Past Performance is Not Indicative of Future Results (2020) • Hacker News Board

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

#91

Post by 3gg » Sat, Jul 31, 2021, 6:33 PM UTC

salawat wrote: post #82

3gg wrote: post #76

Earlier quoted context omitted.

Excellent. One problem in my mind that I don't see discussed enough -- and also not in your other post -- is that there is a large divide between those who use the technology (the cops in this case) and those who supply it, and there is no accountability in any of the two groups when something goes wrong. Like you write in your other post, "the system works (according to an objective function which maximizes arrests.…

I second this. I spend a great deal of time digging through where we've positioned big data models to steer population scale behavior, and very infrequently do the implementers of the system ever stop to analyze the changes they are seeding or think beyond the first or second degree consequences once things take off. That is all part of engineering to me, so by definition, I think many in the field are in fact, out t…

Yes, thank you. Analyzing the effects of our technology should be part of the engineering process. The physicists back where I studied all go through a mandatory ethics class. Us software crowd, well...

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

#92

Post by esfandia » Sat, Jul 31, 2021, 6:35 PM UTC

shreyshnaccount wrote: post #33

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…

I'm in favor of changing the terminology from AI and ML to something along the lines of 'prediction model' so that the idea of machines 'thinking' is replaced with them 'predicting'. it's just easier for our mushy meat brains to think that AI and ML means that it'll lead to general AI or as I like to call it 'general purpose decision maker'. it's all about the language!

We already have "Pattern Recognition", not sure why it got absorbed by Machine Learning (the two terms seemed to co-exist with some overlap on what they covered), and then ML got absorbed by AI.

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

#93

Post by jmull » Sat, Jul 31, 2021, 6:38 PM UTC

nkozyra wrote: post #43

> It’s not sorcery, it’s “magic” – in the sense of being a parlor trick, something that seems baffling until you learn the underlying method, whereupon it becomes banal. I think part of the problem is the belief that human or animal intelligence is somehow more mystical. People who think like this will see an ML implementation solve a problem better and/or faster than a human and counter "well, it's just using statis…

> We're all simple machines.

Great. Prove it. Build the simple machine that acts as a human does. Should be simple, right?

Personally, I don't think there's any magic. But it's not "simple" either.

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

#94

Post by okareaman » Sat, Jul 31, 2021, 6:40 PM UTC

Trasmatta wrote: post #86

okareaman wrote: post #37

> But the idea that if we just get better at statistical inference, consciousness will fall out of it is wishful thinking. It’s a premise for an SF novel, not a plan for the future. My impression of Silicon Valley types like Ray Kurzweil in "The Age of Spiritual Machines" that if we wire up enough transistors somehow consciousness will somehow arise out of the material world. The somehow is not explained. Materialism…

If our brains are receivers to a field of consciousness, why would it be impossible to replicate one of those receivers with a machine? You also seem to have just kicked the can down the road. "Consciousness arises from a field somehow, and the brain acts as a receiver somehow. The somehow is not explained."

I didn't say I knew how. I said I believe materialism is a dead end, by which I mean I doubt the consciousness arises out of atoms configured as neurons. How those neurons receive a conscious field seems a more productive line of inquiry, but for some reason people resist this idea. Not sure why.

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

#95

Post by skmurphy » Sat, Jul 31, 2021, 6:51 PM UTC

3gg wrote: post #76

bhntr3 wrote: post #60

Earlier quoted context omitted.

> Are you at all close to this space? I am. > The example Cory puts on policing My most upvoted comment on this website was discussing this exact scenario. https://news.ycombinator.com/item?id=23655487 Could you perhaps clarify the generalization you're making about me and people like me so I can understand it?

Excellent. One problem in my mind that I don't see discussed enough -- and also not in your other post -- is that there is a large divide between those who use the technology (the cops in this case) and those who supply it, and there is no accountability in any of the two groups when something goes wrong. Like you write in your other post, "the system works (according to an objective function which maximizes arrests.…

   "Don't say that he's hypocritical
   Say rather that he's apolitical
   'Once the rockets are up, who cares where they come down?
   That's not my department!' says Wernher von Braun

   Some have harsh words for this man of renown
   But some think our attitude
   Should be one of gratitude
   Like the widows and cripples in old London town
   Who owe their large pensions to Wernher von Braun"

Tom Lehrer "Wernher von Braun"

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

#96

Post by Dylan16807 » Sat, Jul 31, 2021, 6:51 PM UTC

marcinzm wrote: post #75

foobiekr wrote: post #49

Earlier quoted context omitted.

