In trading deep learning does not work or is even not applicable because price series properties are not fixed spatial but time variant.
How This Hedge Fund Robot Outsmarted Its Human Master ...It makes for a difficult backdrop as Nomura tries to drum up appetite for a strategy with almost no real-world track record. Some prospective institutional clients are so hesitant to stick their neck out for an unfamiliar product that they asked Nomura to remove the term AI from Simplex’s promotional materials. They didn’t want to have to explain how it works to their bosses, he said. “There are very few AI funds currently out there globally,” said Mohammad Hassan, a senior analyst at Eurekahedge in Singapore. “These early adopters of machine-learning methodologies will need to build a good track record before investors give them serious attention.”... http://www.bloomberg.com/news/artic...tsmarts-human-master-as-ai-passes-brexit-test
Have seen this vid before and liked the ideas of the oldest talker of 3 sitting on left. Those utopia topics are nice, i have all sorts of ideas also. Maybe there is "more" outside the frame-system or "box" or univers where rules are even more high-resolution or complex or different and if someone flicks the switch than movie will be rewinded,who knows, but i supose we as humans are not given too much brain flexibility to think about all sorts of big and abstract things, but i think more the society has its work automated the more time general public gets to think about that sort of stuff.(Or play video games) About Ai , I think for computer to percive its suroundigs it needs to construct internal copy of the outside presentation based on input data of some spectrum that can be collected. Later based on the "copy" it should be possible code some flexible algo to explore further possibilities. Best if algo is not always needing fastest hardware but work around the problems instead of brute forcing trough. Also humans could be looked as self repairing bio-computer instead of silicone/transistor. I think we live in interesting times and evolution is growing exponentialy, 100 year or more ago much was based on horses making work ,before slaves were abused by psychopatic people. now alot is automated and so many routine jobs lost. But once someone get some good algo done and the automation goes to not only routine but regular or creative jobs then who knows what will the outcome be. Maybe someone with good imagination can post some thoughts or movies on what may happen to society while transitioning to ai?
It doesn't change one simple problem, the fastest super computers in the world can only process 0.00041% of what the human brain can do when dealing with context. Algorithms are much faster but can only process pre-configured context, ie. known events. The issue arises that the world is not static, so the algorithms need to be adapted, and when they think the AI or Machine Learning or whatever else they decide to call it has the answer, life will throw a curve ball and break all the models. So all the processes are just trying to increase probability, some are very good at it, but for context you will need a person involved. Once you go above 95% automation you are increasing the risk of fat-tail tail events finding a way through the holes in the models, because at that level there so many permutations you can guarantee some were not tested. At 99% automation and above a fat-tail event finding its way through is normally catastrophic.
I believe we will go into "pseudo" economy, as all physical needs will be met by robots, there will still be inphysical human needs unmet. These are things like playful competition between peers in order to determine who gets the hottest girl, or who gets to be the leader, stuff like that.
Tiny Bank-Beating Trading Firm Doesn’t Use Any Human Traders John Detrixhe johndetrixhe October 12, 2016 — 11:00 PM CDTUpdated on October 13, 2016 — 6:21 AM CDT Share on FacebookShare on Twitter Self-learning machines help put XTX among top currency traders Company is expanding into other markets, geographies Share on FacebookShare on Twitter One of the world’s fastest-growing trading shops doesn’t have any traders. XTX Markets Ltd. has emerged as a foreign-exchange powerhouse, relying on programmers and mathematicians to fuel its rise into the global top five earlier this year. Now, after becoming a formidable player in currencies, XTX has its sights set on growing in stocks, commodities and bond markets. But in a world where the difference between profit and loss can be tiny fractions of a second, XTX says it relies more on smarts than speed. Instead of building microwave networks to ferret out prices a microsecond before anyone else, XTX uses mathematical models that are tuned with massive data sets. It says its technology has computing power comparable to some of the world’s top supercomputers. XTX’s rise comes amid a market shift: In an arms race where just about anyone can lease ultra-fast trading systems, it’s harder than ever to get an advantage simply by being fast. That may have opened up an opportunity for firms with sophisticated statistical models, says Greenwich Associates. As XTX co-chief Zar Amrolia sees it, a key trial for the London-based firm comes when it competes in even more markets that are dominated by high-frequency traders, particularly in the U.S. Speed vs. Smarts “Can a smart, machine-learning approach beat probably a less sophisticated algo which is faster?” Amrolia said. “That’ll be a really exciting test.”.. http://www.bloomberg.com/news/artic...-trading-powerhouse-doesn-t-use-human-traders
Differentiable Programming Over the past few years, a raft of classic challenges in artificial intelligence which had stood unsolved for decades were conquered, almost without warning, through an approach long disparaged by AI purists for its "statistical" flavor: it's essentially about learning probability distributions from large volumes of data, rather than examining humans' problem-solving techniques and attempting to encode them in executable form. The formidable tasks it has solved range from object classification and speech recognition, to generating descriptive captions for photos and synthesizing images in the style of famous artists—even guiding robots to perform tasks for which they were never programmed! https://www.edge.org/response-detail/26794