Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Past Performance is Not Indicative of Future Results (2020)

locusmag.com

Post reply on HN
Print view View original

211–220 of 285 posts

Re: Past Performance is Not Indicative of Future Results (2020)

#211

Post by dane-pgp »

username90 wrote: post #198
dane-pgp wrote: post #152

Earlier quoted context omitted.

By studying the atoms configured as neurons, we've managed to develop machines that can learn to play board games and Atari games better than humans, and can write prose and poetry at a convincingly human level. Those skills may not require consciousness, but it's not clear that these machines would be more useful if they could "receive a conscious field". Do you think that animals receive a conscious field? Could we…

Start that argument once we can model insect brains. Mouse brains aren't even on the horizon of what we can do.

> Start that argument once we can model insect brains.

I'm not sure what level of modelling you'd accept, but we appear to be close:

https://pubmed.ncbi.nlm.nih.gov/32880371/

> Mouse brains aren't even on the horizon of what we can do.

I would say that they are "on the horizon", given that the mouse brain connectome has already been published:

https://advances.sciencemag.org/content/6/51/eabb7187

Re: Past Performance is Not Indicative of Future Results (2020)

#212

Post by greggman3 »

dtech wrote: post #153
gonehome wrote: post #135

Earlier quoted context omitted.

I'm not sure I buy that - biology is often messier because of nature related constraints, it gets optimized for other things (energy, head size, etc.) The way a plane flies is quite different than the way a bird flies in complexity - they share an underlying mechanism, but planes don't need to flap wings. It's possible that scaling up does lead to generality and we've seen hints of that. - https://deepmind.com/blog/a…

> It’s hard to predict timelines for this kind of thing, and people are notoriously bad at it. Few would have predicted the results we’re seeing today in 2010. What would you expect to see in the years leading up to AGI? Does what we’re seeing look like failure? Few have predicted a reasonably-capable text-writing engine or automatic video face replacement, but many have predicted self-driving cars would have been re…

self-driving cars are available to consumers now. Search for FSD on youtube and see all the consumers using their self driving cars.

Or, watch the latest Veritasium

https://www.youtube.com/watch?v=yjztvddhZmI

Re: Past Performance is Not Indicative of Future Results (2020)

#213

Post by username90 »

dane-pgp wrote: post #211
username90 wrote: post #198

Earlier quoted context omitted.

Start that argument once we can model insect brains. Mouse brains aren't even on the horizon of what we can do.

> Start that argument once we can model insect brains. I'm not sure what level of modelling you'd accept, but we appear to be close: https://pubmed.ncbi.nlm.nih.gov/32880371/ > Mouse brains aren't even on the horizon of what we can do. I would say that they are "on the horizon", given that the mouse brain connectome has already been published: https://advances.sciencemag.org/content/6/51/eabb7187

They have mapped out the synapse structure, there is no evidence that synapse structure is enough to actually run the brain. If you can run that brain and show it is an accurate representation of a flea brain then you'd have something, but until then I'll believe that the neurons do way more than ML researchers hope they do.

Re: Past Performance is Not Indicative of Future Results (2020)

#214

Post by catears »

ALittleLight wrote: post #210

I don't get why this article conflates machine learning progress and racism. Machine learning is not inherently racist though it may be implemented, intentionally or unintentionally, to produce racist results. It's much easier to solve racism in machine learning models though then in humans and easier to test to confirm that you have corrected it.

There are a whole slew of "advanced AI programs" out there that tells managers who to fire and who to keep, tells judges if someone should go to jail or not, etc.

There are a lot of systems out there where peoples lives are changed forever "because the machine said so".

I'd argue that since machine learning learns only from it's data (produced by it's human creator), it becomes a great tool for baking in unconcious biases in a completely opaque system and amplifying those biases.

Much easier to tell if a human seems to be biased than if the dataset fed to an AI algorithm is biased.

Re: Past Performance is Not Indicative of Future Results (2020)

#215

Post by rapferreira »

epgui wrote: post #179
vijucat wrote: post #111

> Let’s talk about what machine learning is...it analyzes training data to uncover correlations be­tween different phenomena. The author seems to have missed or excluded reinforcement learning and planning algorithms in this definition. My criticism of AI criticism in general is that no one admits that at the root of it, we do not understand thinking (or "consciousness"). We are merely the "recipient" or enjoyer of t…

I believe we do understand, broadly speaking, thinking and consciousness. There remains a lot more to learn, as in anything in science... IMO the main difficulty is that humans have terrible self-awareness or self-insight. We want to believe we're special, we want to believe we're intelligent, we want to believe we're different than machines. We're in denial about that. Our brains aren't any more special than compute…

Are you kidding? Go read Hegel's phenomonology of mind and get back to me on that

Re: Past Performance is Not Indicative of Future Results (2020)

