Comments on: Wolfram Language Artificial Intelligence: The Image Identification Project http://blog.stephenwolfram.com/2015/05/wolfram-language-artificial-intelligence-the-image-identification-project/ Stephen Wolfram's Personal Blog Tue, 28 Nov 2017 21:16:32 +0000 hourly 1 http://wordpress.org/?v=3.4.2 By: Mallo http://blog.stephenwolfram.com/2015/05/wolfram-language-artificial-intelligence-the-image-identification-project/comment-page-1/#comment-1580878 Mallo Tue, 22 Nov 2016 02:24:43 +0000 http://blog.internal.stephenwolfram.com/?p=9922#comment-1580878 Reading the portion about the omission of human faces from an early training set somehow reminded me that, as I age, I am increasingly noticing how much some people I see Reading the portion about the omission of human faces from an early training set somehow reminded me that, as I age, I am increasingly noticing how much some people I see

]]>
By: Kate http://blog.stephenwolfram.com/2015/05/wolfram-language-artificial-intelligence-the-image-identification-project/comment-page-1/#comment-1564490 Kate Thu, 04 Aug 2016 05:46:37 +0000 http://blog.internal.stephenwolfram.com/?p=9922#comment-1564490 This is really a big story to identify an image.. This is my first time I'm here. Glad to see this project! This is really a big story to identify an image.. This is my first time I’m here. Glad to see this project!

]]>
By: Nicholas Gruen http://blog.stephenwolfram.com/2015/05/wolfram-language-artificial-intelligence-the-image-identification-project/comment-page-1/#comment-1455282 Nicholas Gruen Fri, 28 Aug 2015 09:15:10 +0000 http://blog.internal.stephenwolfram.com/?p=9922#comment-1455282 It called a hippo an indian rhinoceros for me - just as it called a cart an indian rhinoceros for you! It called a hippo an indian rhinoceros for me – just as it called a cart an indian rhinoceros for you!

]]>
By: Alexander http://blog.stephenwolfram.com/2015/05/wolfram-language-artificial-intelligence-the-image-identification-project/comment-page-1/#comment-1450580 Alexander Tue, 18 Aug 2015 08:47:38 +0000 http://blog.internal.stephenwolfram.com/?p=9922#comment-1450580 that is so awesome, what a milestone, google tried this and ended up with that creepy deep dream buisness, which is awesome by the way, but not the nail on the head, i still think algorithms like deep dream can be used to obtain expressions of real world objects in such a way as to allow replication via quantum physics, i'd really like to hammer that out with you or someone who might be interested on your team, as post scarcity is of the highest achievement a man could accomplish. that is so awesome, what a milestone, google tried this and ended up with that creepy deep dream buisness, which is awesome by the way, but not the nail on the head, i still think algorithms like deep dream can be used to obtain expressions of real world objects in such a way as to allow replication via quantum physics, i’d really like to hammer that out with you or someone who might be interested on your team, as post scarcity is of the highest achievement a man could accomplish.

]]>
By: Pete Pollock http://blog.stephenwolfram.com/2015/05/wolfram-language-artificial-intelligence-the-image-identification-project/comment-page-1/#comment-1400016 Pete Pollock Fri, 29 May 2015 23:44:09 +0000 http://blog.internal.stephenwolfram.com/?p=9922#comment-1400016 Dr Wolfram CONGRATULATIONS!!! I just saw your preliminary announcement of the image recognition algorithm. I think it's a real breakthrough. Arguably, it's one of the BEST ideas that Wolfram Research has produced. Outstanding job! Keep up the great work!! (Dr.) Pete Pollock, Aerospace Corp, Los Angeles Dr Wolfram

CONGRATULATIONS!!!
I just saw your preliminary announcement of the image recognition algorithm.
I think it’s a real breakthrough. Arguably, it’s one of the BEST ideas that Wolfram Research has produced.
Outstanding job!

Keep up the great work!!
(Dr.) Pete Pollock, Aerospace Corp, Los Angeles

]]>
By: Lars http://blog.stephenwolfram.com/2015/05/wolfram-language-artificial-intelligence-the-image-identification-project/comment-page-1/#comment-1398771 Lars Wed, 27 May 2015 02:40:37 +0000 http://blog.internal.stephenwolfram.com/?p=9922#comment-1398771 Why is is there a limitation to a single "class". I see the Watson services, eg as described in http://www.infoworld.com/article/2880146/machine-learning/ibm-watson-speech-text-visuals-analysis-insights.html, and wonder why the classification has to be unique, and not, say a "vector" of different things, either because the thing to be identified is a "hybrid" of some sorts, or because there may be several items in the picture. Is that a choice as part of "classification" into a unique bucket? Is unique classification something that will limit applications going forward? I assume the class system is predetermined and not flexible; a new training set with new "classes" would have to be generated. Wouldn't it for identification of pictures that fall out of the present class system be useful to also return some "likelihood" or "fit" factor? Why is is there a limitation to a single “class”. I see the Watson services, eg as described in http://www.infoworld.com/article/2880146/machine-learning/ibm-watson-speech-text-visuals-analysis-insights.html, and wonder why the classification has to be unique, and not, say a “vector” of different things, either because the thing to be identified is a “hybrid” of some sorts, or because there may be several items in the picture.

