Mark Cuban: AI and Creativity, Gap Between AI Evals and Impact, New American Open Source AI Model

Captions as YouTube publishes them, usually machine-generated, so product and people names can be misheard.

1. 0:00

   Good morning Legion. Today is the day J1 Championship season officially kicks off. Don't miss the Degen Arena prediction market for the team to the best odds to finish on the podium, as well as those most favored to finish last.

2. 0:13

   We kick off our AI stories today with some recent hot takes from known billionaire Mark Cuban. He shares two of his opinions on how AI is going to affect the future. First, that in the future there will be two kinds of companies those that are grade AI and those that used to be in business.

3. 0:29

   this means, more concretely, that eventually everyone will have to incorporate AI into their work processes. And it's highly likely that we're only at the very beginning of the AI transformation. The second opinion from Mark has to do with artists, and how creative potential will be affected by AI, as he writes.

4. 0:45

   Creators should love AI doesn't make uncreative people creative. It allows creators to become exponentially more creative, creative iteration that used to take hours, days, or weeks can happen in minutes. The different routes you wanted to take but didn't have the money or skill to do.

5. 1:00

   Now you can do the number of gatekeepers drops every day and will eventually hit zero. If you know how to tell a great story. You will love AI. More and more opinions seem to be changing. The AI might just be to benefit the creative potential of humanity as opposed to limiting or stifling it. Are we on the brink of a new Renaissance?

6. 1:18

   And in that renaissance, will blockchain and NFTs known technologies for creator benefits? Will they play a role? Our second story is straight from Silicon Valley, where Ilya outlines the gap between evals and the economic impact of models today.

7. 1:34

   In other words, why are models performing well on hard evaluations but still failing to become serious economic actors? His answer: Reinforcement learning makes the models too narrow minded and unaware, and because of this, they can't do basic things. Or.

8. 1:48

   And more concretely, his second explanation has to do with data. Reinforcement learning. Training environments are geared towards results in scores as opposed to actual real world functionality. Until that's fixed. Can we hope for models to be serious economic agents? Probably not. Our third and final story of the day.

9. 2:07

   There's a new open source model on the block called Rnj-1 base trained on 8 billion parameters. Among the open source models, it scores pretty well. Take a look at the graph on the screen and should remain on your radar. As open source models continue to build out and become more performant in the future.

10. 2:23

    Rnj-1 is the culmination of ten months of hard work by a phenomenal team dedicated to advancing American, state-of-the-art, open source AI. Lots of AI advancements, most of them directionally positive for the development of NEAR. Stick around. We'll do this all again real soon.