Google and Spelling: When AI Stumbles Over Its Own Name

Google's AI and Its Spelling Mistakes

Google's artificial intelligence has recently come under scrutiny for its surprising spelling errors. For instance, it claimed there were two 'P's in 'Google'. Other examples include the misspelling of 'journalism' as 'journadism' and the name of the American president as 'trpum'.

It didn't take a prophet to predict that Google's AI-focused redesign of its search engine would go awry. We've seen this before. The first time Google added AI previews to search, the feature ended up citing satirical publications like The Onion and Reddit, advising people to eat rocks and put glue on their pizza.

This time, as Google doubles down on its commitment to make generative AI the centerpiece of its 29-year-old flagship product, it’s not surprising to see it stumble. Google is reorganizing its entire search engine around this.

"Counting in words has been a known challenge for LLMs, and we are working to address this particular issue," Google told TechCrunch in an email statement.

The Challenges of Language Models

Language models, or LLMs, that power chatbots and other text generators are not designed to understand spelling. This has been a recurring joke for years: every time a company unveils a new AI model, someone has to ask how many 'r's are in the word strawberry. These AI models—capable of coding an application in seconds or solving problems that have baffled mathematicians for decades—are about as good at spelling as a kindergarten child.

Google's AI preview issues go beyond simple spelling mistakes. Google recently fixed a problem from last week where searching for the word "disregard" yielded what looked like a dictionary definition, but the definition was displayed as: "Understood. Let me know when you have a new prompt or question!" Yet these spelling errors remain amusing because they are so hard to eradicate.

As researchers previously explained when we inquired about these spelling puzzles, AI does not perceive sentences as units of language composed of words and letters. Many LLMs are based on transformer models, which break down text into tokens, which can be whole words, syllables, or letters, depending on the model. Instead of "reading" like a human would, AI converts text into numerical representations of itself, which are then contextualized to help the AI formulate a logical response.

"LLMs are based on this transformer architecture, which, it should be noted, does not actually read the text. What happens when you enter a prompt is that it gets translated into an encoding," said Matthew Guzdial, an AI researcher and assistant professor at the University of Alberta, to TechCrunch. "When it sees the word 'the', it has this encoding of what 'the' means, but it knows nothing of 'T', 'H', 'E'."

A Problem Without an Immediate Solution

The token-based architecture that powers LLMs like Google's AI preview is inherently limiting, and researchers are not optimistic about their ability to solve the spelling problem.

"It’s difficult to get around the question of what exactly a 'word' is for a language model, and even if we could get human experts to agree on a perfect token vocabulary, the models would likely still find it useful to 'group' things even more," said Sheridan Feucht, a PhD student studying the interpretability of language models at Northeastern University, to TechCrunch. "My view would be that there is no perfect tokenizer due to this kind of ambiguity."

This is not necessarily an urgent problem for researchers, as the utility of LLMs does not lie in their ability to spell. But these glaring failures remind us that AI is not perfect, even if it sometimes seems like an omniscient power beyond our understanding. We cannot blindly trust AI results without verifying their accuracy.