I dunno, if you use an Android phone with Google maps, by default, you can probably retrieve detailed info about your location and movements unless you've turned that feature off. I happened to have it on for several years, which turned out to be lucky if unintentional, as I used it to successfully defend myself in a legal matter. I could show times and places on a series of maps Google had tracking my phone checking in that were incompatible with the crazy claims of somebody accusing me of some totally irrational and false things. Creepy location tracking, true, but I was damn glad I had the info when I needed it.
I didn't say that it does mean that. I said it has come to be used that way. I support some prescriptiveness in language, but I am a realist about technical terms getting broad to the point of meaninglessness when they enter the mainstream. Fighting this is a frustrating and likely pointless endeavour (see: hacker).
Your theory seems to be that writers at TechCrunch think when they say a startup is using "AI" to improve recruiting or the sales process or even to process large volumes of textual data that they really think they mean artificial general intelligence. I believe they are using a vague catchphrase or abbreviation that gets clicks, and that they know that's not "real AI" but don't care.
Maybe the journalists at the New York Times or Washington Post are a more credulous lot and they really do think that is what it means, but I doubt it. They are all just capitalizing on confusion between Hollywood AI and machine learning for clickbait. That is what gets them paid.
At the same time, AI has come to be used in the tech press and mainstream press to mean "machine learning and related statistical techniques". Obviously this hurts the brains of many of us who still understand AI to mean "strong AI" or the newer moniker, "AGI" (artificial general intelligence), but we sort of have to roll with the language on this one.
Wikipedia insists that machine learning is a subset of AI. OK, sure, I guess that's fine, in that it is one of a series of techniques that can provide reasonable performance in solving certain human-level intelligence tasks.
When I'm speaking with a technical audience I go with more precise terminology - deep learning, reinforcement learning, unsupervised learning and try to avoid the headache, since we all know we are talking about statistical techniques in machine learning and none of it yet comes close to strong AI. Getting the mainstream press to correct their use of the abbreviation AI is about as likely as getting them to correct their use of the word "hacking".
Why? For large memory-requiring jobs and places where the latency of occasional cold start cases isn't acceptable, Lambda isn't the right tool. For periodic batch jobs, components in workflows, or response handling logic where some latency variability is acceptable, it's a great tool and saves a lot of costs over having dedicated but grossly underutilized EC2s.
Plenty of groundbreaking things have happened, just not so much in physics. Which is why despite having a degree in it, I work in an area where far more exciting things are going on these days (machine learning).
Demographic changes, abortion, technology (video games, cashless society, security systems), and economic changes. And maybe also some lead in the air.
AI == machine learning. All of these are machine learned systems. AI doesn't mean thinking magic. And yes, we do actually have DNN-based models that do simple reasoning now. Look up the Facebook bAbI challenge, for example, or question answering systems like FlowQA.
"AI" generally just means machine learning these days. And all of those are machine learned, statistical model-driven systems, not algorithmic systems. While things like SVM or Naive Bayesian classifiers have been around for years, DNN-based systems that perform well on much harder tasks that require large amounts of data to learn have only existed since 2011 or so.
Most of those billion Google Assistant devices were software updates pushed by Google to existing Android devices. This can be better thought of as the count of Android devices, and a small percentage of Google Home devices. Most of the 100 million Alexa devices, on the other hand, are actual devices sold with Alexa enabled, many of which were purchased specifically for use with Alexa.
Apparently they missed "startups 101" - the goal is to sell your company to a bagholder before the exponential growth dies off. Failure to find a bagholder is also known as "failure" in Silicon Valley.
This has nothing to do with AWS auto-scaling. The system that had issues doesn't run in public AWS. I can't say more than that unfortunately, but some random professor speculating based on leaked posts without any knowledge of the actual systems involved is a terrible source of information.
What I said was nobody wants to use it as currency. There are plenty of people who want to buy it and sell it as speculators. They are just betting that there is a greater fool down the road.
