Physicists set to revolutionize big data, AI

[EnumaElish]

When I opened up the article https://www.wired.com/2017/01/move-coders-physicists-will-soon-rule-silicon-valley/ I expected to see quantum computing as the next field physicists are to revolutionize. I was surprised to see it was data management and machine learning. I am happy for physicists. Also, I am a little puzzled as to the physics' edge on this over mathematics, or in fact most engineering fields.

 

[Stavros Kiri]

I would say that is somewhat expected, because Electronics, Solid State physics, Semi-conductor theory etc. (+materials, partially) are all parts (chapters) of physics. Some say "it all starts with physics!". I don't want to be that general, but as physicists we are in fact taught all the basics (at least) for what is needed in the domain in question (including a lot of math, programming etc.) ...

In addition, physicists tend to be multilateral and multidextrous, but one would say so other scientists too.

 

[Aufbauwerk 2045]

I stopped reading Wired a few years ago. I got tired of all the hype and oversimplifications. I don't rely at all on their predictions.

A headline like "move over coders, physicists will soon rule Silicon Valley" is good for attracting attention, which is its purpose. It certainly attracts coders, who are a huge part of the Wired readership, and already have lots of reasons to worry about their jobs. Fear is good bait for readers. It also appeals to the physicists, who listened as I did to numerous arrogant physics professors saying that everything is physics, and that we could easily do anything better than the specialists, if we did not have better things to do.

Perhaps the next headline will be "move over physicists, soon AI will rule physics." Actually, that may be true one day, but as someone working along those lines I know it's not happening "soon." Eventually AI will rule coding, physics, and just about anything else that involves intellectual work. By that time, we won't need magazines like Wired because we will get our information from a real-world Colossus or Alphaville computer.

 

[StatGuy2000]

I had posted a thread about this exact topic earlier.

https://www.physicsforums.com/threads/wired-article-physicists-will-soon-rule-silicon-valley.900681/

 

[StatGuy2000]

Eventually AI will rule coding, physics, and just about anything else that involves intellectual work. By that time, we won't need magazines like Wired because we will get our information from a real-world Colossus or Alphaville computer.

I have my doubts that day will ever arrive, in the sense of AI "ruling" anything.

 

[Aufbauwerk 2045]

I have my doubts that day will ever arrive, in the sense of AI "ruling" anything.

I mean "rule" in the sense that AI now rules chess, and it's apparently making great progress in Go. But hopefully we will never be stupid enough to turn our lives over to AI. I certainly do not want to be ruled by a computer. But I do want the most powerful AI we can imagine as my servant.

 

[Stavros Kiri]

Eventually AI will rule coding, physics, and just about anything else that involves intellectual work. By that time, we won't need magazines like Wired because we will get our information from a real-world Colossus or Alphaville computer.

If it happened + beyond ..., even our brains (and not just our "things") could be all hooked up on line in one big net (internet of brains and things ...), sharing info etc., and directed/standarized by a huge Artificial Inteligence server, or a huge brain (God's ?), depending on which one will prevail ...

May be that's the future of humanity (if the latter is to function properly), but ... way into the future! ...

 

[Aufbauwerk 2045]

If it happened + beyond ..., even our brains (and not just our "things") could be all hooked up on line in one big net (internet of brains and things ...), sharing info etc., and directed/standarized by a huge Artificial Inteligence server, or a huge brain (God's ?), depending on which one will prevail ...

May be that's the future of humanity (if the latter is to function properly), but ... way into the future! ...

I refuse to be assimilated into the Borg. Although I admit a sort of technological "Vulcan mind meld" could be interesting.

I was fascinated by the movie Brainstorm. We could make a recording of our experience and then someone else could play the tape and experience what we did. This goes way beyond virtual reality as it is today.

 

[Stavros Kiri]

I refuse to be assimilated into the Borg. Although I admit a sort of technological "Vulcan mind meld" could be interesting.

I was fascinated by the movie Brainstorm. We could make a recording of our ...

Unless the participation in the net would be voluntary, at will, and with a "switch", just like we log in and out of the net today, but with the help of the central guide (AI or Brain) we would be taught correct behavior and to be and act reasonable at all times! ...

I like and believe too in the potential (evolving) powers of the brain, as well as AI ...

 

[Zarqon]

But hopefully we will never be stupid enough to turn our lives over to AI.

heh, some current events suggest that we are just that stupid...

On a more serious note though, I believe it's not a matter of stupidity, I think it is inevitable that we put our decisions at the hand of an AI.

