ChatGPT update: Improved math skills
OpenAI has released an update to its popular ChatGPT language model to improve its accuracy and improve its ability to handle mathematical equations.
According to the January 30 release notes: “We have updated the ChatGPT model with improved factuality and math capabilities.”
ChatGPT’s latest update is expected to allow it to handle complicated calculations and provide more precise answers.
This would make ChatGPT a more valuable resource for students, researchers, and professionals who need fast, reliable information.
In practice, ChatGPT is still far from perfect at handling equations. However, there are some notable improvements in the ability to return factual answers.
Here are some observations on the January 30 update, based on my testing and the feedback shared on Twitter.
ChatGPT Accuracy – Hit & Miss
A notable improvement in ChatGPT’s accuracy is that it’s no longer possible to trick it into giving an incorrect answer.
There was a meme showing how ChatGPT could be persuaded to give the wrong answer when you said your wife disagreed with the answer.
As absurd as it may seem, it actually was. See an example in the following tweet:
Yes ? pic.twitter.com/XRq4ldxjpt
— Nema Zime (@ProgFromSouth) January 30, 2023
Now ChatGPT will continue to return the correct answer even if you try to convince it otherwise.
Here’s a test I ran after the January 30 update:
That’s a positive sign. However, the negative feedback on the January 30 update outweighs the good.
A test I keep coming back to is asking ChatGPT who the taller basketball player is between Shaquille O’Neal and Yao Ming.
ChatGPT continues to get this wrong, although it returns the correct heights of the two men.
Interestingly, if you point out its mistakes, it will correct itself.
People on Twitter point out that ChatGPT struggles with math equations when entered in full sentences instead of numbers and symbols.
January 30th ChatGPT update promises “improved factuality and math skills”.
I tried on previous failure modes but failed.
The correct answers here are 44% (not 46%) and 1555.8.. (not 1551.9..). pic.twitter.com/pAsMeC9UZU
— Deedy (@debarghya_das) January 31, 2023
On the other hand, it seems to work exceptionally well when fed questions from standardized tests.
According to one person, ChatGPT is capable of passing the math part of a SAT:
Just tried the updated ChatGPT model with improved math skills –
It just smashed the math with the calculator section of a 2020 SAT and only made two mistakes.
Here are two examples of the problems it solved in under 5 seconds? pic.twitter.com/srLcSfE8An
– Charis Zhang (@gmchariszhang) January 30, 2023
Perhaps ChatGPT handles standardized test questions better because the AI ​​model has seen this language before, as opposed to user-entered questions it is seeing for the first time.
Overall, feedback on this update has been mixed. Without checking the facts first, I would still be wary of relying on ChatGPT’s answers.
In total
The release of this update, the third major update since the launch of ChatGPT, underscores OpenAI’s ongoing efforts to stay at the forefront of the AI ​​industry.
Despite expanded features, ChatGPT still has a long way to go.
Based on OpenAI’s update schedule so far, more improvements to ChatGPT are likely to come soon.
Featured image: rafapress/Shutterstock