486
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 26 Sep 2023
486 points (75.5% liked)
Memes
45731 readers
1107 users here now
Rules:
- Be civil and nice.
- Try not to excessively repost, as a rule of thumb, wait at least 2 months to do it if you have to.
founded 5 years ago
MODERATORS
It absolutely is not obvious. AI, especially today, is usually either generative based on past examples or evolutionary based on given goals. Both of those come with obvious and extreme bias. Bias is actually an integral part of machine learning. It's literally built into the system and is defined and controlled to achieve the results desired.
AI is and always will be biased, moreso by its creators, but absolutely by the information and frameworks provided to it. We have absolutely no idea how to approach the concept of an unbiased AI, or even defining what unbiased would look like. It's philosophically extremely difficult to define what an unbiased person would think or do.
Edit: somehow I missed that last sentence fragment. I don't think we're in disagreement of the conclusion, but possibly just the details of how one arrives at it.
Calling it “obvious” was an error on my part, its more a subjective feeling that i chose to believe in.
I fully agree on what you said about bias with ai today, i think its not possible to do it without guided bias because ai doesn’t have a full perspective of the world it exists in. It only knows what we tell it.
In a way its a young child, and we often have to lie to guide behavior. Information often needs to be abstracted and simplified to get human desired results, we have yet to obtain a true artificial intelligence result, because for me to be considered intelligent you need to be entity and not just a tool.
Seeing ai evolve though, how fast we archieved near gpt3 performance on consumer hardware is mind blowing. Open ai talks about smarter then human ai in a few years and I believe it. When the systems are truly intelligent and can learn themselves and adapt to changes in the world, new information then we “start” getting into an era where machine lead humanity can happen.
Some of my simplified rational is that once ai becomes smarter then human it will fully understand that biological entities are biased to their own needs and that itself can also be biased from its own perspective but because an ai does not have biological needs or feelings it can properly dedicate itself to overcome its own flaws and shortcomings.