13:30 Hey, just wanted to say thanks for letting me know about the Gray Swan Arena two months ago. As a person with no pro     See More xperience in the AI industry, I took part in it from curiosity, but surprisingly I ended up getting 54th place in the 'Red-Teaming' competition.    See Less

This is gold 🔥 I just launched a new faceless channel teaching these exact strategies with AI tools. Hope it grows like yours     See Less

America is creating the perfect storm to let ai spiral out of control as the rest of the world joins BRICS...as an American, I feel like a German right before WW2 in that my country is lost     See More barely any control over it.    See Less

So interesting to see benchmark performance AND PRICE. I'd love to see likewise, an estimate of the price of getting the human performance, at least insofar as such a thing makes sense.     See More passing the bar exam would be easy enough to estimate given that it correlates closely to a professional capacity. Some human prices could go as low as minimum_wage per time taken.    See Less

ai models would ultimately push down the value of any field it enters so that figure could still make sense if adjusted for work devaluation     See Less

cool!     See Less

I tried a where is Waldo on a very complicated image and it got it     See Less

Since I'm not a native speaker, I do not follow the "loop my left arm through the gap between my right arm and chest" - ergo - AGI of a non-native has been reached.     See Less

Not an AI expert but have been following the space closely over the last couple of years and would love some thoughts/discussion on something i've been thinking about recently regarding     See More quot;/"AI intelligence".

Namely, why are we fixated on aligning AI intelligence with human intelligence? Why do we, as humans, determine that our intelligence is even the right benchmark (of sorts) to measure against the capability of these AI models?

For example, we don't think "AGI" has been achieved because although these models can do advanced maths/science/etc they fall short at common problems that most humans are able to solve (can't remember off top of my head but examples are given in other videos). A lot of the time i think the AGI definition/argument is pointless and a waste of effort as it's just a title, in the same vein that when can you call a human 'intelligent'...it often depends on background/circumstance/expertise etc

To be clear. i'm not a particular AI advocate or think it's the be-all-and-end-all, just a guy wanting thoughts from people definitely smarter than himself!    See Less

I believe that handling spatial logic is not as difficult as it seems. By constructing a highly realistic 3D virtual environment that adheres to physical laws, and allowing intelligent agent     See More re within this space, the challenge becomes manageable, especially with the advancements in reinforcement learning (RL). The main difference lies in the amount of resources invested. This approach is almost inevitable, whether for creating physically accurate videos through text-to-video (T2V) or image-to-video (I2V), or for extensive robot training in the future. Virtual environments are essential. Once this stage is surpassed, I believe basic AGI will be achieved, akin to AI existing in an abstract virtual environment. The perception layer will essentially mimic low-precision human sensory capabilities. As computational power increases, virtual environments will become indistinguishable from the real world, requiring no new technologies for AI, just substantial investment and integration of existing ones.    See Less