Equities actually represent a claim to a potential future income stream. Some may be very boring and predictable, some may be super speculative and risky - but there is some meaning to them.
Gold and oil have value as commodities because there are alternative uses of these commodities. So yeah, there is plenty of speculation, but you can take gold or oil and "do stuff" with it.
Bitcoin is an investment asset without any meaning sincere there is still no alternate use. Currency use has vanished to near zero while speculation, hoarding, and cybercrime make up essentially all of the transactions.
There's no "reason" because there's no rational valuation mechanism. Cryptocurrency without a mechanism for value stabilization is a scam. Blockchains are clearly useful for certain kinds of distributed trust problems, but Bitcoin is merely one instance that was always marketed as a cryptocurrency but has zero use as a transaction mechanism. Nobody wants to use a currency that may be worth 20% more, or 20% less the next day.
The only valid use case for Bitcoin I've heard described is as an improved version of the offshore banking system. In other words, a mechanism for rich people to launder and hide money. Of course, a cryptocurrency with value stability would sure as hell be a lot more useful and trusted for even this grey market purpose.
Ultimately, Bitcoin's value is driven by grey and black market activity. Money laundering, cybercrime, etc. Investing in Bitcoin is essentially investing in a residual claim on this underbelly of the economy, in the same way that regular fiat currencies are residual claims on national economies, with a healthy dose of mindless speculation and bubblemania thrown into the mix.
Some top-level surfacing of skills has already launched publicly. Can't say anything more than that, other than that this was addressed in the Alexa keynote at AWS re:Invent last week. If you are interested, I encourage you to watch:
Both Google Home and Amazon Alexa have speaker ID - they can differentiate between different voices speaking to them. And some functions can be tied to specific individuals rather than accounts, with more speaker-specific capabilities popping up now.
With PEP 484, you can annotate typed method signatures. And with PEP 526 you can annotate variable types too. If you're willing to go with straight up Python 3.6+ syntax, it's pretty nice looking. mypy can do pretty reasonable type-checking, which is supported in PyCharm and Atom via plugin now.
What's great is you can use it where you want extra checks (more complicated infrastructure code) and just stick with plain old duck typing where you don't (fast and loose scripts).
Very misleading to imply that this device is structurally any different from Google Home/Alexa/Siri/Google Assistant.
It says it is running proactive intelligence locally on the device. OK... but there is no way it's running ASR and NLU locally on a device of this form factor. There may be some notification logic locally on the device, fine, but this is pretty much negligible from a privacy impact perspective.
Agreed. We built a nearly identical system (with OpenCV for morphological analysis and neural networks) about a year ago, as part of a larger AI-powered mobile app development platform.
77% accuracy is not very impressive, but unclear what the training and test sets are here.
The biggest functional win from this is actually getting sensible layout params from a designer's UI mockup - i.e. figuring out should this be right justified/left justified, should there be margin/padding here, etc. We solved that problem pretty well.
Other challenges involve asset up-scaling, background image color extraction, etc. If you can take rough image mockups and output well polished asset packs, with vectorized images, layout files and stub code for developers to work with, that's a pretty significant win.
We got that far with the project, but ended up shifting direction to a somewhat different market where there was more growth potential - literally nobody wanted to invest further in mobile app tools in 2016, AI powered or not.
So yeah, cool proof-of-concept, but as a standalone offering this doesn't create much value. As part of a larger toolchain might be valuable.
Slashdot Mirror
I dunno, if you use an Android phone with Google maps, by default, you can probably retrieve detailed info about your location and movements unless you've turned that feature off. I happened to have it on for several years, which turned out to be lucky if unintentional, as I used it to successfully defend myself in a legal matter. I could show times and places on a series of maps Google had tracking my phone checking in that were incompatible with the crazy claims of somebody accusing me of some totally irrational and false things. Creepy location tracking, true, but I was damn glad I had the info when I needed it.
I didn't say that it does mean that. I said it has come to be used that way. I support some prescriptiveness in language, but I am a realist about technical terms getting broad to the point of meaninglessness when they enter the mainstream. Fighting this is a frustrating and likely pointless endeavour (see: hacker).
