←back to thread

Jagged AGI: o3, Gemini 2.5, and everything after

(www.oneusefulthing.org)
265 points ctoth | 5 comments | 20 Apr 25 14:55 UTC | HN request time: 0.833s | source
Show context
mellosouls ◴[20 Apr 25 17:44 UTC] No.43745240[source]▶
The capabilities of AI post gpt3 have become extraordinary and clearly in many cases superhuman.

However (as the article admits) there is still no general agreement of what AGI is, or how we (or even if we can) get there from here.

What there is is a growing and often naïve excitement that anticipates it as coming into view, and unfortunately that will be accompanied by the hype-merchants desperate to be first to "call it".

This article seems reasonable in some ways but unfortunately falls into the latter category with its title and sloganeering.

"AGI" in the title of any article should be seen as a cautionary flag. On HN - if anywhere - we need to be on the alert for this.

replies(13): >>43745398 #>>43745959 #>>43746159 #>>43746204 #>>43746319 #>>43746355 #>>43746427 #>>43746447 #>>43746522 #>>43746657 #>>43746801 #>>43749837 #>>43795216 #
1. nightmunnas ◴[20 Apr 25 20:35 UTC] No.43746355[source]▶
Low agreeableness will actually be extremely useful in many use cases, such as scientific discovery and of course programming assistance. It's amazing that this venue hasn't been explored more deeply.
replies(2): >>43746369 #>>43747758 #
2. Jensson ◴[20 Apr 25 20:39 UTC] No.43746369[source]▶
Its much easier to sell an agreeable assistant than a disagreeable one, so it isn't that strange the alternative isn't explored.
3. alwa ◴[21 Apr 25 01:15 UTC] No.43747758[source]▶
Why would a bad attitude be helpful in those domains? Are the human partners wont to deliver more effort when you’re mean to them?

Are we talking about something other than Agreeableness in the personality research sense [0]?

The strongest form of your argument I can think of is “willing to contradict you when it thinks you’re wrong”—but you can disagree agreeably, right? The current-gen LLMs certainly have with me, perhaps because my custom prompt encourages them to skepticism—but they do it so nicely!

[0] https://en.m.wikipedia.org/wiki/Agreeableness

replies(1): >>43749468 #
4. Jensson ◴[21 Apr 25 08:11 UTC] No.43749468[source]▶
> but you can disagree agreeably, right?

No, the concepts are linked, agreeable people don't want to be rude and most people see disagreements as being rude no matter how you frame it. You can't call a woman overweight without being rude for example no matter how you frame it, but maybe you want an AI that tells you that you weigh too much.

replies(1): >>43751357 #
5. smallwire ◴[21 Apr 25 12:43 UTC] No.43751357{3}[source]▶
Good point, but calling a woman overweight isn't necessarily a disagreement.