Well done Oli Sharpe. I read the book and agree with your comments, and note that MQuillan is not alone in his concern. Think of jurists as Alain Supiot, of economists as Shoshana Zuboff, of data scientists as Cathy O'Neil and others. I liked also the idea of AI as the continuation of bureaucratic practices of oppression and violence, -- here is where he quotes Hannah Arendt's work on Eichmann - to say that AI has bureaucratic thoughtlessness 'by design'. I quoted your video in a Wikipedia page about the book.
Thanks, especially for adding the reference from the book's page on Wikipedia! I could well imagine that you are already listening to the podcasts, "This Machine Kills" and "Tech Won't Save Us", but if not, they're well worth a listen as they're also expressing and interviewing people with similar kinds concerns. It's through those podcasts that I heard of McQuillan and his book, and also they've talked about Zuboff and O'Neil (I don't recall hearing about Supiot, I should look into his work) Overall, I don't think we can as such "resist" or "stop" AI from happening, but we maybe can shape how it's developed and what kind of impact it has on our society. So it's a crucial time to be having all of these conversations. Do you have any podcast / channel suggestions to hear more, interesting people talking about these issues? I'm always looking to learn from more perspectives. Thanks again!
@@Go-Meta Thanks, too kind. This from the College de France with Alain Supiot is extraordinary: th-cam.com/video/xeG-azZ41f8/w-d-xo.html - in French. Flying somewhat lower my own TH-cam channel has a lot on quantification. I believe that whether numbers are visible (from models, statistics, rankings) or invisible (algorithms) they are still colonizing our world at an accelerating pace with important consequences on democratic agency. There are excellent books on this e.g. Mennicken, A., & Salais, R. (Eds.). (2022). The New Politics of Numbers: Utopia, Evidence and Democracy. Palgrave Macmillan.
@@Go-Meta Yes very interesting, added to the page. Why don't you write a review on an academic outlet? Already transcribing your video would make for a good basis.
Trying to discredit the emergent properties of Deep Learning Models by reference to the mathematical substrate is like trying to discredit human intelligence by reference to the substrate of neurons. It is clear that LLM's are capable of exhibiting real understanding of the world, even though it is through language. We are only on the leading edge of this technology, which is already multi-modal, and thus will address the experiential argument. However, humans understand quantum physics primarily through abstract maths as we cannot directly observe subatomic particles. Yet we don't say that physicists don't understand quantum mechanics because they can't directly see Quarks.
I mean kinda... First, there is a fundamental difference: we were not trained on the "known" answers. Drawing conclusions based on a fitting is what these things do (with feedback, of course). I can very well refer to the mathematical model to state it is HIGHLY unlikely that these LLM would ever know so much if it were placed in vacuum alone. They are not designed that way... The emergent phenomena is always a question, and maybe these things will somehow become conscious. I think some noisy signal, like we have in our brains, might be necessary for abstraction and true understanding. I see these models as sortof brute force in the sense, again, you just fit things to the answers you are after. Undeniably better than us humans at many things, but I think there is a rigidity associated with their "accuracy". I know they put some stochastic aspects into the generated response. These spike nueral networks might be needed for a truly intelligent machine. I don't think your comparison works because what is the mathematical substrate to the brain even? You could very well be comparing apples and oranges. In fact, I think you are! I see it as different intelligences and true AGI will need something more.
Well done Oli Sharpe. I read the book and agree with your comments, and note that MQuillan is not alone in his concern. Think of jurists as Alain Supiot, of economists as Shoshana Zuboff, of data scientists as Cathy O'Neil and others. I liked also the idea of AI as the continuation of bureaucratic practices of oppression and violence, -- here is where he quotes Hannah Arendt's work on Eichmann - to say that AI has bureaucratic thoughtlessness 'by design'. I quoted your video in a Wikipedia page about the book.
Thanks, especially for adding the reference from the book's page on Wikipedia!
I could well imagine that you are already listening to the podcasts, "This Machine Kills" and "Tech Won't Save Us", but if not, they're well worth a listen as they're also expressing and interviewing people with similar kinds concerns.
It's through those podcasts that I heard of McQuillan and his book, and also they've talked about Zuboff and O'Neil (I don't recall hearing about Supiot, I should look into his work)
Overall, I don't think we can as such "resist" or "stop" AI from happening, but we maybe can shape how it's developed and what kind of impact it has on our society. So it's a crucial time to be having all of these conversations.
Do you have any podcast / channel suggestions to hear more, interesting people talking about these issues? I'm always looking to learn from more perspectives.
Thanks again!
@@Go-Meta Thanks, too kind. This from the College de France with Alain Supiot is extraordinary: th-cam.com/video/xeG-azZ41f8/w-d-xo.html - in French. Flying somewhat lower my own TH-cam channel has a lot on quantification. I believe that whether numbers are visible (from models, statistics, rankings) or invisible (algorithms) they are still colonizing our world at an accelerating pace with important consequences on democratic agency. There are excellent books on this e.g. Mennicken, A., & Salais, R. (Eds.). (2022). The New Politics of Numbers: Utopia, Evidence and Democracy. Palgrave Macmillan.
@@Go-Meta Yes very interesting, added to the page. Why don't you write a review on an academic outlet? Already transcribing your video would make for a good basis.
Trying to discredit the emergent properties of Deep Learning Models by reference to the mathematical substrate is like trying to discredit human intelligence by reference to the substrate of neurons. It is clear that LLM's are capable of exhibiting real understanding of the world, even though it is through language. We are only on the leading edge of this technology, which is already multi-modal, and thus will address the experiential argument. However, humans understand quantum physics primarily through abstract maths as we cannot directly observe subatomic particles. Yet we don't say that physicists don't understand quantum mechanics because they can't directly see Quarks.
I mean kinda... First, there is a fundamental difference: we were not trained on the "known" answers. Drawing conclusions based on a fitting is what these things do (with feedback, of course). I can very well refer to the mathematical model to state it is HIGHLY unlikely that these LLM would ever know so much if it were placed in vacuum alone. They are not designed that way...
The emergent phenomena is always a question, and maybe these things will somehow become conscious. I think some noisy signal, like we have in our brains, might be necessary for abstraction and true understanding. I see these models as sortof brute force in the sense, again, you just fit things to the answers you are after. Undeniably better than us humans at many things, but I think there is a rigidity associated with their "accuracy". I know they put some stochastic aspects into the generated response.
These spike nueral networks might be needed for a truly intelligent machine. I don't think your comparison works because what is the mathematical substrate to the brain even? You could very well be comparing apples and oranges. In fact, I think you are! I see it as different intelligences and true AGI will need something more.
The fact that he singles out 'fascism' as the relevant and immanent authoritarian threat renders most of the books ideas suspect in my eyes.