ChatGPT Now Has ‘Improved’ Math Capabilities And Can No Longer Be Bullied Into Giving The Wrong Answer

OpenAI regularly adds updates to ChatGPT to make it better and faster. 

ChatGPT can offer easy solutions to complex queries.

Source : Getty

ChatGPT is an advanced AI-driven chatbot designed to generate surprisingly human-like responses to queries. US-based research lab OpenAI released a prototype version of ChatGPT in November 2022. Ever since then, it appears that humans have taken it upon themselves to prove that the machine-learning-run AI tool isn’t up to the task. Social media was flooded with various users posting how they managed to one-up ChatGPT, either by highlighting its wrong answers or by simply bullying it into admitting a wrong answer. Such shenanigans could very well be a thing of the past now.

On January 30, OpenAI announced that it has upgraded the ChatGPT model “with improved factuality and mathematical capabilities.” Since it’s still a prototype, OpenAI regularly adds updates to the bot to make it better and faster. 

Now, prior to the recent update, ChatGPT had this weird hack that allowed users to exploit it by repeatedly asking it to admit a wrong answer. For example, even if it gave the write answer to a logic-based question, it could still be ‘bullied’ into admitting a wrong answer, if the user repeatedly typed in the wrong answer, challenging ChatGPT’s response.

With the new update, ChatGPT remains adamant and keeps flashing the correct answer, no matter what you tell the chatbot. 

Yeah 😎 pic.twitter.com/XRq4ldxjpt

— Nema Zime (@ProgFromSouth) January 30, 2023

As reported by Search Engine Journal, ChatGPT is still getting basic questions wrong, such as who is taller — Shaquille O’Neal or Yao Ming.

In this screenshot shared by Search Engine Journal, ChatGPT continues to get this height-related question wrong.

However, it does appear that following the latest January 30 update, ChatGPT manages to return correct responses to complex queries.

ChatGPT's Jan 30 update promises "improved factuality and mathematical capabilities".

I tried it on previous failure modes, but it failed.

The right answers here are 44% (not 46%) and 1555.8.. (not 1551.9..). pic.twitter.com/pAsMeC9UZU

— Deedy (@debarghya_das) January 31, 2023

Just tried the upgraded ChatGPT model with improved math capabilities -

It just crushed the math with calculator section of a 2020 SAT and only made two mistakes.

Here are two examples of the problems it was solving in less than 5 seconds🤯 pic.twitter.com/srLcSfE8An

— Charis Zhang (@gmchariszhang) January 30, 2023

Disclaimer: The article majorly includes the responses given by ChatGPT (AI-driven chatbot developed by OpenAI) to various questions/questionnaire and ABP Network Private Limited (‘ABP’) is in no manner liable/responsible for any of such responses. Accordingly, reader discretion is advised.