REGISTER | LOGIN
Breaking News
Maxís Cool Beans

Are Norberts and Norbertinas the Future of the Human Race?

NO RATINGS
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 2   >   >>
DeeJee0
User Rank
Author
Re: ...
DeeJee0   2/15/2017 8:49:12 AM
NO RATINGS
Rick Merritt posted another great article that demonstrates that a "true AI" may need orders of magnitude less computational power than the brain to outperform us: http://www.eetimes.com/document.asp?doc_id=1331346&page_number=1

It describes a startup that uses statistical software to shorten the training fase of machine learning exponentialy. In near all problems, the training fase is the most computational intensive fase.

Two quotes sum it up: "Today's neural network approaches require training with the huge data sets typically only web giants possess.The job takes so much processing power that Google created its own accelerator, Facebook designed an open-source GPU server, and Intel acquired two chip startups to handle it. " and "It uses Python as an inference engine or solver algorithm, opening a door to easier interactions with programmers. The jobs run on standard servers faster than deep-learning techniques, the startup claims."

They made it vastly faster, cheaper and easier for programmers to perform. On top of that they achieve better results with smaller datasets. Likely their method will have its limits but these will get lesser as time passes. The time before machine learning or advanced AI becomes powerfull enough to disrupt society may be shortend quite a lot by this development. We'd better  step up our preparation quite a bit!

Max The Magnificent
User Rank
Author
Re: ...
Max The Magnificent   2/13/2017 3:23:04 PM
NO RATINGS
@DeeJee0: ...combine both and its precision/detection power/manipulation power will be almost "god-like"...

I must admit that I am starting to feel a nagging worry about some of this stuff...

 

realjjj
User Rank
CEO
Re: ...
realjjj   2/13/2017 12:59:19 PM
NO RATINGS
To some extent but that's very little.

However, when corporations or govs can manipulate individual targets at a massive scale, it gets nasty. We, as a society, lose free will. If today, The Prince of Orange would want to nuke Iran, he would likely have some 40% support. Manipulate a fraction of the remaining population and it's done.

With robots there are other complications. They would need to understand what we want and what we need (those can be different) and they need to be very good at it, to become commercially viable. Once good enough, they'll be our friends, lovers, teachers, shrinks and so on. Where are the limits, what's right and wrong, will be complicated...

Max The Magnificent
User Rank
Author
Re: Time travel?
Max The Magnificent   2/13/2017 12:56:19 PM
NO RATINGS
@Alex: ...How does a story from 1993, get in a book in 1977?

It's my stupid time machine on the fritz again -- it's an old model -- you can't get the spare parts these days.

OK -- you caught me -- my bad -- it should have said the book was published in 1997 -- I just fixed it in the article -- thanks for the catch

tb100
User Rank
Author
Re: Time travel?
tb100   2/13/2017 12:01:30 PM
NO RATINGS
Looks like a typo. The link connects to a book from 1997, not 1977.

Alex07004
User Rank
Author
Time travel?
Alex07004   2/13/2017 11:24:49 AM
NO RATINGS
In one paragraph, you says you got a book from 1977 that include the short story Norbert and the System. In the next paragraph, you say that the story was first published in 1993. How does a story from 1993, get in a book in 1977?

DeeJee0
User Rank
Author
Re: ...
DeeJee0   2/13/2017 11:12:50 AM
NO RATINGS
"But we all do that to some extent -- manipulation (or trying to manipulate) each other to gain an advantage, whether it's trying to make a subordinate do something, or trying to get your boss to do something for you..."

A major difference will be that humans have limited chances to learn, practice and optimise this ability. Some people hardly try this at all and are often called naive. Others have maybe one chance a day to practice. Used car salesmen practice most likely a couple of dozen times each day. Seeing a patern? Who would be best at it? The one with more chances to practice each day??? Even then, the used car salesmen will only be able to capture a small part of the available information. If he sees a certain gesture he won't be aible to replay all previous video in an exact manner to search if this behavior happened before.

Now imagine an AI having access to millions of distinct experiences of lying/speaking the truth/twisting the truth over millions of individuals and hunderds of cultures and nationalities. Consider it impossible that an AI would have such information? Not quite! Waring glasses that continuously send video and audio to the cloud suffice. Linking "no, I am not looking to eat in a restaurant" to "you who enter a competitors restaurant 5 minutes later around the corner" is easy as pie. Linking "no, I won't come and work for so few pay" to "you saying to your partner you desperately need that job" is easy as pie too. The restaurant holder will never know for sure if you lie. The employer will not know if you bluff. The AI will know. It will not only have millions of times more data to train its neural net, it will also have access to far more reliable data. Combine both and its precision/detection power/manipulation power will be almost "god-like". 

Having access to far more data and more reliable data allows for far better training of a neural net. That allows for either far better accuracy given the same computing power or exponentially lower computing requirements for the neural net compared to the human brain for the same accuracy or a combination thereof. That is only one of the examples why true AI does require far less computing power then the human brain does; see my comment in http://www.eetimes.com/author.asp?section_id=36&doc_id=1331264

 

Max The Magnificent
User Rank
Author
Re: ...
Max The Magnificent   2/13/2017 10:08:34 AM
NO RATINGS
@realjjj: ...if you understand someone, you can manipulate that person...

But we all do that to some extent -- manipulation (or trying to manipulate) each other to gain an advantage, whether it's trying to make a subordinate do something, or trying to get your boss to do something for you...

Max The Magnificent
User Rank
Author
Re: Interesting stuff....
Max The Magnificent   2/13/2017 9:33:02 AM
NO RATINGS
@David: ...I'll talk to people in supermarkets and trains...

I meant to tell you that they asked me to ask you to stop doing that :-)

realjjj
User Rank
CEO
...
realjjj   2/11/2017 4:45:44 PM
NO RATINGS
As parts of your article point out, not a feature you would want. It wouldn't be polite to wear glasses during a conversation to begin with.

Would be amusing to see how the glasses react to politicians and less amusing when your spouse gets a pair.

Robots utilizing the technology to understand us would be a positive but there would be risks and ethical complications. Note that well before robots get smart enough, the risks come from corporations and govs - if you understand someone, you can manipulate that person.

 

PS: Imagine Trump looking in the mirror with such glasses, assuming he is able to recognize his own image,.

Page 1 / 2   >   >>
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed