Maybe if it doubles as a pork barrel project distributed across multiple states... Actually, as computer work is easy to distribute remotely, this just might work.
- automated translation that actually works in many cases
- reliable text and speech recognition
- auto tagging of photos and videos
- AlphaGo victory, then a stronger version learning with no human game examples
- DeepMind AI mastering many video games with raw screen input and score only
- self-driving cars
- chatbots moving from toy projects into sales and marketing
- generative models making art and music which start getting enjoyable
Really, you have to be a committed denialist to not notice these, or brush them off as "not AI".
...if you want to chuckle about AI never happening. For Mr. Schmidt's warning message, Sputnik works better.
By the way, sustainable fusion could be one of the first problems that we might want to throw at an advanced AI. A lot of it is concerned with plasma field configuration, containment control, and materials design. All problems close to the automation, simulation and optimization techniques where computers already play a big role.
For this goal, it is too late to start obscure government programs to support education. I mean that helps, but is not enough. By now, what would really help is a government-sponsored Manhattan project style lab with top scientists stolen from everywhere, with a virtually unlimited budget, and a firm goal to build the legendary self-improving thinking machine, while keeping it under control.
Schmidt probably hopes that such a project would be based on DeepMind, because they have some sort of a head start, and a great team. That may be the case, but if the government did it with full force, Alphabet would not keep control of the workgroup. At best, they might work out some deal for a share in the intellectual property.
I guess the 26% figure is just some mean estimation from some modelling based on the typical benefits from ML-based resource optimization, projected development rate, etc. I guess there was no appropriate room to share the underlying assumptions, methodology, the margin of error, etc, because it was not a scientific conference.
I think it also makes sense for all countries to invest in this technology, just because it pays for itself. The US needs to pay attention and try to match or exceed China's commitment only as long as they want to continue seeing themselves as a major world power.
Actually, the Sputnik analogy is quite good. Sputnik also did not just happen overnight. There was research, rocket engineering, etc. In the end, Sputnik was just a huge wake-up call for US to catch up in a more focused way. It said "we are right there, over your head, in the ultimate high ground, with a research tool for now".
I don't think that anyone claims that the AI "Sputnik" has taken off already. The argument is that maybe the US could focus better this time without waiting for that ultimate wake-up call. Because it would be a similarly strong message sent in the stock markets, and perhaps in the military balance, as well.
It is true that once advanced AI technology is built by anyone, it is much easier for others to replicate it, stealing or not. It is just that this particular field can bring huge economic and military advantage extremely fast, before anyone can catch up, so some people are afraid in case that first mover advantage falls to some power they perceive as a fierce competitor, or a potential opponent.
There is no magic in "emergent behavior", and you do not "accidentally" get e.g. intelligent self-replicating robots by "stitching together" some 3D printers, motors and chips.
But there are a lot of very smart people working hard to move the ML and AI forward, and a lot of rich and influential people backing them, because the potential rewards are huge, and so is the risk of your competition arriving there first.
Also, there are many useful intermediate results, like self-driving cars, even if the general AI stays beyond our reach for another century.
As for the moral questions, they will be considered more seriously and solved quickly, once the general AI exists at least at the ape- or infant-level prototype, with a clear timeline for the next steps. Remember how people here said self-driving cars would never work even if technology existed, because we would never work out the laws? Turns out, as they gradually become a real thing, the laws catch up quickly, including the resolution of any moral issues, because the benefits are too huge to ignore. We will have some controversy, like with animal rights or equal opportunities today, but we will also have a working framework.
But more importantly, smart AI-powered systems reduce the need for custom code. We will not have Terminator-style humanoid robots replacing programmers in their cubicles. We will just need fewer programmers.
I think the closest threat from near-time AI/ML technology is AI-enhanced war logistics, planning, and battlefield tactics (drones, etc.) These technologies can provide huge advantages to otherwise underdog malicious powers, and it is hard to limit the proliferation because it is almost purely knowledge-based, as opposed to rockets and nukes. That's why every nation is working like crazy to develop an advantage in this area.
I would count the Chinese huge high speed rail network as one of the modern seven wonders. And they seem to enjoy great economical benefits from it. It is funny how in the game of Civilization, everyone knows to prioritize the roads, but in the real world most countries cannot pull off such huge infrastructure projects with reasonable speed, or at all.
I hope China does manage to take their rail network global, as planned.
I would count the Chinese huge high-sped rail network as one of the modern day seven wonders. And it looks like they are ripping huge economical benefits from it. It is funny how in the game of Civilization everyone knows to prioritize the roads, but in the real world most countries fail to pull off such massive infrastructure projects with reasonable speed.
