@jcoglan
he/him : you may know me as @mountain_ghosts on twitter : I wrote some books you can buy from https://shop.jcoglan.com
Relays (1)
- wss://relay.ditto.pub – read & write
Recent Notes
somehow I only recently learned that LLMs need the whole conversation fed back to them on each prompt so their i/o cost scales as O(n^2), something that would be considered completely unacceptable in almost any other production network-accessible software
1
this story about openai "breaking containment" is such bullshit if you know anything about how computer security is supposed to be done. if a program accesses a remote system it's because you decided ...
if your "containment" is "telling the robot to be a good boy" you're not a serious person
this story about openai "breaking containment" is such bullshit if you know anything about how computer security is supposed to be done. if a program accesses a remote system it's because you decided to let it do that. LLMs don't have magical powers that can work around "we unplugged the ethernet"
1
no computing appliance, no matter how complicated, can ever gain capabilities beyond those of a turing machine, or by church-turing, the untyped lambda calculus. anything a computer does must arise fr...
you can't actually make a computing system that is anything but a large graph of these operations. you can make it as big as you like but it will only ever do those 3 things. therefore you need to show that those 3 things and nothing more can give rise to consciousness
one thing that irks me about the "complexity -> consciousness" argument is that computer people do not have a well defined idea of what "complexity" is, it is a "I know it when I see it" thing. wherea...
no computing appliance, no matter how complicated, can ever gain capabilities beyond those of a turing machine, or by church-turing, the untyped lambda calculus. anything a computer does must arise from some composition of `x`, `\x -> e`, and `x y` (names, abstractions and applications)
1
one thing that irks me about the "complexity -> consciousness" argument is that computer people do not have a well defined idea of what "complexity" is, it is a "I know it when I see it" thing. whereas we do have a very rigorous formal model of what a computer can do
1
as I have been saying: LLMs are a form of anti-computing
LLMs have taken steve jobs' "a computer is a bicycle for the mind" and replaced it with driving a huge american pickup truck to the corner store that's a five-minute walk away
one reason that "AI" is not producing huge shifts in profitability is that, for most businesses, raw production speed of software has not been the bottleneck on generating or increasing revenue for at least as long as I've been working in tech (20 years)
you know what else can turn poorly defined specs into poorly written programs for the cost of a software developer salary
I'm hearing the perpetual motion machine costs a lot to run? huh
1
if only there was a way of telling a computer what to do that didn't require paying data centre landlords four figures a month
you know what else can turn poorly defined specs into poorly written programs for the cost of a software developer salary
1
if only there was a way of telling a computer what to do that didn't require paying data centre landlords four figures a month
1