Therefore, it is also plausible that general intelligence is achievable only through robust-first computing.
-
Show this thread
-
Indeed it's promising to explore how to create cognitive machinery based on robust first computing.
1 reply 0 retweets 0 likesShow this thread -
Replying to @IntuitMachine
Every useful computational system involves both isolation from, but also interaction with, its physical implementation and environment. Traditional computing's hardware determinism (for the former) plus explicit input/output (for the latter) was a place to start, but not to end.
1 reply 0 retweets 1 like -
Replying to @livcomp
Are you thus implying that a robust-first platform will require fuzzier boundaries between its constituent components? If so, what are the characteristics of these boundaries that differentiate itself from the sharper boundaries of classical computers?
1 reply 0 retweets 0 likes -
Replying to @IntuitMachine
Not so much 'requires' as 'accepts'. Non-deterministic execution is a fact of life for truly large-scale computing. So the distinction between 'implementation' and 'environment' cannot be made completely crisp, and we'll all just have to roll with that.
2 replies 0 retweets 2 likes -
Replying to @livcomp @IntuitMachine
Jean Tardy points at the difference between technical design (outside-in) and 'existential design' (inside-out). The former assumes a fully structured universe that imposes functional structure on a piece of raw material, the latter a seed that conquers the outside chaos.
5 replies 0 retweets 4 likes -
Replying to @Plinz @IntuitMachine
Not sure I follow what a 'fully structured universe' is, but yeah, I begin and end with implementable systems. I get antsy playing with cognitive concepts unless I can at least somewhat explicitly imagine how I'd implement them.https://www.youtube.com/watch?v=DBZBQSBdcwY&t=100 …
1 reply 0 retweets 1 like -
Replying to @livcomp @IntuitMachine
When we engineer a system, we start with a deterministic substrate and a deterministic set of tools that we use to manipulate the substrate to add a function that interfaces with the existing ones. Living systems must carry all structure into the environment by themselves.
2 replies 0 retweets 0 likes -
Replying to @Plinz @IntuitMachine
I engineer without determinism all day long.
2 replies 0 retweets 0 likes -
Can you expound a bit more on this? How do you handle non-determinism and still be able to implement something?
1 reply 0 retweets 0 likes
He calls it 'best effort computing':https://www.youtube.com/watch?v=I4flQ8XdvJM …
-
-
I was hoping he would explain it with his current self.
1 reply 0 retweets 0 likes -
Replying to @IntuitMachine @Plinz
That video still holds up. Also perhaps the AAAI Q&A, with like: Q: But if determinism fails, doesn’t that mean chaos? A: [Not at all. When we shift to] best-effort hardware, software becomes obligated to look beyond efficiency and embrace robustness.. https://www.aaai.org/ocs/index.php/AAAI/AAAI16/paper/view/11987 …
1 reply 0 retweets 3 likes - 4 more replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.