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

The AI Misinformation Epidemic

approximatelycorrect.com

Post reply on HN
Print view View original

241–250 of 286 posts

Re: The AI Misinformation Epidemic

#241

Post by digi_owl »

arjie wrote: post #221
digi_owl wrote: post #201

Earlier quoted context omitted.

Very much so. More and more i feel that he PC and smartphone has just been about recreating the mainframe using smaller components. The laser may well have been the last really new thing to come about.

With enough work one can convince oneself that each 'discovery' or 'invention' is a trivial follow-up of existing technology. DNA sequencing is in the post-laser era. Can you build a convincing argument as to why it's not new but the laser is new?

the first laser was built in the 60s, DNA was isolated in 1878.

Re: The AI Misinformation Epidemic

#242

Post by kenjackson »

knieveltech wrote: post #240
kenjackson wrote: post #237

Earlier quoted context omitted.

But that's because most of the money and talent has focused on the digital realm. Progress is going where there is the best ROI.

Which is kind of the point. ROI is not a strong predictor of societal improvement.

It may be, but there isn't a real good metric for societal improvement. And it seems like progress would always have ebbs and flows as different technologies have new opportunities that are squeezed -- and then other technologies look plumper, so they are squeezed, and so on.

And that's fine. I don't think anyone is complaining today about timepieces not keeping even more accurate time. Clocks today (in whatever device they are embedded in), while not 100% accurate are pretty good. The ROI in improving them further just isn't there -- in part because the good they'd serve humanity isn't there.

Re: The AI Misinformation Epidemic

#243

Post by hasbroslasher »

SkyMarshal wrote: post #147
ExactoKnight wrote: post #26

The AI and Singularity hype irks me, because I'm genuinely in agreement with Peter Thiel's argument that technological progress is actually decelerating relative to how it was moving 60 years ago once you look beyond the advancements we have experienced in information technology and finance.

>because I'm genuinely in agreement with Peter Thiel's argument that technological progress is actually decelerating relative to how it was moving 60 years ago Is it possible we just covered most of the low-hanging fruit in the 1800s and 1900s, and now the only remaining discoveries that are really ground-breaking are the really difficult ones, hence the slower rate?

There's an (underrepresented) idea when looking at the history of tech called a Kondratiev_wave[0] that says global economic changes happen in a series of predictable peaks and valleys. If the recent trends of computing really took off in the late 1990's, we should expect the wave to stagnate between 2015 and 2030, and recede into stagnation and negative growth around 2030-2050.

I don't think this is foolproof, but it _is_ a tool and a perspective for analyzing history, and one that runs counter to the most common narrative of increasing acceleration in technological development. The physics way of saying this is that velocity is always positive (we're always going up or near-neutral) but the rate of acceleration isn't spectacularly greater than the new ones. If we really are talking about the rate of change of economic/tech growth (the rate of change of market size/capability), then we're talking about acceleration!

Anyway, this was a whole rant, but my personal take is that we tend to underestimate the amount of turbulence in history, leading us to believe that the past was a stable, peaceful and simpler time, while the present day is going nuts. This kind of thinking has fueled messianism, apocalyptic cults, futurism, dreams and nightmares for centuries, and will continue to if we don't blow ourselves up. We forget that even in places like Ancient Rome the rate of change in tech and human growth was still positive and rapid, that during the Dark Ages science and technology were still exploding in the Middle East, that the differences between 1650 and 1750 were vast in terms of human thought, that between 1850 and 1950 they were seemingly greater. There was no time of mythological stasis before the present day, and it's unlikely that there ever will be as long as we're still human.

[0] https://en.wikipedia.org/wiki/Kondratiev_wave

Re: The AI Misinformation Epidemic

#244

Post by abecedarius »

SkyMarshal wrote: post #147
ExactoKnight wrote: post #26

The AI and Singularity hype irks me, because I'm genuinely in agreement with Peter Thiel's argument that technological progress is actually decelerating relative to how it was moving 60 years ago once you look beyond the advancements we have experienced in information technology and finance.

>because I'm genuinely in agreement with Peter Thiel's argument that technological progress is actually decelerating relative to how it was moving 60 years ago Is it possible we just covered most of the low-hanging fruit in the 1800s and 1900s, and now the only remaining discoveries that are really ground-breaking are the really difficult ones, hence the slower rate?

It's possible, even plausible, but I think it's not so. My guess is that in the decades around 1900 America and western Europe were considerably more dynamic -- there was less overall resistance to innovation. To pick a tiny example, driver's licenses weren't required in all states until, iirc, the 1950s. Radio had decades of development before the FCC. Some doctors tested new treatments on themselves. Immigration from Europe to the U.S. was unlimited. I'm not promoting going back to those days -- I don't want to get into that argument -- but I'd like to reach better understanding of how much the growth slowdown since 1970ish is about the intrinsic difficulty of innovating vs. how much is about the way we collectively think and coordinate.

Here's some fruit that looks low-hanging today relative to its value: atomically precise manufacturing. http://www.foresight.org/roadmaps/Nanotech_Roadmap_2007_main...

