U.S. CitiesCity-Data Forum Index
Go Back   City-Data Forum > General Forums > Economics
 [Register]
Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
View detailed profile (Advanced) or search
site with Google Custom Search

Search Forums  (Advanced)
Reply Start New Thread
 
Old 05-22-2018, 04:33 PM
 
5,633 posts, read 4,237,577 times
Reputation: 10681

Advertisements

Quote:
Originally Posted by Storm Eagle View Post
So because one person was wrong that means everyone is wrong? The technology we have now is a lot more advanced then what we had back then.

Unfortunately it is not just one person who overestimated the effect of technology. It happens all the time with even otherwise smart individuals. Ever heard of Theranos?
Reply With Quote Quick reply to this message

 
Old 05-22-2018, 06:07 PM
 
25,044 posts, read 11,703,774 times
Reputation: 11730
Quote:
Originally Posted by jrkliny View Post
One of the leaders in the field of cognitive science, computers and artificial intelligence, a genius and Nobel laureate, has made the following prediction: "machines will be capable, within twenty years, of doing any work a man can do."

This prediction was made by Herbert A Simon in 1965. Oops.

Uh huh. in 1965 that was a foolish prediction to say the least. The human brain is estimated to be comparable to about 38 teraflops (on the way too low end) to 1 exoflop of processing power. in 1985 we reached 1.9 gigaflops with the Cray-2 supercoputer.


In 2018 I can rent 880 teraflops of ai processing power on the amazon cloud for $25/hr.


Its only been recently that we have reached this sort of processing power. And we are seeing rapid improvements in the AI field after decades of frustratingly slow progress. And while the most obvious are specific purpose AI, there is a LOT of progress on general AI as well. Its a mistake to think this is not going to occur.
Reply With Quote Quick reply to this message
 
Old 05-22-2018, 06:09 PM
 
25,044 posts, read 11,703,774 times
Reputation: 11730
Quote:
Originally Posted by jrkliny View Post
Unfortunately it is not just one person who overestimated the effect of technology. It happens all the time with even otherwise smart individuals. Ever heard of Theranos?

Theranos was a scam. HArdly something I would base your predictions on.


Heres the other side of the coin, moores law is a exponential curve. Seeing the long flat progress behind us really doesn't prepare us for that massive jump at the end.



And Quantum computers are seeing even faster progress.
Reply With Quote Quick reply to this message
 
Old 05-22-2018, 09:48 PM
 
5,633 posts, read 4,237,577 times
Reputation: 10681
Quote:
Originally Posted by greywar View Post
Theranos was a scam. HArdly something I would base your predictions on.

..... .
Sure it was a scam. That is the whole point. Even major corporations and major investors were sucked in. How I cannot imagine because any of tens of thousands of us who are familiar with the field of lab testing knew it was a scam.


The Theranos board was filled with a cast of very well known experts in their fields. They got sucked in. Walgreens was sucked in. A great many serious investors were sucked in to a total of close to a $billion. It happens all the time with tech. A good deal of the hype about tech is just a more watered down scam to attract investors. Look at one of the chief flim flam artists, Elon Musk. Sure there is some actual tech and who knows it might actually succeed but most of what comes from Musk is shear BS.


I have no idea where AI is going but I see no evidence of a massive tech change. AI has been around for a long, long time and processing power is never been the major limitation.
Reply With Quote Quick reply to this message
 
Old 05-22-2018, 10:01 PM
 
Location: Ruidoso, NM
5,171 posts, read 4,756,071 times
Reputation: 4206
Quote:
Originally Posted by jrkliny View Post
processing power is never been the major limitation.
Then what is?
Reply With Quote Quick reply to this message
 
Old 05-22-2018, 11:21 PM
 
5,633 posts, read 4,237,577 times
Reputation: 10681
Quote:
Originally Posted by rruff View Post
Then what is?
Programming.


Self driving cars are a good example. The software is partially based on programming that is written by humans and partly based on self learning based on neural networks and similar designs. Both are severely limited. Feedback is the big limitation for the so called self learning process. Look at what happens with us as we learn. We are prewired for managing in the world but it still takes unbelievable amounts of feedback for us to learn to even just maneuver in space and to recognize objects and to begin to shift through the useless data and process data of value. How do you provide feedback to a machine so it can start to learn even relatively simple concepts?


There are some other serious issues but they quickly get to complicated to discuss on a forum like this. Another simple limitation is fixing the programming and stopping the "learning". If you want a machine to do some tasks, you need to achieve the requisite programming. Then the learning needs to be stopped. The programming and function needs to be tested and then it can be replicated for use in additional units. To progress the software needs to continue to evolve and improve. That means letting a machine "learn" again. What was previously "learned" is now subject to change. The machine can easily start to make mistakes that it did not make before and all of the testing has to start from zero.


