The 'algorithms' Facebook etc run are just adaptive neural networks. They're trained on a set of user data points across millions of users to increase screen time, leading to more advertising impressions and revenue. There's nothing new or novel about neural networks. The first computational model of neural nets was created in 1943 by W…
The 'algorithms' Facebook etc run are just adaptive neural networks. They're trained on a set of user data points across millions of users to increase screen time, leading to more advertising impressions and revenue. There's nothing new or novel about neural networks. The first computational model of neural nets was created in 1943 by Warren McCulloch and Walter Pitts. Frank Rosenblatt's Perceptron was a trainable neural network built in 1957. The number of layers and size/width of the data sets have increased with computing power, but the 'algorithm' is the same.
This is why most of our algorithmically-provided "feeds" are getting really lame and tiresome. The feeds home in on your interests until you get sick of seeing the same SAMEness over and over.
What I'm saying is there's nothing to see "under the hood". You can't look at the internals of a neural network and make any sense of them. They're not algorithms with clearly defined logic.
On the outside they're simple revenue-optimizing machines. The inputs are your/our data, the output is slightly higher revenue than a human-designed heuristic model could provide.
We can see behind the curtain, and we're not impressed with the wizard. I think the great opting-out of social media is due. It has to provide some meaningful benefit to society, not just move billions of dollars from public to private hands. No one has ever said that of a business, not in the USA anyway. The measure has always been growth of shareholder value.
The 'algorithms' Facebook etc run are just adaptive neural networks. They're trained on a set of user data points across millions of users to increase screen time, leading to more advertising impressions and revenue. There's nothing new or novel about neural networks. The first computational model of neural nets was created in 1943 by Warren McCulloch and Walter Pitts. Frank Rosenblatt's Perceptron was a trainable neural network built in 1957. The number of layers and size/width of the data sets have increased with computing power, but the 'algorithm' is the same.
This is why most of our algorithmically-provided "feeds" are getting really lame and tiresome. The feeds home in on your interests until you get sick of seeing the same SAMEness over and over.
What I'm saying is there's nothing to see "under the hood". You can't look at the internals of a neural network and make any sense of them. They're not algorithms with clearly defined logic.
On the outside they're simple revenue-optimizing machines. The inputs are your/our data, the output is slightly higher revenue than a human-designed heuristic model could provide.
We can see behind the curtain, and we're not impressed with the wizard. I think the great opting-out of social media is due. It has to provide some meaningful benefit to society, not just move billions of dollars from public to private hands. No one has ever said that of a business, not in the USA anyway. The measure has always been growth of shareholder value.
I don't think right wing voters ever get tires of algorithms feeding them head lines that justify them hating liberals.