Tom Schmidt
👤 SpeakerVoice Profile Active
This person's voice can be automatically recognized across podcast episodes using AI voice matching.
Appearances Over Time
Podcast Appearances
I was looking actually at some of the new AI models that, again, in the spirit of letting people choose their own path, there's sort of this newfound open source AI renaissance.
And so people are pretty happy about GLM 5.2, which is a Chinese model that came out and has...
performance that is on par, maybe he's surpassing Opus 4.8 and people say, hey, if you orchestrate it correctly, it's similar to Fable or OpenRouter, which is founded by Alex Atala, pharmacy of OpenSea.
We are recording a podcast with him next week.
Oh, cool.
Also came out with this sort of hybrid model that allows you to stitch together many different models to get sort of table-level performance.
And so it feels like there's this cool undercurrent and sort of similarities actually between crypto, what's happening in the AI industry right now, in terms of a move towards open source and sort of self-sovereign compute, which is, I think, a much more optimistic vision for the future than the alternative.
I think we see some kind of fruits of that today.
I think just in terms of like, hey, how do these things ultimately get funded?
How do you actually access, compute for this if it gets restricted on an inference provider level?
Maybe there's a way to tap into a more distributed inference network and actually access some of these models.
But
Today, it's more of a personality and sort of ethos alignment, I would say, where you see a lot of early crypto people coming to AI and vice versa, but ultimately driven by the same sort of ideas that, again, go back to kind of the early days of computing.
Totally.
And I think there's also this sort of private compute component in terms of just data.
You know, people, I think, trust, you know, these chatbots with a whole lot of personal information that they probably would not trust even Google with, not realizing that, hey, this is, you know, can be retained for, you know, arbitrary amounts of time by these companies that's subject to, you know, subpoena, can be trained on a good leak, whatever.
And I think that that
It seems unsustainable to me that maybe a future is one where, yeah, people have had their entire personal lives held on a single company's servers.
And it seems like there's a move towards having local or distributed private data, maybe accompanied with some sort of remote compute.
But it does seem like these two are kind of going to collide at some point.