Crypto x AI: Investing Across The Stack | Crypto x AI Event

Bullish:

• $AIAGENTS - AI agents are expected to become major consumers of products and a significant part of the economy within the next 5 years, potentially revolutionizing the consumer side of various industries.

  • “I actually think it's very possible that the most widely used product is consumed by bots, and bots become your users. I think that what we underestimate is right now, most products are consumed by human beings. And I really think in the next 5 years, agents or bots or whatever you wanna call them, humanoid robots, whatever, software is gonna be a huge part of the consumer side of the economy moving forward.”

    Casey·1:08:03
• $EMBODIEDAI - Embodied AI, which combines language models with sensory inputs, is expected to dramatically enhance AI capabilities and applications.

  • “I think once you hook these language models up with eyes, ears, ability to understand their physical world, it it's just gonna blow our minds and what they can do.”

    Casey·56:01
• $CRYPTOAI - Open source AI is showing promise in the crypto sector, with improving quality of open source models serving as a foundation for crypto AI development.

  • “The first thing I would say is that the skepticism is well warranted, and we're going day by day here. It's a really nascent category, but I do think there's a lot of promise here and some early signs that it's gonna materialize into a really robust sector. One thing that I think is underappreciated is how how good open source is getting and how that is a foundation for crypto AI.”

    Casey·4:03
• $DECENTCOMP - Decentralized compute networks are seen as crucial for competing with large tech companies in AI model training, potentially enabling the creation of larger and more effective networks.

  • “I think if we think that we can build larger, more effective networks than a meta or an OpenAI to train these kinds of things, then the only way that we're gonna be able to do that is through a decentralized setup because that will be the only way we can, you know, source enough compute to do that.”

    Jake·15:05

  • “I do think if we can make some progress there, we're gonna see a number of models, you know, probably reaching, like, 50,000,000,000 parameters quite soon, which are gonna start to become actionable, like, you know, in like, intellect 1, from prime intellect, just just train 10,000,000,000 parameters. When we get to, like, around a 100,000,000,000, I think people will be able to use that for, you know, kind of basic LLM purposes.”

    Jake·1:06:02
• $DATAOWN - Technologies enabling private data ownership and model fine-tuning are gaining traction, allowing users to maintain privacy while benefiting from AI model improvements.

  • “And it's you know, I really like what Bagel that Jake mentioned is doing where they can enable you to fine tune, a my a model that is remains private. The weights remain private on data that can also remain private to, say, the user or the the customer that has that data, and then get, like, a a fine tune model over that that you can then query.”

    Will·52:02