in Snippets

Interesting video showing some of the ‘failure modes’ of large language models such as ChatGPT. On the one hand it is comforting to recognise that these tools can reveal themselves to be fallible machines. On the other, I fear that despite this, people won’t heed these flaws and they will integrate them into everything anyway.

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.

  1. @adoran2 I share your fear.

    Spend some time with ChatGPT and you start to see its limitations.

    ChatGPT might be fine for getting help with an English essay. But technology that constantly makes stuff up shouldn’t be used in life-or-death situations.

  2. @jakelacaze Yeah, absolutely spot on. The fact that people don’t really understand how an LLM works — or, at least, why it works as well as it does — definitely precludes it being integrated to a point where it ‘takes the wheel’ on anything. But I fear that we’ll collectively just do it anyway.