#216

Post by hirako2000 »

emrah wrote: post #205
jstx1 wrote: post #16

> I am an AI skeptic. I am baffled by anyone who isn’t. I don’t see any path from continuous improvements to the (admittedly impressive) ‘machine learning’ field that leads to a general AI - I share the skepticism towards any progress towards 'general AI' - I don't think that we're remotely close or even on the right path in any way. - That doesn't make me a skeptic towards the current state of machine learning thoug…

> It doesn't need to solve all of humanity's problems to be a great tool. As a side note, I'd like to say humanity's own intelligence is actually able to come up with solutions to its problems, we don't need AGI for that. Humanity is unable to implement those solutions for reasons beyond technical. How an AGI would get over those hurdles I have no idea

Humanity has been able to introduce at last as many problems as it managed to solve. The big question is what is AI/ML trying to accomplish.

Re: Past Performance is Not Indicative of Future Results (2020)

#217

Post by hirako2000 »

username90 wrote: post #208
blamestross wrote: post #202

Earlier quoted context omitted.

https://en.m.wikipedia.org/wiki/AnimatLab We totally do emulate organisms on that scale. The real challenge is simulating the sensory inputs and the feedback loop between the outputs, the environment as the body acts, then new inputs. Disembodied simulations of nerual networks don't work. They are part of a body, an environment, and all the feedback loops that come with it. It sounds like you really just want to see…

You don't need sensors, you just need a virtual room. > We totally do emulate organisms on that scale. The There is no evidence those emulations actually emulates those organisms. They just built a neural net in the same structure and assumes the cells doesn't matter. But cells are really smart and can navigate environments on their own, they are intelligent beings in their own right, and building a flea using a thou…

Also note, organisms not only navigate the environment, they interact with it, handle their own capture and consumption of energy from it,and reproduce. Autonomously.

Re: Past Performance is Not Indicative of Future Results (2020)

#218

Post by radu_floricica »

akomtu wrote: post #137
radu_floricica wrote: post #14

I think he's doing a bit of bait and switch there. Knowing reliably whether arrests are genuinely racist or if winks are flirtatious is superhuman intelligence. > But the idea that if we just get better at statistical inference, consciousness will fall out of it is wishful thinking. I'm a mostly disinterested spectator in current AI research, and even I know that it's not all about that. Just google "AI alignment" fo…

I think the definition of racism in this context can be simple. If the rate of false positives for blacks is significantly higher than the average across the nation, then it's racism. Significantly higher can mean "one stddev higher".

Exactly my point - I happen to disagree. We're not going to have AI come and tell us the Truth, at least not from day one. They'll help, but not with _this_ kind of questions. This is why I commented, because they seem specifically chosen to be some of the most complicated questions we face.

(On topic: blacks could commit more crime then average. They could do it because of systemic racism, but in this case the _arrests_ are not racist.)

Re: Past Performance is Not Indicative of Future Results (2020)

#219

Post by ALittleLight »

catears wrote: post #214
ALittleLight wrote: post #210

I don't get why this article conflates machine learning progress and racism. Machine learning is not inherently racist though it may be implemented, intentionally or unintentionally, to produce racist results. It's much easier to solve racism in machine learning models though then in humans and easier to test to confirm that you have corrected it.

There are a whole slew of "advanced AI programs" out there that tells managers who to fire and who to keep, tells judges if someone should go to jail or not, etc. There are a lot of systems out there where peoples lives are changed forever "because the machine said so". I'd argue that since machine learning learns only from it's data (produced by it's human creator), it becomes a great tool for baking in unconcious b…

I don't agree with that. If someone says "The AI is biased" we can just look at the data it was trained on or come up with a set of test cases to establish the bias. We can update the training data or method to remove the bias too.

If someone says a person is biased they can just say "No I'm not" and then how do we tell? We can't really make a person do a million judgments and find statistical evidence. We can't dive into how the possibly biased person came to have or not have biases.

Re: Past Performance is Not Indicative of Future Results (2020)

#220

Post by feanaro »

username90 wrote: post #208
blamestross wrote: post #202

Earlier quoted context omitted.

https://en.m.wikipedia.org/wiki/AnimatLab We totally do emulate organisms on that scale. The real challenge is simulating the sensory inputs and the feedback loop between the outputs, the environment as the body acts, then new inputs. Disembodied simulations of nerual networks don't work. They are part of a body, an environment, and all the feedback loops that come with it. It sounds like you really just want to see…

You don't need sensors, you just need a virtual room. > We totally do emulate organisms on that scale. The There is no evidence those emulations actually emulates those organisms. They just built a neural net in the same structure and assumes the cells doesn't matter. But cells are really smart and can navigate environments on their own, they are intelligent beings in their own right, and building a flea using a thou…

Have you seen these C. elegans emulations in a robot body?

https://www.youtube.com/watch?v=YWQnzylhgHc

https://www.youtube.com/watch?v=xu_oYLmPX9U

Post reply on HN