The problem with this line of reasoning is that it can be used as a non-constructive counter to any observation about AI failure. It’s always more and more training data or errors in the training set. This really is a god-of-the-gaps answer to the concerns being raised.

No, my point is that if two systems show very similar classes of errors but at different thresholds with one trained on significantly more data than the more likely conclusion is that there isn't enough data in the other.

Don't most high-end machine learning solutions have more training data than a human could consume in a lifetime?

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

#97

Post by username90 » Sat, Jul 31, 2021, 6:52 PM UTC

marcinzm wrote: post #75

foobiekr wrote: post #49

Earlier quoted context omitted.

The problem with this line of reasoning is that it can be used as a non-constructive counter to any observation about AI failure. It’s always more and more training data or errors in the training set. This really is a god-of-the-gaps answer to the concerns being raised.

No, my point is that if two systems show very similar classes of errors but at different thresholds with one trained on significantly more data than the more likely conclusion is that there isn't enough data in the other.

They aren't very similar errors, ML solutions are equally accurate as humans in at a glance performance but longer and humans clearly wins. I'd say that the system is similar to humans in some ways, but humans have a system above that which is used to check if the results makes sense or not, that above system is completely lacking from modern ML theory and it doesn't seem to work like our neural net models at all (the brain isn't a neural net).

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

#98

Post by jstx1 » Sat, Jul 31, 2021, 6:52 PM UTC

esfandia wrote: post #92

shreyshnaccount wrote: post #33

Earlier quoted context omitted.

I'm in favor of changing the terminology from AI and ML to something along the lines of 'prediction model' so that the idea of machines 'thinking' is replaced with them 'predicting'. it's just easier for our mushy meat brains to think that AI and ML means that it'll lead to general AI or as I like to call it 'general purpose decision maker'. it's all about the language!

We already have "Pattern Recognition", not sure why it got absorbed by Machine Learning (the two terms seemed to co-exist with some overlap on what they covered), and then ML got absorbed by AI.

ML is still widely used and is much more common than AI as a term. So I wouldn't say that it has been absorbed by AI but their use sometimes overlaps depending on the target audience.

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

#99

Post by m0rphy » Sat, Jul 31, 2021, 7:00 PM UTC

ML or not, at the most fundamental level, classical computers simply do not possess the type of logic that's truly reflective of our reality. Its binary nature forces it to always resolve any single statement to either a true or false answer only.

A very simple example. If we ask our classical computer this question "are people currently supportive of COVID-19 vaccines?", then it would probably give us a straight answer of either a "yes" or "no" based on statistical inference of the percentage of total people who have received vaccinations at this point.

At its most fundamental level, classical computers just cannot comprehend a reality that could resolve that answer to both "Yes" and "No" in a single statement, which btw is possible in a quantum computing environment under its superposition state.

In our reality, some people who may not be fully supportive of the vaccines, but under special circumstances they may be forced to receive it because of workplace requirements, pressures from their loved ones, etc...

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

#100

Post by username90 » Sat, Jul 31, 2021, 7:07 PM UTC

Reimersholme wrote: post #25

swayvil wrote: post #20

I see no path from "observation" to "model" that does not involve an arbitrary (aesthetic? Nonrational, human-necessitating?) choice. This would suggest that "general" AI is impossible. ON THE OTHER HAND There is a variety of general AI, called an "optimizer". It starts with something better than a void. Maybe that's the path we should be looking at.

Well, human thinking relies on prior models/filters for understanding the world as well so that would invalidate us as having general intelligence too?

Human thinking includes building new models/filters for understanding the world, not just applying old ones. And that isn't used for learning, we do it all the time when solving any kind of challenging problem or even for simple problems like trying to recognize a face. Computer models might never compete with human performance unless they can learn how to solve a problem as it is solving it, because that is what humans do.