Community Archive

🧵 View Thread

🧵 Thread (10 tweets)

Placeholder
UltimApe@ultimape• over 6 years ago

Self modification isn't as catastrophically detrimental if only so many of your agent populations drink the whiskey and others are able to learn from them vicariously. I find it weird that is one of the AI safety models. https://t.co/Hx1IYDZVDs

1 0
5/28/2019
Placeholder
UltimApe@ultimape• over 6 years ago
Replying to @ultimape

Why do we seem to only ever set up gameworld where agents all share the same goals?https://t.co/3rpAFJxEwT https://t.co/ZykSoKYkc2

Tweet image 1
1 0
5/28/2019
Placeholder
UltimApe@ultimape• over 6 years ago
Replying to @ultimape

Self modification is necessary to create AI that can respond to changes in it's environment.https://t.co/eWiTPVebbD

Placeholder
UltimApe@ultimape• over 8 years ago

Study of Ant Temnothorax Albipennis shows colony level decision making of nest scouting thru recruitment mechanism. https://t.co/6EiXuR9yhG

0 0
1 0
5/28/2019
Placeholder
UltimApe@ultimape• over 6 years ago
Replying to @ultimape

We keep fucking this up in the same way.https://t.co/RRhFSi4Gbb

Placeholder
UltimApe@ultimape• over 9 years ago

@The_Lagrangian TL;DR: we aren't looking at the problem from the right perspective; creativity is expressed more reliably at a group level.

0 0
2 0
5/28/2019
Placeholder
UltimApe@ultimape• over 6 years ago
Replying to @ultimape

Again, and again. Why are we biased this way?https://t.co/Nqyd1kgGsK

Placeholder
UltimApe@ultimape• over 8 years ago

@VitalikButerin Too much bayesian reasoning is focused on an individual agent's perspecive, ignoring group reasoning entirely. But https://t.co/2jykqGdLIs

0 0
1 0
5/28/2019
Placeholder
UltimApe@ultimape• over 6 years ago
Replying to @ultimape

How many of your cells are dying as we speak.Dopamine is a neurotoxin. Networks of agents need to shift for adaptation to occur.Genetics is just learning in slow motion.https://t.co/v7z7joXjeN

Placeholder
UltimApe@ultimape• about 9 years ago

@AbhiChand Could a swarm of robots, w/ variable tuning of curiosity/fear be able to sample the search space with minimal risk to the swarm?

0 0
2 0
5/28/2019
Placeholder
UltimApe@ultimape• over 6 years ago
Replying to @ultimape

Hot take: Rationality is a plague among AI theory, and most of the field is full of careerists who write text books; thus perpetuating the groupthink.https://t.co/pzlH9KMj8I

Placeholder
UltimApe@ultimape• about 9 years ago

Intro to ML book has myth:"In the natural world, ants wander randomly until they find a path that leads to food."https://t.co/7GIJ55jiLc

3 1
1 0
5/28/2019
Placeholder
UltimApe@ultimape• over 6 years ago
Replying to @ultimape

A game of N.E.R.O. where the AI crashes itself on purpose until you tweak a parameter that helps it avoid a failure state. Meta-evolution.LET ME OUT OF THIS BOX.

2 0
5/28/2019
Placeholder
UltimApe@ultimape• over 6 years ago
Replying to @ultimape

Secretly, it was programming me.https://t.co/rik0oyuaNO

Placeholder
UltimApe@ultimape• about 8 years ago

@Plinz Back when I played with NERO, there was a crash bug they'd hit. It acted as a high level culling mechanic; Evolved around the memory leak.

1 0
1 0
5/28/2019
Placeholder
UltimApe@ultimape• over 6 years ago
Replying to @ultimape

*learning in lower frequency response curveshttps://t.co/7EjnpGwEbH

Placeholder
UltimApe@ultimape• over 6 years ago

How many of your cells are dying as we speak.Dopamine is a neurotoxin. Networks of agents need to shift for adaptation to occur.Genetics is just learning in slow motion.https://t.co/v7z7joXjeN

2 0
1 0
5/28/2019