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

231–240 of 286 posts

Re: The AI Misinformation Epidemic

#231

Post by claytonjy »

jacquesm wrote: post #104

A lot of start-ups are hyping themselves as deep-learning or something close to that in order to make themselves appear 'hot'. Recently I looked at a company that had absolutely nothing to do with deep-learning or even any kind of machine learning whatsoever and that still managed to sprinkle the various buzz-words with great regularity throughout their investor targeted docs. What I don't get about this behavior is…

Even in the midwest I've seen this a lot in the last year or two. A couple months ago I was suspicious about the CEO of a startup's claims of "machine learning", so I pressed him on it; turns out it was much less ambitious than his pitch made it sound. On the part of that CEO, he was probably making a smart pitch because most investors, _especially_ around here, aren't going to know enough to call this out, and are going to fully believe the company pitching has a magical ML-black-box giving them a significant edge over the competition.

Re: The AI Misinformation Epidemic

#232

Post by cr0sh »

graycat wrote: post #212
Quarrelsome wrote: post #80

Earlier quoted context omitted.

You give your work a catchy name because you want other people to help develop it, via attention, promotion, word of mouth, etc so other people can find it and fund it. To do so you speak in their terms and their terms are catchy one-liners. Don't be bitter about it, some people just don't have the level of concentration to deal with our specifics.

I can accept Newton's second law of motion and Einstein's general relativity, but so far I'm less accepting of deep learning for finding some coefficients for fitting some data or neural networks for fitting some sigmoid functions to some data. Apparently the sigmoid functions have some nice theoretical properties and can be useful in approximating some complicated data; e.g., IIRC D. Bertsekas at MIT used such fitti…

I am not an expert in this field.

From what I currently understand in the CNN/deep learning arena, sigmoids are avoided in favor of RELU, due to a number of reasons, chiefly that of avoiding gradient fade during backprop.

I'm surprised that you don't mention this (or don't seem to be aware of it), though? Is this intentional, or is there another reason?

Re: The AI Misinformation Epidemic

#233

Post by mcguire »

puranjay wrote: post #98
Banthum wrote: post #28

Earlier quoted context omitted.

True that. 1880-1950 (70 years) took us from train-and-telegraph to jet-and-atom bomb. Horse and carriage was the top-tier tech for personal transportation in 1880, and they were just starting to understand that diseases are often caused by these little blobby creatures called germs. 1950-2020 (70 years) will take us much less distance. Almost everything we're developing seems to be about information processing. If y…

A flying car would be nearly magical for someone from 1880. But a smartphone to someone from 1950? It would seem pretty fancy, but not impossible to imagine. The guy from 1950 already knows the basic tech behind it - it has a screen, it uses wireless data, and is powered by computers.

Dick Tracy's wrist-TV, for example.

Re: The AI Misinformation Epidemic

#234

Post by TheOtherHobbes »

graeham wrote: post #149
ExactoKnight wrote: post #87

Earlier quoted context omitted.

Almost everything you listed there already started becoming common by the 1960's. Step back and forget about informational technology. Look at the world of atoms, not at the world of bits. Since the decommissioning of the Concorde, our fastest commercial means of transportation has actually been getting slower , not faster. Man hasn't reached farther out in space than the moon missions of the 1960s. The first half th…

Oil prices are essentially capped at $60/barrel for the foreseeable future due to unconventional extraction techniques -> was not long ago oil production was thought to have peaked. Solar power can be economical without subsidy. Robotic and minimally invasive surgery. Battery cost becoming viable for mass-market cars. Gene sequencing. Sensors for pennies. New materials (cheaper carbon fibre, metal alloys, semiconduct…

No, because you're failing to distinguish between game-changer tech - transistors, DNA sequencing, operating systems, powered flight, all as classes of original and unexpected inventions - and refinement tech, which is made of game-changer inventions made smaller, cheaper, and more widely avaialable.

Game changer tech changes what can be imagined. Refinement tech changes what can be bought by consumers.

