Machine learning algorithms (i.e. the kind of shit that can look

>>18119683
you cannot comprehend the powers of God.

Machines only function in binary, physical bandwidth, missing out a whole lot of other layers, so I highly doubt they could access other realms.

File: dolphin_woke_af.jpg (174KB, 790x559px) Image search: [Google]

174KB, 790x559px

>>18119944
This DESU. A machine intelligence as we understand digital technology, would be the equivalent of an entirely left brained individual. Dysfunctional to say the least. By the time you understand what the alternative to digital information is, in regards to formatting intelligence structures via physical components, you'll be sufficiently spiritually "woke" that just befriending a dog, or horse or something would be sufficient.

I mean, dolphins have bigger brains than you or I do. What's that tell you?

>>18119683
Computers are used to regulate the flow of traffic in cities. The way lights are timed is modeled in software to try and best control the flow with the least congestion. They go as far as to model places of business, how many people live and work in an area, and how much business they do per day. All of this to predict primarily where people will be, and what time of the day they'll be there. Whether lights are set to timers or are set to change on demand depends on this.

The same way that traffic is modeled is also used to model crime in cities. They can predict where crimes will occur, and even what type of crime it'll be. Entirely like in Minority Report sans identifying a specific offender. Jurisdictions that pay for the data service see a drop in crime rate by using this data to alter officer patrol routes. It in effect makes police more efficient by somewhat predicting the future.

Most everything in life revolves around set patterns, including human behavior. If you quantify those patterns then over a large enough number set you can make very accurate predictions on the probability of certain things occurring. These calculations aren't particularly hardware intensive. A couple racks of Xeons would give you enough power to model a major US city. Most of what needs improving is still the software side of things.

Fun fact: The DMZ in Korea has used automated defenses for a couple decades. There are computer controlled turrets that use machine learning to recognize the shape of a person, a vehicle, small animals, and other misc objects so they don't have any accidental discharges.
The same type of system used by self driving cars.

>>18119683
What you are referring to specifically are deep learning approaches, resulting in applications like deepdream.
Yes, given enough training they can perform excellent at classification, however the deepdream approach of using the neural network to create output is more or less a gimmick.

>>18119683
Computers have already scraped Twitter posts to make successful predictions in the stock market.