Your theory seems to be that writers at TechCrunch think when they say a startup is using "AI" to improve recruiting or the sales process or even to process large volumes of textual data that they really think they mean artificial general intelligence. I believe they are using a vague catchphrase or abbreviation that gets clicks, and that they know that's not "real AI" but don't care.
Maybe the journalists at the New York Times or Washington Post are a more credulous lot and they really do think that is what it means, but I doubt it. They are all just capitalizing on confusion between Hollywood AI and machine learning for clickbait. That is what gets them paid.
At the same time, AI has come to be used in the tech press and mainstream press to mean "machine learning and related statistical techniques". Obviously this hurts the brains of many of us who still understand AI to mean "strong AI" or the newer moniker, "AGI" (artificial general intelligence), but we sort of have to roll with the language on this one.
Wikipedia insists that machine learning is a subset of AI. OK, sure, I guess that's fine, in that it is one of a series of techniques that can provide reasonable performance in solving certain human-level intelligence tasks.
When I'm speaking with a technical audience I go with more precise terminology - deep learning, reinforcement learning, unsupervised learning and try to avoid the headache, since we all know we are talking about statistical techniques in machine learning and none of it yet comes close to strong AI. Getting the mainstream press to correct their use of the abbreviation AI is about as likely as getting them to correct their use of the word "hacking".
In fact... this kind of service has always been held up as a shining example of how businesses are *supposed* to make money off of GPL.
Why? For large memory-requiring jobs and places where the latency of occasional cold start cases isn't acceptable, Lambda isn't the right tool. For periodic batch jobs, components in workflows, or response handling logic where some latency variability is acceptable, it's a great tool and saves a lot of costs over having dedicated but grossly underutilized EC2s.
"democratic forum"... "trash heap full of retards and emotionally-stunted dipshits" - I fail to see how these are materially different?
Just look at what we do with our democratic elections in the United States right now.
Plenty of groundbreaking things have happened, just not so much in physics. Which is why despite having a degree in it, I work in an area where far more exciting things are going on these days (machine learning).
Demographic changes, abortion, technology (video games, cashless society, security systems), and economic changes. And maybe also some lead in the air.
ROFD beats Blockchain every time. Regular Old Fucking Database.
AI == machine learning. All of these are machine learned systems. AI doesn't mean thinking magic. And yes, we do actually have DNN-based models that do simple reasoning now. Look up the Facebook bAbI challenge, for example, or question answering systems like FlowQA.
"AI" generally just means machine learning these days. And all of those are machine learned, statistical model-driven systems, not algorithmic systems. While things like SVM or Naive Bayesian classifiers have been around for years, DNN-based systems that perform well on much harder tasks that require large amounts of data to learn have only existed since 2011 or so.
Most of those billion Google Assistant devices were software updates pushed by Google to existing Android devices. This can be better thought of as the count of Android devices, and a small percentage of Google Home devices. Most of the 100 million Alexa devices, on the other hand, are actual devices sold with Alexa enabled, many of which were purchased specifically for use with Alexa.
It's all about the GDP/IQ ratio.
Apparently they missed "startups 101" - the goal is to sell your company to a bagholder before the exponential growth dies off. Failure to find a bagholder is also known as "failure" in Silicon Valley.
This has nothing to do with AWS auto-scaling. The system that had issues doesn't run in public AWS. I can't say more than that unfortunately, but some random professor speculating based on leaked posts without any knowledge of the actual systems involved is a terrible source of information.
Source: I work at Amazon.
What about 1) traffic caused by bad drivers 2) hours of our lives wasted every day by having to focus on roads while we commute.
What I said was nobody wants to use it as currency. There are plenty of people who want to buy it and sell it as speculators. They are just betting that there is a greater fool down the road.
Equities actually represent a claim to a potential future income stream. Some may be very boring and predictable, some may be super speculative and risky - but there is some meaning to them.
