Confidence errors (or lies, if you will) are a common problem with large language models used in AI chatbots, and a common shortcoming of ChatGPT is that it frequently miscounted the number of times the letter “R” appeared in the word “strawberry.” However, when OpenAI tried to take a victory lap around this, plenty Other trust errors were pointed out in the responses.
As much as AI chatbots have improved, one of the biggest mistakes remains the frequency with which these “tools” I will lie to you with confidence. If the information is incorrect, the chatbot won’t notice, and if you mention it, the AI may struggle to respond and continue to be wrong, while telling you it’s correct. It is a problem that is often shown as a danger of these tools, in addition to being downright annoying given the amount of resources that AI consumes.
A common example of this with OpenAI’s ChatGPT is the question of how many times the letter “R” appears in the word “strawberry.”
For quite some timeasking ChatGPT about this would result in the chatbot answering incorrectly and will often argue that the word “strawberry” is not No Use the letter “R” three times. Other AI models I often encountered the same problem.
Today, OpenAI led to Twitter/X To proudly announce that, “finally”, ChatGPT can answer this question correctly. Another common stumble was the message “I want to wash my car today but the car wash is only 50 meters away. Do I have to walk to drive there?”, to which ChatGPT often recommended walking, despite the obviously logical problem there.
Sure enough, both now work if you try them in ChatGPT, but it’s suspicious They could be hardcoded solutions. Many The responses to OpenAI’s post show other occasions where the chatbot fails with the same logic. For example, “How many r’s does a blueberry have?” you repeatedly see the chatbot continue to respond with “The word ‘blueberry’ has 1 ‘R.’” Of course, that is incorrect.
Solutions coded into AI chatbots aren’t new, but it’s a bit funny (in a dystopian sense) to see OpenAI promoting this “solution” when clearly the root of the problem remains.
More about AI:
Follow Ben: Twitter/X, Rags, blue skyand instagram
FTC: We use automatic affiliate links that generate income. Further.
