this post was submitted on 28 Apr 2025
105 points (98.2% liked)

Saturday Morning Breakfast Cereal

681 readers
113 users here now

Posts and discussion about the webcomic Saturday Morning Breakfast Cereal by Hugo Award-winning author Zach Weinersmith (and related works)

https://www.smbc-comics.com/

https://www.patreon.com/ZachWeinersmith

@ZachWeinersmith@mastodon.social

New comics posted whenever they get posted on the site, and old comics posted every day until we catch up in a decade or so

founded 6 months ago
MODERATORS
 

http://www.smbc-comics.com/comic/consciousness-9

Alt textThe key to real AGI is to make 4,000 different small AIs that hate each other.

Bonus panelBonus panel

top 10 comments
sorted by: hot top controversial new old
[–] Albbi@lemmy.ca 18 points 2 weeks ago
[–] will_a113@lemm.ee 8 points 2 weeks ago (1 children)

That alt text about making a bunch of AIs that hate each other is basically how antagonistic machine learning works.

[–] davidgro@lemmy.world 1 points 2 weeks ago

Yes, but that's only like 2 or so networks.

[–] davidgro@lemmy.world 5 points 2 weeks ago (1 children)

I've often wondered if the quest for Artificial General Intelligence has failed so far because even Natural General Intelligence doesn't exist - exactly as expressed in this comic.

My evidence is the way brain injuries and such can take out Very precise functionality (stuff like prosopagnosia) instead of just generally making things worse in every case.

[–] m_f@discuss.online 3 points 2 weeks ago (1 children)

Yeah, I think we've already achieved human-like AI, simply because of how low the bar is set by humanity. People tend to overrate the human experience IMO. Part of that is tied up in people's belief in souls and whatnot

[–] davidgro@lemmy.world 2 points 2 weeks ago

I disagree with the wording "human-like" because they don't at all work the way humans do (and that's part of my first point - it may be the wrong approach entirely to try to generalize as we seem to do) current systems are better at certain narrow tasks which don't involve hard facts and are extremely bad at others, which they still perform with full confidence and mislead people who think they are thinking.

But yeah, your main point is right: these things easily and regularly passing the Turing Test is a statement about human unintelligence and not artificial intelligence.

[–] DragonTypeWyvern@midwest.social 3 points 2 weeks ago (1 children)

Literally the plot of Echopraxis

[–] m_f@discuss.online 5 points 2 weeks ago (2 children)

This book, I assume? https://en.wikipedia.org/wiki/Echopraxia_(novel)

I'll have to check it out, I don't think I've read any of his stuff before, but I've heard Blindsight recommended a lot

[–] davidgro@lemmy.world 3 points 2 weeks ago* (last edited 2 weeks ago)

I need to re-read that one myself, especially now that we have

spoilerworking "Chinese rooms" (LLMs) in the real world.

Note that the book is free on the author's site.

[–] Yaky@slrpnk.net 2 points 2 weeks ago

I just finished Blindsight, and I am a bit torn in my opinions. It had really thought-provoking ideas about aliens, mind, conscience and subconscious, which stuck with me. But some of the descriptions, events, and characters are difficult to follow and reasoning (including key points in the story) does not make sense.