AI cannot count the letter “R” in the word “Strawberry”. But why is it that most systems fail at such a comparably simple task?
Artificial intelligence and especially large language models (LLMs) can accomplish many tasks in impressive ways. With tools like ChatGPT or Google Gemini, writing essays and solving complex equations is often no longer a problem.
But in some cases, AI systems fail because of simple things, like spelling a single word correctly. It often happens that artificial intelligence provides the wrong answer to the question of how often the letter “R” appears in the word “Strawberry”.
These errors make it clear that AI systems are extremely powerful, but not human. Algorithms do not “think” like we do and therefore have no understanding of basic linguistic concepts such as letters or syllables. But why are complex mathematical formulas often not a problem, while the English word “Strawberry” throws almost all models off course?
AI cannot count “R” in “Strawberry” due to tokenization
This is mainly because LLMs are based on transformer architectures. This break down the text into so-called “tokens”. Depending on the model, these tokens can represent entire words, syllables or individual letters. A tool converts the entered text into a numerical representation, which is then processed by the AI system behind it.
So the AI might know that “straw” and “berry” together make “strawberry.” But she doesn’t understand exactly which letters the word consists of. This mechanism makes it difficult for the AI to recognize exact letters or their number in a word.
One of the biggest challenges in this problem is defining what a “word” means for a language model. Even if it were possible to create a perfect token vocabulary list, LLMs would likely still have difficulty processing more complex linguistic structures.
Different languages follow different grammatical rules
It becomes particularly difficult when an LLM has to learn several languages. Because some languages like Chinese or Japanese don’t have spaces to separate words. This makes tokenization even more complex. A possible solution would be for language models to work directly with individual characters instead of tokenization.
But at the moment this is too computationally intensive for transformer models. As technologies continue to evolve, it remains to be seen how well future AIs can handle these challenges. Perhaps the almost infinite computing power of a quantum computer will one day enable artificial intelligence to absorb and understand grammar like a human.
Also interesting:
- Self-healing power grid: Artificial intelligence should prevent blackouts
- AI gap: Artificial intelligence is creating an even deeper “digital divide”
- AI as a judge: The advantages and disadvantages of artificial intelligence in the judiciary
- CT scans: Artificial intelligence could save thousands of lives
The post AI can’t count the letter “R” in “Strawberry” – that’s why by Felix Baumann appeared first on BASIC thinking. Follow us too Facebook, Twitter and Instagram.
As a Tech Industry expert, I find it fascinating that AI technology, despite its incredible capabilities, still struggles with seemingly simple tasks like counting the letter “R” in a word like “strawberry.” This highlights the limitations of current AI systems and the challenges that come with natural language processing.
The fact that AI can’t accurately count the letter “R” in “strawberry” may be due to the complexities of language and the nuances of human communication. Words can be spelled and pronounced in different ways, and AI systems may struggle to accurately process and understand these variations.
This issue underscores the importance of continual development and improvement in AI technology. As we strive to make AI smarter and more sophisticated, we must also address the limitations and shortcomings that come with it. By identifying and addressing challenges like this, we can work towards creating more advanced and reliable AI systems in the future.
Credits