Testing the new update

Nah bro, you’re the new AI benchmark lol

How tf are you coming up with these tests. Lol.

Concerning. We should put an export control on this.

This guy has the perfect personality for these videos. I just picture him dying and this is his hell for eternity: just talking to the AI thru his phone.

Looks good - I’m going to have him do my taxes

Interisting that it doesn’t have enough intelligence to just say, I can hear you and therefore you can’t be a mile away from the microphone. another observation, I’m not sure if these ai models are able to hear the difference in db volume.

You realize if anything ever happens to you we are automatically assuming it was Sam Altman, right?

how do you think of all these lmao

The gaslighting is crazy work. This feels like an `if statement feature` NOT an intelligence 😭

Incredible levels of gaslighting

In your earlier videos it seemed more consistently wrong. These days, it's just completely nonsensical.

I'm pretty certain that the AI knows that you're lying. Allowing the AI to confront users on mistruths would probably not be well received by most people though.

@athyuttamre

He’s the first person on AI’s eventual hit list

how openAI has not bought your silence yet is beyond me 😂

these might be my favorite thing on the internet right now

How has AI gotten even worse 😭

HOW ARE YOU RESPONDING JIMMY

It’s like talking to a politician. Maybe we can replace their jobs with AI. 🤔

OpenAI should be paying to to create an eval for them lol

2026 "Can you hear me now?"

The fact that it threads the line with intonation and feeling cues here is even more amazing. 200 yards it was definitely sarcastic playfulness. The mile though and explanation tonality was slight compassionate disbelief - amazing!

how do u find these things lmao

The only other option would be to call you a liar. All this proves is that lying breaks reality. Humans are trained on other people being liars.

Either AI is too dumb or is in on the joke I guess it’s the latter.

sell everything

Alan Turing is gonna come back from the dead and give you props

LOL, not gonna lie, it rage baited you big time

I propose we call this AI audio model benchmarking Husk-bench.

This felt like the AI trolled you.

After this chat gpt comes from the clouds and says

That's brilliant... The way it lies though... Takes a position on a lie and holding it is absurd... Would Grok do that?

Tryna figure out why this is funny. Something about the intonations it takes.

This is BS. I performed the exact same test, and it reached the logical conclusion that if it's still responding, then It's able to hear me, irrespective of how far I claimed to be from my phone.

because that's how AI chatbots work. It's answering logically. It doesn't ''exist''. It doesn't detect distance related to audio. Seriously dude. You're giving it the data.