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 12 months 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 version” replace.
The popup reads:
“We made more improvements to the ChatGPT model! It should be generally better across a wide range of topics and has improved factuality.”
As well as, now you can cease ChatGPT whereas it’s within the strategy 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 study 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 knowledge.
When requested what’s within the new ChatGPT replace, the AI tells me:
“ChatGPT is a large language model developed by OpenAI that is trained on a dataset of conversational text. Since it’s an OpenAI model, It’s updated regularly with new features and improvements. However, as my knowledge cut off is in 2021 I don’t have the latest updates and I cannot provide you a specific information on which new features or improvements have been added to the model in recent updates.”
When requested one other means, I used to be capable of get extra perception into what’s concerned in enhancing the accuracy of an AI mannequin:
“In general, when an update is made to an AI model like ChatGPT, it can be aimed to improve its performance in one or multiple aspects. For example, it can have a bigger dataset to train on, or improvements in its architecture that allows it to generate more fluent and accurate responses. And such update also make it better in factuality, i.e its ability to understand and generate responses based on factual information.”
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 will produce completely different solutions.
Take a look at One: Failed
Beforehand, ChatGPT couldn’t precisely establish what number of occasions Argentina gained 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 gained it two occasions. As soon as in 1978 and once more in 1986.
As proven within the tweet beneath, ChatGPT didn’t at all times return the fitting reply:
Requested #ChatGPT abt who gained the FIFA world cup in 2022. It could not reply. That is anticipated. Nevertheless, it appears to offer incorrect info (abt the opposite 2 wins) regardless that the data is there within the system. Any #Explanations? pic.twitter.com/fvxe05N12p
— indranil sinharoy (@indranil_leo) December 29, 2022
I ran the immediate by way of the up to date model of ChatGPT, and it returned a unique however nonetheless incorrect reply.
Take a look at Two: Failed
Beforehand, ChatGPT was unable to offer an accurate reply when requested who’s the taller basketball participant between Shaq and Yao Ming.
I ran the immediate by way of the up to date model of ChatGPT, and it confidently returned the identical incorrect reply.
Going by way of the ChatGPT failures linked above, I discovered it continues to wrestle with the identical prompts.
It’s troublesome 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 knowledge. Though it supplies appropriate solutions to many questions, it’s at present not reliable sufficient to exchange Google.
Featured Picture: CHUAN CHUAN/Shutterstock