Gold and oil have value as commodities because there are alternative uses of these commodities. So yeah, there is plenty of speculation, but you can take gold or oil and "do stuff" with it.
Bitcoin is an investment asset without any meaning sincere there is still no alternate use. Currency use has vanished to near zero while speculation, hoarding, and cybercrime make up essentially all of the transactions.
There's no "reason" because there's no rational valuation mechanism. Cryptocurrency without a mechanism for value stabilization is a scam. Blockchains are clearly useful for certain kinds of distributed trust problems, but Bitcoin is merely one instance that was always marketed as a cryptocurrency but has zero use as a transaction mechanism. Nobody wants to use a currency that may be worth 20% more, or 20% less the next day.
The only valid use case for Bitcoin I've heard described is as an improved version of the offshore banking system. In other words, a mechanism for rich people to launder and hide money. Of course, a cryptocurrency with value stability would sure as hell be a lot more useful and trusted for even this grey market purpose.
Ultimately, Bitcoin's value is driven by grey and black market activity. Money laundering, cybercrime, etc. Investing in Bitcoin is essentially investing in a residual claim on this underbelly of the economy, in the same way that regular fiat currencies are residual claims on national economies, with a healthy dose of mindless speculation and bubblemania thrown into the mix.
Some top-level surfacing of skills has already launched publicly. Can't say anything more than that, other than that this was addressed in the Alexa keynote at AWS re:Invent last week. If you are interested, I encourage you to watch:
https://www.youtube.com/watch?...
Both Google Home and Amazon Alexa have speaker ID - they can differentiate between different voices speaking to them. And some functions can be tied to specific individuals rather than accounts, with more speaker-specific capabilities popping up now.
With PEP 484, you can annotate typed method signatures. And with PEP 526 you can annotate variable types too. If you're willing to go with straight up Python 3.6+ syntax, it's pretty nice looking. mypy can do pretty reasonable type-checking, which is supported in PyCharm and Atom via plugin now.
What's great is you can use it where you want extra checks (more complicated infrastructure code) and just stick with plain old duck typing where you don't (fast and loose scripts).
Sure. But Alexa isn't that. You can turn the mic off any time you want, there's a button for that.
And Amazon has committed publicly and legally to not sharing your audio with third party developers. That isn't going to change.
This story is completely false. Somebody misunderstood something they read or just wrote it as pure clickbait to rile up the privacy crowd.
Here, LMGTFY:
https://en.wikipedia.org/wiki/...
and
https://en.wikipedia.org/wiki/...
Wow, sorry if those were the second Google results rather than the first. I'm sure it was really hard to disambiguate these acronyms for you.
Moreover, if you don't know what these acronyms mean, you are likely not qualified to comment on the privacy implications of products in this domain.
Very misleading to imply that this device is structurally any different from Google Home/Alexa/Siri/Google Assistant.
It says it is running proactive intelligence locally on the device. OK... but there is no way it's running ASR and NLU locally on a device of this form factor. There may be some notification logic locally on the device, fine, but this is pretty much negligible from a privacy impact perspective.
Agreed. We built a nearly identical system (with OpenCV for morphological analysis and neural networks) about a year ago, as part of a larger AI-powered mobile app development platform.
77% accuracy is not very impressive, but unclear what the training and test sets are here.
The biggest functional win from this is actually getting sensible layout params from a designer's UI mockup - i.e. figuring out should this be right justified/left justified, should there be margin/padding here, etc. We solved that problem pretty well.
Other challenges involve asset up-scaling, background image color extraction, etc. If you can take rough image mockups and output well polished asset packs, with vectorized images, layout files and stub code for developers to work with, that's a pretty significant win.
We got that far with the project, but ended up shifting direction to a somewhat different market where there was more growth potential - literally nobody wanted to invest further in mobile app tools in 2016, AI powered or not.
So yeah, cool proof-of-concept, but as a standalone offering this doesn't create much value. As part of a larger toolchain might be valuable.