Community Archive

🧵 View Thread

🧵 Thread (16 tweets)

Placeholder
james, the giant peach@JimmyRis• over 6 years ago

It seems unfair to demand explainability from our AI brethren when we don't hold humans to the same standard

1 0
6/10/2019
Placeholder
james, the giant peach@JimmyRis• over 6 years ago
Replying to @JimmyRis

Explainable humans are boring, why demand explainable AI?

0 1
6/10/2019
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @JimmyRis

@JimmyRis Its easier to plant malicious behavior behind a black box that we trust and claim ignorance, vs planting a human spy.The former is also scalable. Having a piece of malware that does something based on some statistical inference would be nigh undetectable.

0 0
7/30/2022
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @ultimape

@JimmyRis This lack of trust and perception that an algorithm is lying/manipulating us is at the core here. Feeling like google's suggestion algorithm is trying to hide a search term. Or that algorithmic bias is intentional (or an artifact of) a corrupt system? Rational given unknowns.

0 0
7/30/2022
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @ultimape

@JimmyRis Imagine an AI doctor that prescribes a bunch of pills that end up killing a bunch of people. We might wanna know why the AI decided this was the best course of action. With a black box, we must necessarily assume it was subterfuge as one possibility. https://t.co/svdg8A0Llu

0 0
7/30/2022
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @ultimape

@JimmyRis Granted, there is room here for plausible deniability within the AI explanation, but we can't even begin to route out that kind of thing in a black-box model because we are only left scratching our heads.

0 0
7/30/2022
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @ultimape

@JimmyRis Even in contrived cases of unexplainable human actions (as given in this quote tweet ) we are still able to / desire to interrogate why a destructive behavior occurred. https://t.co/gmJDRARI8q

0 0
7/30/2022
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @ultimape

@JimmyRis If a human does something unexpected, we don't shrug and give up with "whoops, guess we'll never know".This is what we are faced with currently, and our imaginations run wild.

0 0
7/30/2022
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @ultimape

@JimmyRis I find it interesting that so many people have adopted this talking point about "we don't expect this of humans" when we do! We have entire court systems and systems of law based on interrogation & seeking explanation / justice.We should absolutely hold AI to this standard.

0 0
7/30/2022
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @ultimape

@JimmyRis I feel like someone is trying to manipulate the public by getting inside of our OODA loop by slipping in a framing bias as a psyop.But its not like I study propaganda or anything.

0 0
7/30/2022
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @ultimape

@JimmyRis It's one thing to frame this under a framework of unknowability and the nature of axioms of knowledge, but an entirely other thing to say "guess we can't know, better just put up with potential abuses by proxy agents ran by people in positions of power & control here".

0 0
7/30/2022
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @ultimape

@JimmyRis The latter is defeatist attitude and is a natural conclusion of this kind of framing. I don't accept that, so I reject the framing to subvert it.

0 0
7/30/2022
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @ultimape

@JimmyRis I feel its far too easy to get caught up in the dazel and amazing potential of these systems, and get trapped within the PR hype cycle. There is a lot of promise in it and I to am looking forward to the future with powerful Ai.

0 0
7/30/2022
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @ultimape

@JimmyRis But it would be folly to stick our head in the sand here.https://t.co/cTCf1Vh8Iw

Placeholder
UltimApe@ultimape• over 4 years ago

Demons are coming.https://t.co/crMsiPbM3F

1 0
0 0
7/30/2022
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @ultimape

@JimmyRis Daemons. Sleeper (computer) agents. Hidden control surfaces triggered out of systems of inference."Demons are coming" is a pun.https://t.co/5uOm9tpwVu https://t.co/zcmZLhX24V

Tweet image 1
Placeholder
UltimApe@ultimape• over 3 years ago

Demons are coming. https://t.co/yxIG7EiAgk

1 1
0 0
7/30/2022
Placeholder
UltimApe@ultimape• about 3 years ago
Replying to @ultimape

@JimmyRis But my paranoia runs deep. I wanna build my own chips.https://t.co/xi0RLy3vr7

Placeholder
UltimApe@ultimape• over 4 years ago

It came from inside the CPU!"In this talk we will discuss the methods of remote pwning of almost every Intel based system, manufactured since 2010 or later." https://t.co/Oy3Ndpx8Vc

1 0
1 0
7/30/2022