Consider for example any company making business decisions; we know that humans are fairly bad at rational decisions due to a large amount of psychological conditions and limited potential of keeping everything relevant to the bigger picture in your mind at the same time. An AI could likely do this much better, and if some companies start using AI's for decisions the rest will be forced to follow unless we ban them. The same goes for military decisions/strategies in conflicts, once AI's become better at it you have no choice but to turn your lives over to their rule or the other side who does has an increased chance to win with heavy consequences.

 

[Stavros Kiri]

On a more serious note though, I believe it's not a matter of stupidity, I think it is inevitable that we put our decisions at the hand of an AI.

But there is always a risk that the process may loose control over humans (just like in some movies). Especially if we are talking about Projective Networks and self-evolving learning computers and AI.

 

[StatGuy2000]

heh, some current events suggest that we are just that stupid...

On a more serious note though, I believe it's not a matter of stupidity, I think it is inevitable that we put our decisions at the hand of an AI.

Consider for example any company making business decisions; we know that humans are fairly bad at rational decisions due to a large amount of psychological conditions and limited potential of keeping everything relevant to the bigger picture in your mind at the same time. An AI could likely do this much better, and if some companies start using AI's for decisions the rest will be forced to follow unless we ban them. The same goes for military decisions/strategies in conflicts, once AI's become better at it you have no choice but to turn your lives over to their rule or the other side who does has an increased chance to win with heavy consequences.

Humans may be bad individually at rational decisions due to cognitive limitations (we are boundedly rational -- bounded rationality is a concept developed by political economist and AI researcher Herbert Simon back in the 1950s), but it is also the case that a diverse group of humans (diverse as being those with a diverse ethnic, religious, gender, social and career backgrounds) have been known to make more effective decisions. There is a research group led by economist and complexity researcher Scott Page at the University of Michigan that has examined this phenomenon.

http://sites.lsa.umich.edu/scottepage/research-2/diversity-research/

At any rate, the question is whether an AI would in fact make decisions that is "better", in the sense of making decisions that incorporate or are aligned with human values. Berkeley computer scientist and AI researcher Stuart Russell, among others, has sought to address this very concern in a number of different articles and talks, including this article he has written for Scientific American.

https://people.eecs.berkeley.edu/~russell/papers/sciam16-supersmart.pdf

You can see other articles/talks/videos about this from Russell on his website.

https://people.eecs.berkeley.edu/~russell/

 

[dkotschessaa]

When I opened up the article https://www.wired.com/2017/01/move-coders-physicists-will-soon-rule-silicon-valley/ I expected to see quantum computing as the next field physicists are to revolutionize. I was surprised to see it was data management and machine learning. I am happy for physicists. Also, I am a little puzzled as to the physics' edge on this over mathematics, or in fact most engineering fields.

Most mathematicians I know would look at a big messy data set the way my cat looks at canned cat food in pate form. I'm not too surprised.-Dave K

 

[EnumaElish]

If the political polls of 2016 are any indication, "big data" is in dire need for a new approach IMO.

 

[dkotschessaa]

If the political polls of 2016 are any indication, "big data" is in dire need for a new approach IMO.

I think they are more of an indication of the public's failure to understand probabilities, and a call for general numeracy.

-Dave K

 

[StatGuy2000]

If the political polls of 2016 are any indication, "big data" is in dire need for a new approach IMO.

This is off-topic, but I thought I'd address this here. The major issue with the political polls of 2016 was the issue regarding survey non-response, which affects whether the polls actually surveyed a representative sample of voters (in survey analysis, the gold standard is to sample from a representative random sample). And there had been a number of analyses of the polling data which indicated a close election, with Clinton being given a 52% or a 49% of the two-party vote, depending on which model you're looking at. A slight shift in voter turnout could easily result in a Clinton loss, as it turned out.

You can see more about this from Columbia University statistics professor Andrew Gelman on his blog entry dating back to December of last year.

http://andrewgelman.com/2016/12/08/19-things-learned-2016-election/

Bottom line is -- there is no indication of anything wrong with the methodology, and no suggestion from this example that "big data" is in need of a new approach.

 

[dkotschessaa]

This is off-topic, but I thought I'd address this here. The major issue with the political polls of 2016 was the issue regarding of survey non-response .<snip>..there is no indication of anything wrong with the methodology, and no suggestion from this example that "big data" is in need of a new approach.

Yeah, these issues are in the first chapter of an introductory statistics book.

If I say there is a 30% chance of something happening, and it happens, that doesn't mean I was wrong.

-Dave K