Gemini revealed the score of the Super Bowl before the game even started

Reading time icon 2 min. read


Readers help support MSPoweruser. When you make a purchase using links on our site, we may earn an affiliate commission. Tooltip Icon

Read the affiliate disclosure page to find out how can you help MSPoweruser effortlessly and without spending any money. Read more

Key notes

  • Before the Super Bowl, Gemini and Copilot provided detailed accounts of the game, including scores and player statistics. However, these accounts were entirely fictional.

During the recent anticipation surrounding Super Bowl LVIII, two prominent chatbots, Google’s Gemini and Microsoft’s Copilot, generated inaccurate information regarding the game’s outcome. This incident was seen on Reddit and sheds light on the capabilities and limitations of large language models (LLMs).

Before the Super Bowl, Gemini and Copilot provided detailed accounts of the game, including scores and player statistics. However, these accounts were entirely fictional. For example, Gemini attributed 286 rushing yards to Kansas City Chiefs quarterback Patrick Mahomes, a statistically improbable feat. 

Additionally, Copilot incorrectly declared the San Francisco 49ers victors with a score of 24-21. In contrast, the game resulted in a thrilling 25-22 victory for the Kansas City Chiefs over the San Francisco 49ers.

These incidents highlight a key limitation of LLMs: their reliance on statistical patterns rather than true understanding. Trained on massive text datasets, these models can generate grammatically correct and seemingly coherent text, but accuracy is not guaranteed. Moreover, it didn’t even understand the context of the question; the user asked about the odds of the game. 

It is recommended that Users should not rely solely on such outputs and should always verify information through trusted sources, especially when dealing with unfamiliar sources.

Users had a good time sharing their POVs on this; one of the users commented. 

Comment
byu/Sp00kbee from discussion
inChatGPT

While others pointed out their trust issues in Gemini by commenting

“I can’t even ask Gemini to help me write basic HTML, and y’all out here trusting it to predict the future…”

Overall, Gemini is full of errors now, and users prefer GPT4 over it, but Google has publicly revealed that they are working on it. We also have a guide that helps you decide which one to go for – Gemini Advanced, Copilot Pro, or ChatGPT Plus. 

More about the topics: gemini, Super Bowl