For a REST API these days documentation has gotten very easy due to the Open API standard. For most major frameworks you don't even need to write any sort of definitions, you just put annotations in your code. Then run it through Swagger UI (or similar) and it generates standardized, informative, and even interactive documentation. It's really beautiful.
The definition of AI has changed. We all need to get used to this. If you throw a fit every time somebody says "AI" without meaning some kind of magical super machine, all of your hair is going to fall out. The meanings of words change over time. Just go with it.
The modern definition is basically just any implementation of machine learning. Which is funny, because the phrase "machine learning" also used to be a buzzword that nobody in the industry actually used unless talking to the media. We'd be more likely to say "Computational Statistics" or "Function Estimation". So, in summary, the modern definition of AI is: statistics on a computer.
In ML that's called "Feature Importance". Some algos can do it. For example, if you're running a decision tree based model such as Gradient Boosting, you can determine which particular pieces of data were considered important in a prediction.
But I'd be willing to bet that's not what Watson is using. IBM is more concerned with making headlines than functional products. So they'll of course go right for the Neural Nets (Works like the human brain!1!) . Which is of course a black box that requires a huge amount of human work to get any idea at all about why it made a decision.
I worked with one who worked out VERY well. But he was a coder before he started the bootcamp. We hired him as a web dev, and the bootcamp was for Javascript. But he was coding for many years before taking the course, so it's not like the course taught him to code, it just helped him code something new.
Similar situation: I hired an IOS dev that learned from a Udacity course, and he's a rock star. But again, he knew how to code in something other than Swift before he started the Udacity course.
This sounds great for people that want to commit to using an algo for an extended timeframe. But since it's etched, you won't get any benefit from the constant stream of papers about deep learning that are being released. Keep in mind that the big ML competitions like ImageNet usually have a new high score record set every year. Not really a field in which I want to be behind the curve. Everybody could start using a spicy new activation function tomorrow for all we know. It's happened before.
Not to mention, training speed just isn't the problem that it used to be. You can get amazing results off your CPU if you've got some cores. Sometimes people will train giant monstrosity models that take a week on a pair of Titans, but nobody actually uses those models, that's just companies like Google and Facebook trying to show off.
Rent Google/Amazon's K80s if you're just dipping your toe in. Or buy an external GPU. This will give you all of the horsepower that you need, even if you're training giant models.
You would think that Google would own search via API, but actually Google's APIs for searching the whole web are surprisingly shitty. Bing's on the other hand perform just as well, require no configuration, and are far less expensive.
This is great news for anybody who uses these voice-based devices. The app ecosystem on these devices has been an absolute sewer, because only apps with a monetization strategy get published. There is a tonne of room for some very cool little applications that use voice inside the home for quick information requests and the like. But as is, nobody is going to build those apps because they cost money to run. Not very much money, but you still need to host a service somewhere. The worst case scenario for somebody up until now who was developing an Alexa Skill was to have it blow up in popularity, and then get stuck with a large hosting bill.
The difference is that the company delivering groceries to your door owns cars and groceries. Uber doesn't own cars. They don't just rely on an app, the app is their entire product. They are an orchestration layer only. Which makes them very much a tech company.
It does sound like the same shenanigans at play, although nobody is admitting to purposely breaking the Chinese one, like what happened with Tay.
This type of chatbot is a predictor about what will be said next in a conversation, based off of the words that have already been said. In the case of Microsoft Tay, it was being trained from Twitter. So all anybody had to do was make sure it was trained on their tweets, and they could make it say anything. If it sees a pair of tweet like "That dog is awesome" and the response "It must have a highly varied diet", then as long as this is the only time it's seen the phrase "That dog is awesome", if you say "That dog is awesome" to it, it will respond with "It must have a highly varied diet".
Of course, it was the 4chins that noticed this first in Tay's case, so it ended up saying some awful stuff. And I'm assuming that's what happened with Propaganda Bot too. Of course it's also possible that it just got some negation mixed up. Negation words in NLP are super hard to deal with and cause all kinds of headaches. So maybe it was supposed to say that it WAS a huge fan of the Communist Party. But seeming that it mentioned wanting to go to America, that sounds like it's just parroting some rapscallion.
This gets into why I think this whole Markov chain based thing is a total dead end in terms of AI. It just produces very convincing nonsense. If anybody is interested in the tech though, check out Karpathy's blog post "On The Unreasonable Effectiveness of RNNs". It's what kicked off the popularity of the technique.
Very old apps that aren't supported anymore are very likely broken at this point anyway. IOS isn't all that kind about reverse compatibility. Most apps need at least a little touching up each time a new version is released.
I find it ironic that Google invests so heavily in online education programs, but only hires people who have gone through the higher education song and dance. They straight up claim that their Udacity Android nano-degree will get you a job in the field, but how many of those grads are they hiring?
I'm not getting this either. People keep blaming Javascript, but we are talking about native IOS here, so it's either Swift or Objective-C. And it's not like they're going to have a bunch of graphics and video in the package, other than logos, backgrounds, etc. All that stuff would presumably be stored on a CMS. So what is all this bulk? Awfully fishy. Too bad it isn't Android we're talking about, or we could just decompile the APK and take a looksee.
Slashdot Mirror
void onDraw (Canvas canvas)
Implement this to do your drawing.
And big surprise, no mention in the style guide of updating documentation when something becomes deprecated.
For a REST API these days documentation has gotten very easy due to the Open API standard. For most major frameworks you don't even need to write any sort of definitions, you just put annotations in your code. Then run it through Swagger UI (or similar) and it generates standardized, informative, and even interactive documentation. It's really beautiful.
