contextualizing machines wrt energy
Dr Brian Keating Dan Dennett: His Final Interview https://youtu.be/5r8vMk0Zgds?t=1389 (machine transcribed - obvious errors & sentence structure not yet added…)

Brian Keating:

Speaking about you replication selection, you make the case in the book and based on your earlier works, you know that language is almost also a certainly uh an evolved process and there are higher order and lower order languages although we don't see you know Denise Evans walking around, we still do see primitive languages with uh you know even non-written languages that exist on the planet. I interviewed your cross town rival uh Noom Chomsky and he sort of made a persuasive case to me had implications for artificial intelligence and that was that you know the communication there's so much non-verbal and and even most durable form of of communication might be generated non-verbally and that made me think what extent can these LLMs that you mention in the book as well can they ever achieve you know a Turing test level? We’ll talk about the deficiencies and problems you have with the Turing test later but um, could it could a llm that doesn't have a body and you know doesn’t have the ability to pop a circuit, you know to cause it to feel pain when it does something wrong could it ever hope to evolve you know or or to to be present itself as a as almost you know human level artificial intelligence if it doesn’t have embodiment?

Daniel Dennett

Well of it has to have embodiment in once and you have to have some some Hardware to run the software on one of the points that I like to stress these days is that brains are not at all like Foman machines, they're not they’re not uh one of the amazing things about computers and touring was very clear about this is that they have to be very bureaucratic, they have to be very rigid have an operating system and they have to do you have to know exactly what they’re going to do in order to program them, their design depends on the uniformity on the fact that there’s billions trillions of exactly identical elements almost to the yes to the atomic level, your flip-flops your your registers and you have timing pulses and all brains aren't like that, brains are made of billions of individualistic neurons runs no two are exactly alike and they don't act on the quite the same time scale so it’s an entirely different underlying structure. Now could you nevertheless build such a structure in a silicon uh digital computer, yeah sure you could because you can you can simulate it in the same I mean you can simulate an analog machine in a digital machine, you can simulate a parallel machine in a serial machine you pay big prices for that in time and and energy, LLMs are incredible energy Hogs compared with human brains, energy is really key and some people for instance Terry Deacon argues that for all all the wonders of computers they ended up getting us to explore the wrong part of design space because all the computers that were designed following touring and F noon and the like were in a sense he calls them parasitic, they didn't have to worry about energy they were provided for by their plug into the power and this means that all the designs all of the space that we’ve explored has been space that depends on there being a sort of Shepherd nursaid to take care of the machine to make sure its energy, no no neuron no circuit has to worry about whether it’s going to be alive or whether it gets enough energy whereas the neurons in your brain are working for a living, they they will die if they don't connect and so your brain is made of neurons and G cells even more that that the the neurons are looking for work and there’s no HR Director there’s no Human Resources Director and our neuron resource director and when when a part of the brain dies neurons hungry for work will take over those tasks we don't have anything much like that in in digital computers yet so brains are they are computers of course they’re not radiators, they're not for cooling the blood, they're they're control centers they are the the control headquarters for movable arms and legs for for mobile things that’s what brains are, are and so they're not much like digital computers Still Still you could simulate all in principle, in principle…

Do you like what you are reading? Subscribe to receive updates.

Unsubscribe anytime

Attributes
name
contextualizing machines wrt energy
summary
Dr Brian Keating Dan Dennett: His Final Interview https://youtu.be/5r8vMk0Zgds?t=1389 (machine transcribed - obvious errors & sentence structure not yet added…)