Actually self driving cars are a poor example. It is probably better to start with much simpler tasks to understand how to use and implement AI. The first example I saw was in the 70s. It was thought that a computer could understand pattern recognition and could learn through neural network techniques to identify and count the various types of white blood cells. At the minimum it seemed that a computer could identify the majority of white cells and flag others for human review. Forty years later we still have not gotten to that point. Considering the millions of tedious manual 100 cell counts that are performed every year, there is considerable economic importance to this task. So you tell me what is it that does not seem to work?
Reply With Quote Quick reply to this message
 
Old 05-22-2018, 11:37 PM
 
Location: Ruidoso, NM
5,171 posts, read 4,756,071 times
Reputation: 4206
Quote:
Originally Posted by jrkliny View Post
So you tell me what is it that does not seem to work?
They seem to be doing it...: https://www.google.com/search?q=neur...ient=firefox-b
Reply With Quote Quick reply to this message
 
Old 05-23-2018, 07:24 AM
 
5,633 posts, read 4,237,577 times
Reputation: 10681
Quote:
Originally Posted by rruff View Post
Trying and doing are different. There is not one FDA approved system in routine use. The last experimental system I saw made so many errors; i.e., misclassifications, that it was useless. I would think someone would be close. This is baby work in comparison with all the wild talk about the future of AI.
Reply With Quote Quick reply to this message
 
Old 05-23-2018, 08:57 AM
 
12,405 posts, read 9,251,495 times
Reputation: 8869
Quote:
Originally Posted by jrkliny View Post
Programming.


Self driving cars are a good example. The software is partially based on programming that is written by humans and partly based on self learning based on neural networks and similar designs. Both are severely limited. Feedback is the big limitation for the so called self learning process. Look at what happens with us as we learn. We are prewired for managing in the world but it still takes unbelievable amounts of feedback for us to learn to even just maneuver in space and to recognize objects and to begin to shift through the useless data and process data of value. How do you provide feedback to a machine so it can start to learn even relatively simple concepts?


There are some other serious issues but they quickly get to complicated to discuss on a forum like this. Another simple limitation is fixing the programming and stopping the "learning". If you want a machine to do some tasks, you need to achieve the requisite programming. Then the learning needs to be stopped. The programming and function needs to be tested and then it can be replicated for use in additional units. To progress the software needs to continue to evolve and improve. That means letting a machine "learn" again. What was previously "learned" is now subject to change. The machine can easily start to make mistakes that it did not make before and all of the testing has to start from zero.


Actually self driving cars are a poor example. It is probably better to start with much simpler tasks to understand how to use and implement AI. The first example I saw was in the 70s. It was thought that a computer could understand pattern recognition and could learn through neural network techniques to identify and count the various types of white blood cells. At the minimum it seemed that a computer could identify the majority of white cells and flag others for human review. Forty years later we still have not gotten to that point. Considering the millions of tedious manual 100 cell counts that are performed every year, there is considerable economic importance to this task. So you tell me what is it that does not seem to work?
Very well put! Without good algorithms, raw processing power does little. This is what so many of Ray Kurzweil's followers don't get!
Reply With Quote Quick reply to this message
 
Old 05-23-2018, 11:58 AM
 
Location: Ruidoso, NM
5,171 posts, read 4,756,071 times
Reputation: 4206
Quote:
Originally Posted by ncole1 View Post
Very well put! Without good algorithms, raw processing power does little. This is what so many of Ray Kurzweil's followers don't get!
Does my algorithm have a mental-health problem? | Big Think

I think everyone who works in AI gets this very well.

There is still only so much that you can teach a cockroach or mouse, vs a human or superhuman. Give it a little time...
Reply With Quote Quick reply to this message
Please register to post and access all features of our very popular forum. It is free and quick. Over $68,000 in prizes has already been given out to active posters on our forum. Additional giveaways are planned.

Detailed information about all U.S. cities, counties, and zip codes on our site: City-data.com.


Reply
Please update this thread with any new information or opinions. This open thread is still read by thousands of people, so we encourage all additional points of view.

Quick Reply
Message:

Over $104,000 in prizes was already given out to active posters on our forum and additional giveaways are planned!

Go Back   City-Data Forum > General Forums > Economics
Follow City-Data.com founder on our Forum or

All times are GMT -6.

2005-2018, Advameg, Inc.

City-Data.com - Archive 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35 - Top