Re: The AI Misinformation Epidemic

#245

Post by nostrademons »

visarga wrote: post #123
nostrademons wrote: post #69

Earlier quoted context omitted.

I think AI is accurate, but it's actually short for "aggregated intelligence". Most AI algorithms can't actually generate "intelligence" as we know it. Rather, they aggregate the intelligence of the humans who contributed to their training data, and then generalize it to situations where those humans are not present. If the humans are stupid (as we saw with Microsoft's Tay chatbot), the AI is going to be pretty stupi…

> Most AI AlphaGo would be an example. It was trained by self play.

Bootstrapped by supervised learning though: it was trained with the results of 30 million human games before the reinforcement learning kicked in.

Re: The AI Misinformation Epidemic

#246

Post by jfoutz »

cglouch wrote: post #220
jfoutz wrote: post #54

Earlier quoted context omitted.

don't options solve this? Buy a put at 50, watch it fall, then exercise. Doesn't work out, and you're only out the premium.

Options solve it in the sense that you limit your downside, but you can still very well lose your premium. Even if you're 100% convinced that $TSLA is overvalued, markets can remain irrational longer than you can remain solvent, and your puts may very well expire worthless.

Well, yes. But that's a different thing than the infinite risk of a naked short.

Re: The AI Misinformation Epidemic

#247

Post by nostrademons »

SCHiM wrote: post #94
nostrademons wrote: post #69

Earlier quoted context omitted.

I think AI is accurate, but it's actually short for "aggregated intelligence". Most AI algorithms can't actually generate "intelligence" as we know it. Rather, they aggregate the intelligence of the humans who contributed to their training data, and then generalize it to situations where those humans are not present. If the humans are stupid (as we saw with Microsoft's Tay chatbot), the AI is going to be pretty stupi…

Expert systems (which is what you're describing) are generally considered to belong in the AI category.

Not just expert systems: any supervised learning system also fits into this category (including neural nets), as do Markov chains, collaborative filtering, Bayesian inference, anything that involves counting actions that humans take.

Re: The AI Misinformation Epidemic

#248

Post by VHRanger »

gone35 wrote: post #215
VHRanger wrote: post #134

Earlier quoted context omitted.

If the paper gets traction and the model specification is indeed not robust, the first thing you'll see in a few months is something like "paper xyz: comment" driving holes in the methodology getting even more traction. Empirical microeconomics is fairly open about methodology flaws and critiques. Also, by the way, you see pretty much the same type of thing in a ton of fMRI neuroscience, medical and psychology studie…

Your response, and the ostensible fact my critique went completely over your head, probably shows you should definitely spend more time familiarizing yourself with the discipline, before going around defending it. You should start with 'calibration' in economics. No, it is not quite what you (seem to) think it is. No, it is not quite "the same type of thing" as p-hacking and low-powered studies in psychology. (Whose…

My point is that it's very easy to check "model calibration" (eg. "Plug values from outside data"). Just run the code with different values.

Because it's so easy to check those things (assuming the data is not proprietary or whatnot) I'd argue it's a much lesser problem than the "garden of forking paths" in lab experiments where it's much harder to test robustness of the result.

Moreover, I don't think anyone intelligent is foolish enough to take the coefficients in an econometric study literally; at least I would hope not.

Re: The AI Misinformation Epidemic

#249

Post by notahacker »

coldtea wrote: post #183
meri_dian wrote: post #154

Earlier quoted context omitted.

The fact that a relatively low income individual has access to a large number of goods and services which were once considered exclusive to the wealthy is a clear indication that technological progress has continued at a steady pace. While 'slowed down' or 'sped up' are difficult to quantify, the fact that 3 billion human beings have gained access to smartphone technology over the last decade seems to support the ide…

> The fact that a relatively low income individual has access to a large number of goods and services which were once considered exclusive to the wealthy is a clear indication that technological progress has continued at a steady pace. No, it's just a clear indication of market efficiencies and/or better engineering. It doesn't say anything about what the grandparent asked for: the rate and magnitude of new scientifi…

But half of the "world of atoms" things the grandparent listed were also market efficiencies or better engineering. Trading average speed for vast improvements in energy efficiency in aircraft in order to make flight accessible to the masses certainly involved vastly more progress of a technical and scientific nature in the late twentieth century than the cited early twentieth century example of wider replication of well-understood principles of plumbing to give the masses indoor toilets.

Re: The AI Misinformation Epidemic

#250

Post by mcguire »

CuriouslyC wrote: post #142
Fricken wrote: post #17

Welcome to the club. Now you know how Economists, healthcare professionals, skateboarders, basket weavers and anybody else whose domain knowledge runs deeper than the average joe's feels whenever their area of technical expertise becomes the subject du jour for the public at large.

On a deeper level, economists generally don't know what they're talking about. They create linear models of nonlinear systems, they assume Gaussian distribution when fat tails are common and their assumptions about rationality don't align with people's actual behavior. Trusting the models of economists to manage economies gives us the ability to trade a little bit of efficiency for a large amount of risk.

References:

Steve Keene, Debunking Economics. Ignore everything but the actual arguments.

Richard Thaler, Misbehaving.

Post reply on HN