I hope China does manage to take their rail network global, as planned.
Linkedin wants to have their cake and eat it, too. The users post their data for all interested parties to see, unless they put some explicit restrictions (e.g. friends only). Linkedin then add all sorts of artificial limits on visibility, search, and god forbid you try to fetch that data with a script. Suddenly it is no longer the person's data shared as they want, but Linkedin's data intended for monetization.
I understand they have expenses incurred by careless bots. It is possible to traffic shape the active connections, or provide a reasonable API, without being greedy and hypocritical, obfuscating the data that is not yours, and pretending it is about the user protection.
Our best bet to get a closer look any sooner is to use our Sun as a gravitational lens. It is still a challenge, because we would need to put a telescope at the correct side of the Sun at about 550 AU, far beyond the orbit of Pluto, but it is much closer to our technological reach than actual interstellar probes. NASA is thinking about this project: https://www.technologyreview.c...
I would say it is a stunning improvement over previous years, when robots mostly just struggled to stay on their feet and try to get to the ball. Given the better-than-linear improvement rate in vision, battery, movement control, team coordination, etc, I expect this to become an enjoyable show in a couple of years.
I have been running the same PC box for 13 years. In this time, I have upgraded/replaced:
monitor: twice
video card: 2 or 3 times
HDD/SSD: 3 or 4 times
CPU: once (core quad Q6600 to cheap used core 2 QX9770)
RAM: 2 or 3 times
added small things like WiFi, webcam, etc as needed
I could easily afford each incremental upgrade or replacement, solving any specific bottlenecks I encountered.
I will probably buy a laptop (because they've gotten fast enough) when the motherboard breaks down, or some groundbreaking new tech goes mainstream, but so far this box just keeps working fine.
No mechanism in physics for consciousness?! It is well established that both consciousness and intelligence are based in electrochemical phenomena in the brain, and we are mapping them out in greater detail every day. Just because we do not have a full map and understanding yet, does not mean we never will. All knowledge was new at some time. No need to mystify it.
Slashdot Mirror
Hype bubbles happen; so do real advances. Our understanding of computers has advanced a lot since early 80s.
Maybe if it doubles as a pork barrel project distributed across multiple states... Actually, as computer work is easy to distribute remotely, this just might work.
- reliable text and speech recognition
- auto tagging of photos and videos
- AlphaGo victory, then a stronger version learning with no human game examples
- DeepMind AI mastering many video games with raw screen input and score only
- self-driving cars
- chatbots moving from toy projects into sales and marketing
- generative models making art and music which start getting enjoyable
Really, you have to be a committed denialist to not notice these, or brush them off as "not AI".
...if you want to chuckle about AI never happening. For Mr. Schmidt's warning message, Sputnik works better.
By the way, sustainable fusion could be one of the first problems that we might want to throw at an advanced AI. A lot of it is concerned with plasma field configuration, containment control, and materials design. All problems close to the automation, simulation and optimization techniques where computers already play a big role.
For this goal, it is too late to start obscure government programs to support education. I mean that helps, but is not enough. By now, what would really help is a government-sponsored Manhattan project style lab with top scientists stolen from everywhere, with a virtually unlimited budget, and a firm goal to build the legendary self-improving thinking machine, while keeping it under control.
Schmidt probably hopes that such a project would be based on DeepMind, because they have some sort of a head start, and a great team. That may be the case, but if the government did it with full force, Alphabet would not keep control of the workgroup. At best, they might work out some deal for a share in the intellectual property.
I guess the 26% figure is just some mean estimation from some modelling based on the typical benefits from ML-based resource optimization, projected development rate, etc. I guess there was no appropriate room to share the underlying assumptions, methodology, the margin of error, etc, because it was not a scientific conference.
I think it also makes sense for all countries to invest in this technology, just because it pays for itself. The US needs to pay attention and try to match or exceed China's commitment only as long as they want to continue seeing themselves as a major world power.
You have to plan ahead and hedge your bets, as a country as much as a corporation. And don't pretend there are no actual achievements, either.
Actually, the Sputnik analogy is quite good. Sputnik also did not just happen overnight. There was research, rocket engineering, etc. In the end, Sputnik was just a huge wake-up call for US to catch up in a more focused way. It said "we are right there, over your head, in the ultimate high ground, with a research tool for now".
I don't think that anyone claims that the AI "Sputnik" has taken off already. The argument is that maybe the US could focus better this time without waiting for that ultimate wake-up call. Because it would be a similarly strong message sent in the stock markets, and perhaps in the military balance, as well.
