Sep 14, 2023·edited Sep 15, 2023Liked by daniel bashir
Another fascinating talk. A pivotal quote from the 1st hour:
"We are not climbing towards physics here. We are climbing toward a different kind of global abstraction that give us the fundamental forces governing everyday life. We are not trying to find the computations governing nature as nature.. We are trying to find the mental language for the gnomics for the world of human life and human experience and human objects." Here, Grietzer is exploring some Kantian switch-backs through the symbolist-connectionist debate. It strikes me that the SC debate is in search of its own critique or synthesis. Or at the very least, the debate is following philosophical trajectories very similar to or are harkening back to those predating the arrival of Kant's critiques.
I personally found Grietzer's command of the Kantian dictum to be impressive and his efforts to synthesize it with ML and AI quite promising and fascinating. While I am not personally a Kantian--more a critical realist--I still think there is much to learn from the systematic nature of the approach. To what extent can mind be encompassed by Kantian descriptions of mind --- or ML or AI computational approximations of mind? Are these models steadily encroaching on what we previously thought was ineffable or even impossible to say? There is a lot more to say... but I will have to leave that for a post of my own.
You're doing a great job bringing in the (continental) humanities perspective on AI. There's more to AI and art than the rise of 10x content creators, and more to the philosophy of AI than trolley problems. I suppose literature studies with its ties to NLP and the linguistic turn of philosophy has an advantage at this point in this regard.
To add some hopefully constructive criticism: I think the discussion could have benefitted from having had some lines of thought intercepted earlier. Certainly on these topics it's hard for the guest to anticipate the depth and scope of the answers, and to know the audience. Meanwhile it is obvious that you (the interviewer) are very well-versed in these matters. With that joint caliber, you can afford risking some stray bullet shots.
Thanks, I really appreciate the feedback (and you were too kind in how you phrased it)! I certainly could have done more to pace things and make sure we built up some of the ideas a bit more incrementally. I definitely don’t want these interviews to have too high a barrier to entry
So many words. After 2.5 hours, I still have no idea what the point of this monologue was. I heard about autoencoders and philosophy. Maybe a conversational style of interview could be more effective + enjoyable for the listeners?
If you’ve listened in before, the podcast format is conversational—as I said in another comment I definitely could have interrupted more, but leaned too far in the other direction
Yes, this was meant as a constructive comment. I actually have listened to most episodes of the podcast. And in general the questions are very, very sparse, and most episodes definitely feel like monologues. This is fine with some guests, but IMO it feels very laborious with others. The highlights of this very episode were actually the few times you intervened to summarize the past 10 minutes of monologue or so. Again, this is just a constructive suggestion for the future! :)
Thanks! I try not to insert myself too much since I view the point as hearing from the guest, but you’re right that the best strategy probably depends on the person.
And I really appreciate the feedback (and thanks for listening as well)! I don’t hear much so it can be hard to calibrate (eg what I think of as a conversational format) 🙂
Another fascinating talk. A pivotal quote from the 1st hour:
"We are not climbing towards physics here. We are climbing toward a different kind of global abstraction that give us the fundamental forces governing everyday life. We are not trying to find the computations governing nature as nature.. We are trying to find the mental language for the gnomics for the world of human life and human experience and human objects." Here, Grietzer is exploring some Kantian switch-backs through the symbolist-connectionist debate. It strikes me that the SC debate is in search of its own critique or synthesis. Or at the very least, the debate is following philosophical trajectories very similar to or are harkening back to those predating the arrival of Kant's critiques.
I personally found Grietzer's command of the Kantian dictum to be impressive and his efforts to synthesize it with ML and AI quite promising and fascinating. While I am not personally a Kantian--more a critical realist--I still think there is much to learn from the systematic nature of the approach. To what extent can mind be encompassed by Kantian descriptions of mind --- or ML or AI computational approximations of mind? Are these models steadily encroaching on what we previously thought was ineffable or even impossible to say? There is a lot more to say... but I will have to leave that for a post of my own.
You're doing a great job bringing in the (continental) humanities perspective on AI. There's more to AI and art than the rise of 10x content creators, and more to the philosophy of AI than trolley problems. I suppose literature studies with its ties to NLP and the linguistic turn of philosophy has an advantage at this point in this regard.
To add some hopefully constructive criticism: I think the discussion could have benefitted from having had some lines of thought intercepted earlier. Certainly on these topics it's hard for the guest to anticipate the depth and scope of the answers, and to know the audience. Meanwhile it is obvious that you (the interviewer) are very well-versed in these matters. With that joint caliber, you can afford risking some stray bullet shots.
Thanks, I really appreciate the feedback (and you were too kind in how you phrased it)! I certainly could have done more to pace things and make sure we built up some of the ideas a bit more incrementally. I definitely don’t want these interviews to have too high a barrier to entry
So many words. After 2.5 hours, I still have no idea what the point of this monologue was. I heard about autoencoders and philosophy. Maybe a conversational style of interview could be more effective + enjoyable for the listeners?
If you’ve listened in before, the podcast format is conversational—as I said in another comment I definitely could have interrupted more, but leaned too far in the other direction
Yes, this was meant as a constructive comment. I actually have listened to most episodes of the podcast. And in general the questions are very, very sparse, and most episodes definitely feel like monologues. This is fine with some guests, but IMO it feels very laborious with others. The highlights of this very episode were actually the few times you intervened to summarize the past 10 minutes of monologue or so. Again, this is just a constructive suggestion for the future! :)
Thanks! I try not to insert myself too much since I view the point as hearing from the guest, but you’re right that the best strategy probably depends on the person.
And I really appreciate the feedback (and thanks for listening as well)! I don’t hear much so it can be hard to calibrate (eg what I think of as a conversational format) 🙂