Australia's largest stock trading and investment forum Australia's #1 stock forum

AI needs training and source data. If that source data is biased...

First 92 Last

Share
Profile

Follow
It investor

3,845 Posts.

275

27/03/24

09:03

Post #: 73075698
Share

Originally posted by Parsifal: ↑

there is more then one type of AI and it is used in so many different ways. There are also many different AI “models”. For example in image generation you can use a range of tools and they each tend to have a particular style. However no matter which ai program is used it is recognisably AI (deep fakes are another matter) And I’m yet to see one that can produce hands that are accurate

language models are also of various quality and accuracy. But can be incredibly useful if you know enough to pick mistakes. Universities and tafes are having to address how to pick cheating

https://en.wikipedia.org/wiki/Google_Brain

. Ai is also currently unpredictable. This makes for interesting reading - seems like trying to be “balanced” brings its own problems

https://www.semafor.com/article/03/01/2024/googles-ai-problems-expose-deeper-industry-dilemma

Expand

AI needs training and source data. If that source data is biased or faulty then the output is. Images are a representation of source data. AI will always be a representation of where the developers point the model to consume data.
1 Upvote

Reply

There are more pages in this discussion • 1 more message in this thread...

First 92 Last

SPONSORED BY