It is true that once advanced AI technology is built by anyone, it is much easier for others to replicate it, stealing or not. It is just that this particular field can bring huge economic and military advantage extremely fast, before anyone can catch up, so some people are afraid in case that first mover advantage falls to some power they perceive as a fierce competitor, or a potential opponent.
There is no magic in "emergent behavior", and you do not "accidentally" get e.g. intelligent self-replicating robots by "stitching together" some 3D printers, motors and chips.
But there are a lot of very smart people working hard to move the ML and AI forward, and a lot of rich and influential people backing them, because the potential rewards are huge, and so is the risk of your competition arriving there first.
Also, there are many useful intermediate results, like self-driving cars, even if the general AI stays beyond our reach for another century.
As for the moral questions, they will be considered more seriously and solved quickly, once the general AI exists at least at the ape- or infant-level prototype, with a clear timeline for the next steps. Remember how people here said self-driving cars would never work even if technology existed, because we would never work out the laws? Turns out, as they gradually become a real thing, the laws catch up quickly, including the resolution of any moral issues, because the benefits are too huge to ignore. We will have some controversy, like with animal rights or equal opportunities today, but we will also have a working framework.
usual corporate insanity, with a touch of bad luck
I knew this was possible, but did not know it was widespread. Thanks for sharing.
which will be slow but wireless and cheap at $200.
They simply misspelled "Visual Studio". https://cloudblogs.microsoft.c...
It already does: http://www.techrepublic.com/ar...
But more importantly, smart AI-powered systems reduce the need for custom code. We will not have Terminator-style humanoid robots replacing programmers in their cubicles. We will just need fewer programmers.
I think the closest threat from near-time AI/ML technology is AI-enhanced war logistics, planning, and battlefield tactics (drones, etc.) These technologies can provide huge advantages to otherwise underdog malicious powers, and it is hard to limit the proliferation because it is almost purely knowledge-based, as opposed to rockets and nukes. That's why every nation is working like crazy to develop an advantage in this area.
This only means the profit margin on them will go from 90% to 99%.
I would count the Chinese huge high speed rail network as one of the modern seven wonders. And they seem to enjoy great economical benefits from it. It is funny how in the game of Civilization, everyone knows to prioritize the roads, but in the real world most countries cannot pull off such huge infrastructure projects with reasonable speed, or at all.
I hope China does manage to take their rail network global, as planned.
I would count the Chinese huge high-sped rail network as one of the modern day seven wonders. And it looks like they are ripping huge economical benefits from it. It is funny how in the game of Civilization everyone knows to prioritize the roads, but in the real world most countries fail to pull off such massive infrastructure projects with reasonable speed.
I hope China does manage to take their rail network global, as planned.
Linkedin wants to have their cake and eat it, too. The users post their data for all interested parties to see, unless they put some explicit restrictions (e.g. friends only). Linkedin then add all sorts of artificial limits on visibility, search, and god forbid you try to fetch that data with a script. Suddenly it is no longer the person's data shared as they want, but Linkedin's data intended for monetization.
I understand they have expenses incurred by careless bots. It is possible to traffic shape the active connections, or provide a reasonable API, without being greedy and hypocritical, obfuscating the data that is not yours, and pretending it is about the user protection.
It is also reported that the terrorists used US Dollars to fund their activities! Ban US Dollar!!!1
Our best bet to get a closer look any sooner is to use our Sun as a gravitational lens. It is still a challenge, because we would need to put a telescope at the correct side of the Sun at about 550 AU, far beyond the orbit of Pluto, but it is much closer to our technological reach than actual interstellar probes. NASA is thinking about this project: https://www.technologyreview.c...
I would say it is a stunning improvement over previous years, when robots mostly just struggled to stay on their feet and try to get to the ball. Given the better-than-linear improvement rate in vision, battery, movement control, team coordination, etc, I expect this to become an enjoyable show in a couple of years.
I have been running the same PC box for 13 years. In this time, I have upgraded/replaced:
I could easily afford each incremental upgrade or replacement, solving any specific bottlenecks I encountered.
I will probably buy a laptop (because they've gotten fast enough) when the motherboard breaks down, or some groundbreaking new tech goes mainstream, but so far this box just keeps working fine.
No mechanism in physics for consciousness?! It is well established that both consciousness and intelligence are based in electrochemical phenomena in the brain, and we are mapping them out in greater detail every day. Just because we do not have a full map and understanding yet, does not mean we never will. All knowledge was new at some time. No need to mystify it.