Is that a choice as part of “classification” into a unique bucket?
Is unique classification something that will limit applications going forward?
I assume the class system is predetermined and not flexible; a new training set with new “classes” would have to be generated. Wouldn’t it for identification of pictures that fall out of the present class system be useful to also return some “likelihood” or “fit” factor?

]]>
By: Frank Klaassen http://blog.stephenwolfram.com/2015/05/wolfram-language-artificial-intelligence-the-image-identification-project/comment-page-1/#comment-1394442 Frank Klaassen Wed, 20 May 2015 13:00:01 +0000 http://blog.internal.stephenwolfram.com/?p=9922#comment-1394442 As ProgrammingCloud AI effort, are you planning to integrate something like Watson or even Watson itself? Currently the power Watson has combined with Wolfram is amazing but they are not well integrated (manual work). To have that integrated into the language would be a great step. As ProgrammingCloud AI effort, are you planning to integrate something like Watson or even Watson itself? Currently the power Watson has combined with Wolfram is amazing but they are not well integrated (manual work). To have that integrated into the language would be a great step.

]]>
By: Admin http://blog.stephenwolfram.com/2015/05/wolfram-language-artificial-intelligence-the-image-identification-project/comment-page-1/#comment-1394034 Admin Tue, 19 May 2015 14:25:52 +0000 http://blog.internal.stephenwolfram.com/?p=9922#comment-1394034 Thanks for your question Philipp. That would be expensive! Actually, the ImageIdentify function does not access the Wolfram|Alpha API, which is what you referenced. It is part of the Wolfram Language, and costs Cloud Credits when used. It averages out to about 5 Cloud Credits per use of the function. Each Wolfram Programming Cloud plan provides thousands of Cloud Credits per month. Even WPC Free includes 1,000 Cloud Credits per month, so please start with the free account while you explore the Wolfram Language. We look forward to seeing what you create! Thanks for your question Philipp. That would be expensive! Actually, the ImageIdentify function does not access the Wolfram|Alpha API, which is what you referenced. It is part of the Wolfram Language, and costs Cloud Credits when used. It averages out to about 5 Cloud Credits per use of the function. Each Wolfram Programming Cloud plan provides thousands of Cloud Credits per month. Even WPC Free includes 1,000 Cloud Credits per month, so please start with the free account while you explore the Wolfram Language. We look forward to seeing what you create!

]]>
By: Marius Buliga http://blog.stephenwolfram.com/2015/05/wolfram-language-artificial-intelligence-the-image-identification-project/comment-page-1/#comment-1391503 Marius Buliga Sat, 16 May 2015 12:08:13 +0000 http://blog.internal.stephenwolfram.com/?p=9922#comment-1391503 I have tried it with several pictures of my artificial creatures and it worked or it failed in amusing ways. From one side, for pictures where I used a color scheme which I hoped it is suggestive for micro-organisms, the classifier returned BACTERIA. From another side, for pictures made with a color scheme which is more easy for explanations about how these creatures work, it returned JELLY BEAN. Which is accurate though! I have tried it with several pictures of my artificial creatures and it worked or it failed in amusing ways. From one side, for pictures where I used a color scheme which I hoped it is suggestive for micro-organisms, the classifier returned BACTERIA. From another side, for pictures made with a color scheme which is more easy for explanations about how these creatures work, it returned JELLY BEAN. Which is accurate though!

]]>
By: J Neville Groff http://blog.stephenwolfram.com/2015/05/wolfram-language-artificial-intelligence-the-image-identification-project/comment-page-1/#comment-1390832 J Neville Groff Fri, 15 May 2015 11:39:31 +0000 http://blog.internal.stephenwolfram.com/?p=9922#comment-1390832 interesting, the mistakes show its not thinking like a human. CNNs will again pass as we discover the underlying mechanics of thought and realize modelling wetware in silicon is a long, inefficient way around. interesting, the mistakes show its not thinking like a human.
CNNs will again pass as we discover the underlying mechanics of thought and realize modelling wetware in silicon is a long, inefficient way around.

]]>