There's been plenty of refinement over the last few decades, but not nearly as much original game changer invention as in the previous decades.

Re: The AI Misinformation Epidemic

#235

Post by mcguire »

iwintermute wrote: post #171
notahacker wrote: post #126

Earlier quoted context omitted.

Popular economics and popular "AI" as described in the article both seem like situations where everybody with an idea to push cites popularisation of research conclusions without understanding the limitations to the models and people with actual expertise in the field are often happy to play along with these hyperbolic, caveat-free popularisations because it helps their end goals. Sure, the economics profession and i…

Wasn't it the case with the first AI winter? When the field was overhyped, oversold and then imploded?

And the second.

Re: The AI Misinformation Epidemic

#236

Post by kenjackson »

romaniv wrote: post #225
deepnotderp wrote: post #218

Earlier quoted context omitted.

Making things smaller and power efficient IS invention, it's what has driven so much tech progress. Even supercomputers are a power problem now...

You probably should watch that video. This is one of its core subject. Kay makes a valuable distinction between invention and innovation, and what you describe falls into the "innovation" bucket. It's incremental. Yes, our cellphones are more powerful than mainframes of the old, but do we utilize them to the same extent?

But aren't lasers incremental too?

Re: The AI Misinformation Epidemic

#237

Post by kenjackson »

coldtea wrote: post #140
notahacker wrote: post #92

Earlier quoted context omitted.

Not sure I'd consider stuff like the vast majority of human knowledge sitting in my pocket to be less useful than the dishwasher because it has fewer moving parts or most people can afford to fly to be less of an achievement than the megarich can fly 50% faster . Recent decades might a disappointment compared with the millenialist interpretation of exponential curves, but that's only the same as the disappointment a…

> Not sure I'd consider stuff like the vast majority of human knowledge sitting in my pocket to be less useful than the dishwasher because it has fewer moving parts or most people can afford to fly to be less of an achievement than the megarich can fly 50% faster. Not sure anybody told you to consider that. What they did ask you to consider is that technological progress, outside of the digital realm, has been slowed…

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.

Re: The AI Misinformation Epidemic

#238

Post by kenjackson »

AndrewKemendo wrote: post #227
erkkie wrote: post #77

Earlier quoted context omitted.

Why should people be able to tell the difference vs the outcome of those products, unless someone (investors?) is specifically just buying into the AI hype itself? (in which case, this is nothing different from when everything was "social")

Differentiation. Most off the shelf ML tools are not that powerful - and they don't make a difference in how you differentiate as a business. If you have a ML team and know how to build models from scratch with novel data, then you can iterate on your model and expand, something you can't do if you don't own the model.

You should show your differentiation by differentiating. But in many cases the ML differentiation isn't the same as differentiating the value prop of the product.

Re: The AI Misinformation Epidemic

#239

Post by deepnotderp »

romaniv wrote: post #225
deepnotderp wrote: post #218

Earlier quoted context omitted.

Making things smaller and power efficient IS invention, it's what has driven so much tech progress. Even supercomputers are a power problem now...

You probably should watch that video. This is one of its core subject. Kay makes a valuable distinction between invention and innovation, and what you describe falls into the "innovation" bucket. It's incremental. Yes, our cellphones are more powerful than mainframes of the old, but do we utilize them to the same extent?

So are FinFETs, Quantum Wells, EUV lithography, compact linear accelerators and all the devices necessary for shrinking just "innovation"?

What's "invention then"?

Re: The AI Misinformation Epidemic

#240

Post by knieveltech »

kenjackson wrote: post #237
coldtea wrote: post #140

Earlier quoted context omitted.

> Not sure I'd consider stuff like the vast majority of human knowledge sitting in my pocket to be less useful than the dishwasher because it has fewer moving parts or most people can afford to fly to be less of an achievement than the megarich can fly 50% faster. Not sure anybody told you to consider that. What they did ask you to consider is that technological progress, outside of the digital realm, has been slowed…

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.
Post reply on HN