> This class of bug seems to be in the harness, not in the model itself. It’s somehow labelling internal reasoning messages as coming from the user, which is why the model is so confident that “No, you said that.”
from the article.
I don't think the evidence supports this. It's not mislabelling things, it's fabricating things the user said. That's not part of reasoning.
This is great, and I'm not knocking it, but every time I see these apps it reminds me of my phone.
My 2021 Google Pixel 6, when offline, can transcribe speech to text, and also corrects things contextually. it can make a mistake, and as I continue to speak, it will go back and correct something earlier in the sentence. What tech does Google have shoved in there that predates Whisper and Qwen by five years? And why do we now need a 1Gb of transformers to do it on a more powerful platform?
> I generally only attempt to scrutinize government action, and not government reason for action
This might be the wildest opinion I've read.
You're onboard with the US bombing another country ("I like the war"), but you don't know, or care WHY. You just think it was a good idea.
"Random citizens are at such an information disadvantage that I think it would be impossible to have an informed opinion as an outsider on the reasoning."
I'm trying to give you the benefit of the doubt here, but if you re-read your own words, you've just said a random citizen like yourself can't possibly know enough to have an informed opinion, yet you gave us your opinion, which is that you think they should have bombed Iran.
Trump casually talks about destroying the energy infrastructure, power plants, desalination plants etc. This is one of the most controversial things that the Russians do in Ukraine - attack the grid when it's cold to try and freeze people to death. To willingly deprive a country of 100,000,000 people of water and power coming into summer would surely be a war-crime.