The definition of AI has changed. We all need to get used to this. If you throw a fit every time somebody says "AI" without meaning some kind of magical super machine, all of your hair is going to fall out. The meanings of words change over time. Just go with it.
The modern definition is basically just any implementation of machine learning. Which is funny, because the phrase "machine learning" also used to be a buzzword that nobody in the industry actually used unless talking to the media. We'd be more likely to say "Computational Statistics" or "Function Estimation". So, in summary, the modern definition of AI is: statistics on a computer.
In ML that's called "Feature Importance". Some algos can do it. For example, if you're running a decision tree based model such as Gradient Boosting, you can determine which particular pieces of data were considered important in a prediction.
But I'd be willing to bet that's not what Watson is using. IBM is more concerned with making headlines than functional products. So they'll of course go right for the Neural Nets (Works like the human brain!1!) . Which is of course a black box that requires a huge amount of human work to get any idea at all about why it made a decision.
Harder to turn a buck on mushrooms though.
The go-to product for AR before both of these releases was an open source project called ARToolkit... The plot thickens!
+1 Depressing
I worked with one who worked out VERY well. But he was a coder before he started the bootcamp. We hired him as a web dev, and the bootcamp was for Javascript. But he was coding for many years before taking the course, so it's not like the course taught him to code, it just helped him code something new.
Similar situation: I hired an IOS dev that learned from a Udacity course, and he's a rock star. But again, he knew how to code in something other than Swift before he started the Udacity course.
This sounds great for people that want to commit to using an algo for an extended timeframe. But since it's etched, you won't get any benefit from the constant stream of papers about deep learning that are being released. Keep in mind that the big ML competitions like ImageNet usually have a new high score record set every year. Not really a field in which I want to be behind the curve. Everybody could start using a spicy new activation function tomorrow for all we know. It's happened before. Not to mention, training speed just isn't the problem that it used to be. You can get amazing results off your CPU if you've got some cores. Sometimes people will train giant monstrosity models that take a week on a pair of Titans, but nobody actually uses those models, that's just companies like Google and Facebook trying to show off.
Rent Google/Amazon's K80s if you're just dipping your toe in. Or buy an external GPU. This will give you all of the horsepower that you need, even if you're training giant models.
Calculus is a good example of this. So much easier to understand once you've been coding for a while.
You would think that Google would own search via API, but actually Google's APIs for searching the whole web are surprisingly shitty. Bing's on the other hand perform just as well, require no configuration, and are far less expensive.
This is great news for anybody who uses these voice-based devices. The app ecosystem on these devices has been an absolute sewer, because only apps with a monetization strategy get published. There is a tonne of room for some very cool little applications that use voice inside the home for quick information requests and the like. But as is, nobody is going to build those apps because they cost money to run. Not very much money, but you still need to host a service somewhere. The worst case scenario for somebody up until now who was developing an Alexa Skill was to have it blow up in popularity, and then get stuck with a large hosting bill.
The difference is that the company delivering groceries to your door owns cars and groceries. Uber doesn't own cars. They don't just rely on an app, the app is their entire product. They are an orchestration layer only. Which makes them very much a tech company.
It does sound like the same shenanigans at play, although nobody is admitting to purposely breaking the Chinese one, like what happened with Tay.
This type of chatbot is a predictor about what will be said next in a conversation, based off of the words that have already been said. In the case of Microsoft Tay, it was being trained from Twitter. So all anybody had to do was make sure it was trained on their tweets, and they could make it say anything. If it sees a pair of tweet like "That dog is awesome" and the response "It must have a highly varied diet", then as long as this is the only time it's seen the phrase "That dog is awesome", if you say "That dog is awesome" to it, it will respond with "It must have a highly varied diet".
Of course, it was the 4chins that noticed this first in Tay's case, so it ended up saying some awful stuff. And I'm assuming that's what happened with Propaganda Bot too. Of course it's also possible that it just got some negation mixed up. Negation words in NLP are super hard to deal with and cause all kinds of headaches. So maybe it was supposed to say that it WAS a huge fan of the Communist Party. But seeming that it mentioned wanting to go to America, that sounds like it's just parroting some rapscallion.
This gets into why I think this whole Markov chain based thing is a total dead end in terms of AI. It just produces very convincing nonsense. If anybody is interested in the tech though, check out Karpathy's blog post "On The Unreasonable Effectiveness of RNNs". It's what kicked off the popularity of the technique.
Offices are unrealistically expensive, open spaces are distracting, and cubicles are depressing. I don't even know what I want anymore.
Very old apps that aren't supported anymore are very likely broken at this point anyway. IOS isn't all that kind about reverse compatibility. Most apps need at least a little touching up each time a new version is released.
X-files plot to turn us into alien chimeras.
Sign me up!
I find it ironic that Google invests so heavily in online education programs, but only hires people who have gone through the higher education song and dance. They straight up claim that their Udacity Android nano-degree will get you a job in the field, but how many of those grads are they hiring?
See, it's because the name "Comp Sci" is misleading. They should instead call this program "Programming circa 1961"
I'm really curious about where Musk is getting these ideas from. It certainly couldn't have been from anybody in the industry.
It would also help if investors had a clue instead of just giving money to the guy in a flashy suit that screams buzzwords at them.
I'm not getting this either. People keep blaming Javascript, but we are talking about native IOS here, so it's either Swift or Objective-C. And it's not like they're going to have a bunch of graphics and video in the package, other than logos, backgrounds, etc. All that stuff would presumably be stored on a CMS. So what is all this bulk? Awfully fishy. Too bad it isn't Android we're talking about, or we could just decompile the APK and take a looksee.
They've stolen the term AI. Just let it go man. You'll get less stress headaches that way.
The right side of the hippocampus is the part of the brain that makes you a square.