OpenAI’s ChatGPT Update Brings Improved Accuracy


OpenAI, the corporate behind the favored conversational AI mannequin, ChatGPT, has launched an replace to enhance the chatbot’s accuracy.

Following an prolonged interval of downtime on Tuesday, ChatGPT is up and working with a brand new mannequin.

That is the primary replace to ChatGPT this yr and the second replace to the mannequin since its launch in November.

A popup message lists the adjustments in what OpenAI calls the “Jan 9 model” replace.

OpenAI’s ChatGPT Update Brings Improved AccuracyScreenshot from:, January 2023.

The popup reads:

“We made extra enhancements to the ChatGPT mannequin! It must be typically higher throughout a variety of subjects and has improved factuality.”

As well as, now you can cease ChatGPT whereas it’s within the means of producing a response, which is a closely requested characteristic.

Particulars within the launch notes are restricted, so I examined a number of prompts in ChatGPT to be taught extra in regards to the new replace. Right here’s what I found.

Information Nonetheless Restricted To 2021

Whereas ChatGPT could also be extra correct now, its data remains to be restricted to 2021 information.

When requested what’s within the new ChatGPT replace, the AI tells me:

“ChatGPT is a big language mannequin developed by OpenAI that’s educated on a dataset of conversational textual content. Because it’s an OpenAI mannequin, It’s up to date commonly with new options and enhancements. Nevertheless, as my data lower off is in 2021 I don’t have the newest updates and I can not present you a particular info on which new options or enhancements have been added to the mannequin in latest updates.”

When requested one other manner, I used to be in a position to get extra perception into what’s concerned in bettering the accuracy of an AI mannequin:

“Usually, when an replace is made to an AI mannequin like ChatGPT, it may be aimed to enhance its efficiency in a single or a number of elements. For instance, it could actually have a much bigger dataset to coach on, or enhancements in its structure that permits it to generate extra fluent and correct responses. And such replace additionally make it higher in factuality, i.e its skill to know and generate responses based mostly on factual info.”

Accuracy Stays Questionable

OpenAI claims ChatGPT can now present extra factual solutions.

I examined that declare by consulting a GitHub repository of ChatGPT failures and working a number of prompts to see if it might produce totally different solutions.

Take a look at One: Failed

Beforehand, ChatGPT couldn’t precisely establish what number of instances Argentina received the FIFA World Cup.

Disregarding the 2022 World Cup win, as a result of ChatGPT’s data is proscribed to 2021, it ought to say Argentina has received it two instances. As soon as in 1978 and once more in 1986.

As proven within the tweet beneath, ChatGPT didn’t all the time return the best reply:

I ran the immediate by means of the up to date model of ChatGPT, and it returned a special however nonetheless incorrect reply.

OpenAI’s ChatGPT Update Brings Improved AccuracyScreenshot from:, January 2023.

Take a look at Two: Failed

Beforehand, ChatGPT was unable to supply an accurate reply when requested who’s the taller basketball participant between Shaq and Yao Ming.

I ran the immediate by means of the up to date model of ChatGPT, and it confidently returned the identical incorrect reply.

OpenAI’s ChatGPT Update Brings Improved AccuracyScreenshot from:, January 2023.

Going by means of the ChatGPT failures linked above, I discovered it continues to battle with the identical prompts.

It’s tough to pinpoint the areas wherein ChatGPT can return extra correct responses. It might be useful if OpenAI might present particular particulars within the launch notes of future updates.

That mentioned, watch out when utilizing ChatGPT as a supply of data. Though it gives appropriate solutions to many questions, it’s at the moment not reliable sufficient to switch Google.

Supply: OpenAI

Featured Picture: CHUAN CHUAN/Shutterstock


Scroll to Top