Brian Maucere
👤 SpeakerAppearances Over Time
Podcast Appearances
This is from thenewscientist.com, by the way.
I cannot read the full story because I'm not subscribed.
But from what I can tell here, Kenneth Payne at King's College London set three leading large language models, which are 5.2.
He said Claude Sonnet 4 and Gemini 3.
Flash.
Okay.
Well, I already see a problem.
But okay.
Against each other.
Why Flash?
Hit each other in a simulated war games.
The scenarios involved intense international standoffs, including border disputes, competition for scarce resources, and existential threats to regime survival.
And then this is where it starts to cut off.
But anyway, you get the gist of it.
In this particular case, AI was suggesting using nuclear weapons in 95% of the issues.
Now, what's that I don't know about this one?
What were the controls in this study?
Why did he decide to use three Flash along with 5.2?
That's an odd choice for me.
Why would you go with Flash?