Why ChatGPT fails | Language Model Limitations EXPLAINED

How it is possible that ChatGPT makes so many incorrect statements and spits out wrong facts? We explain why ChatGPT isn't a truth oracle.
β–Ί Sponsor: Arize AI. Sign up for Arize:
Free industry certification in ML observability:
Learn about embedding drift from Arize research:

Check out our daily #MachineLearning Quiz Questions:
πŸ“œ ChatGPT blog:
πŸ“œ Evaluating the Factual Consistency of Large Language Models Through Summarization:
πŸ“œ WebGPT: Improving the Factual Accuracy of Language Models through Web Browsing:
πŸ“œ Behavior cloning is miscalibrated:
πŸ“Ί ChatGPT vs. Sparrow:
πŸ“œ Transformers as Algorithms: Generalization and Stability in In-context Learning:
πŸ“œ Do Prompt-Based Models Really Understand the Meaning of their Prompts?:

Thanks to our Patrons who support us in Tier 2, 3, 4: πŸ™
Dres. Trost GbR, Siltax, Edvard GrΓΈdem, Vignesh Valliappan, Mutual Information, Mike Ton

Outline:
00:00 ChatGPT spits out wrong facts
02:15 Arize AI (Sponsor)
03:40 How does ChaGPT / a language model work?
05:53 Why ChatGPT generates nonsense
06:38 Confidence and clarifications
07:21 Limits of behavioral cloning
09:04 Phrasing
09:21 Jail breaks
09:45 Is ChatGPT even usable?

β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€
πŸ”₯ Optionally, pay us a coffee to help with our Coffee Bean production! β˜•
Patreon:
Ko-fi:
β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€β–€

πŸ”— Links:
AICoffeeBreakQuiz:
Twitter:
Reddit:
YouTube:

#AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​
Music 🎡 : Intentions – Anno Domini Beats
Video editing: Nils Trost

25 Comments

  1. It’s always fun to see these new chatbots say something silly, but it’s even nicer to have an idea of why it happens πŸ˜€ Thanks for the video! πŸ˜€

    PS: I found the secret πŸ˜‰

  2. well, there’s no way to say which is the fourth child since you don’t specify the order they were born

    1. Right, that’s how I interpreted the question also. Maybe you’d have to say something like β€œThe names of three children are Lili, Lala, and Lulu. What’s the name of the remaining child in the list?” (Also, the question breaks a convention in Englishβ€”you _wouldn’t say_ someone β€œhas four children” and then just name three of them. It’s much more likely to say β€œThe other three children are…” so maybe it’s not surprising that ChatGPT got β€œconfused.”)

    1. ​@AI Coffee Break with Letitia This is a great summary of the fundamental flaws behind ChatGPT. Millions of people are just trusting it blindly now. This video is perfect for sharing and informing them. Nice work as always!

  3. Thanks for the critical and nuanced view on these language models. While I’m excited by the cool research and its potential, that potential is sullied by the tech bros with an interest in capitalizing on it as quickly as possible without considering the consequences. Like a lot of what we call AI, it amounts to a very sophisticated parlor trick; interpolation at a high enough dimensionality will look exactly like extrapolation to the credulous. I don’t know much about linguistics (or language models), but my intuition is that we’ll never have a truly conversational model as long as the models are optimized solely for communication, like language models that predict text expression of language. Chomsky makes the compelling argument that the structure of language is optimized for some other internal tasks, and it’s actually inefficient for communication.

  4. i asked the same question to chatgpt. I get a different answer which is also incorrect. It says Letitia Nkungu from Canada

  5. Language models don’t have semantic. Hence it is people’s idiocy to see semantics when looking only into the syntax. Before being a language model, it has to be fully about language, first.

  6. Great video! I tried many times to stop the video at 5:13 to see what was flashed on the screen but couldn’t manage it! πŸ€”

    1. I think there are shortcuts to step through a video frame by frame. I don’t know it by heart, though. Another way would be to watch that segment with 0.25 speed.

    2. @DerPylz I slowed it down to 0.25x and was able to stop it to see the message flashed “Why do you go through the video frame by frame” or something like that. πŸ‘

  7. The community is taking this seriously and exposing the vulnerabilities of this model to the public. Amazing video. Good effort in aggregating and summarising the information.
    Thank you Ms coffee bean

  8. Very good video! Thank you!

    Here just a shy note: In “Katys father …” you forgot the apostrophe: “Katy’s father”. With the apostrophe ChatGPT gives the right answer:

    “The name of the fourth child is Katy. The question states that “Katy’s father” has four children, and the names of three children are given as Lili, Lala, and Lulu. Since the question refers to Katy’s father and does not mention any other children, Katy must be the fourth child. Therefore, the complete list of children is Lili, Lala, Lulu, and Katy.”

Leave a Reply

Your email address will not be published. Required fields are marked *

© 2024 EVERYTHING CHATGPT - WordPress Theme by WPEnjoy