dust has settled now, and after having some time to really dig into GPTs and their capabilities we were relieved to realise for most professional use cases there are still plenty of reason to use a service like Chat Thing over GPTs.

OpenAI GPTs compared to Chat Thing

At first glance it might seem like OpenAI GPTs and Chat Thing provide very similar functionality, and while GPTs do a great job of making custom AI chat bots more accessible to everyone, in a lot of areas they are severely lacking features hampering their overall usefulness.

In most cases, if you want to integrate AI chat bots into your business you are still probably better off using a service like Chat Thing. Of course we would say that, but we will let you make up your own mind after reading the points below.

Data Sources

OpenAI GPTs allow you to add "knowledge" to your bot by uploading up to 10 files. There are also size limits on these files, and you get charged $0.20 per day per gigabyte per bot! This is a great way of adding data to basic bots, but the limitations make it very impractical for anything more complicated.

Chat Thing's data sources

Chat Thing's data sources are leaps and bounds ahead of GPTs knowledge files in many ways:

  • Automatically connect to 8+ different types of data source, not just files
  • Automatically refresh data source content to keep it up to date
  • No limits on the number of documents in a data source
  • Advanced functionality such as web crawling

Bots you can use anywhere

When you create a GPT its forever imprisoned within the ChatGPT interface. That might be a bit dramatic, but it does make GPTs a lot less useful. Not only that but you can only use GPTs if you are a $20 a month ChatGPT plus subscriber, so this hugely limits who can use your bot if you share it publicly.

With Chat Thing if you share your bot via a URL anyone can use it regardless of whether they are a ChatGPT or Chat Thing subscriber. In addition to that, you can integrate your bot over many different channels such as Slack, Telegram, Discord, WhatsApp all in a few clicks. That's not possible with GPTs at the moment.

With Chat Thing you can even add an customisable embeddable widget to your site meaning you can use Chat Thing as a support agent helping to answer customer queries.

Chat Thing embeddable widget

In summary here's the way Chat Thing makes your bots more shareable:

  • 6+ channels, interact with your bot via the web, Slack, Discord, Telegram, WhatsApp and API.
  • Shareable bot URLs that are useable by everyone, not just ChatGPT subscribers
  • Embeddable and customisable chat widgets
  • Password protect bots

Stopping hallucinations

One of the biggest issues with large language models such as GPT is their tendency to hallucinate and make up responses and data. This can be very detrimental to the usefulness of AI bots as depending on your use case the accuracy of responses might be very important. You don't want your support bot making things up when a customer has a problem, that will only anger them more!

With OpenAI GPTs, the only way to fight against hallucinations is via prompt engineering. You can write something along the lines of "Only answer questions using the content in the provided documents. If the answer isn't there say 'I do not know'". This technique works some of the time, but especially as the chat history gets longer the bots tend to stop paying as much attention to the prompt and can still hallucinate.

In Chat Thing you can use prompt engineering to prevent against hallucinations just like you can with GPTs, but we have also gone further and developed something we call response guards. Response guards basically get the bot to double check its responses before errors before sending them back to the user. This massively improves the reliability of the bot in giving back accurate information.

A screen shot of the response guard interface

Conversation Logs

With OpenAI GPTs you have no way of seeing the conversations users have with your bot. Whilst this is good for privacy, it does limit GPTs for certain use cases. If you want to use them for customer support, lead generation, or many other uses having visibility of conversations will be essential.

Chat Thing provides you with full searchable conversation history so you can see how users are interacting with your bot. In addition to that user's can mark responses as positive or negative so you can see where your bot is lacking and make improvements in your data sources, or prompts etc.

Not only that, but you can export all your bot's conversations as a CSV if you need to work with the data in other tools, and we are working more advanced features, such as bot usage statistics and graphs. None of which is available in GPTs.

A screen shot of the Chat Thing conversations interface

A screenshot of the new bot statistics screen

Personal Support

Although getting started with creating an AI bot using GPTs or Chat Thing is quick and easy, tuning them to work exaclty how you want can be quite an art. There are many settings to tweak. Should I set a custom prompt? How do I change the tone? How do I stop hallucinations? Whats the best way to stop hallucinations. The list goes on!

With OpenAI GPTs you are completely on your own trying to figure out these issues, but with Chat Thing we have dedicated support with knowledge of helping people make thousands of bots ready to help whenever you need it.

Don't just take our word for it, we get messages like the one below in our Discord community all the time.

A message praising Chat Thing and our amazing support in Discord

We have 850 members in our Discord where you can get supports, or if email is more your style you can always email us there too and we will be happy to help. You won't get this kind of service setting up your GPTs from OpenAI.

Adaptability and new features

At Chat Thing we have the advantage of being a small nimble team that can adapt to customer requests and ideas at the click of your fingers. We have sooooo much in the pipeline we are really exited about, and we believe we can innovate at a faster pace than larger companies.

We are on the cusp of releasing a new and improved bot dashboard, and then shortly after that we will be releasing the first version of functions giving your bots powers they have never had before.

Our long term vision for Chat Thing is to build a product that provides users with the easiest and most powerful way to integrate AI assistants into your websites and products. Chat Thing, should not only be a support tool, but a pro-active co-pilot that can not only guide users but take actions on their behalf!

Summary

In summary we are really pleased about the additional exposure OpenAI GPTs will provide to the bots industry. They are awesome for quickly dipping your toes in the world of AI bots, but for more professional use cases you will still need to reach for a service with more advanced features such as Chat Thing.