3 ms·

They have a metric ton of data about you. I think they just don’t use it much because they probably realized that unless you get super creepy about it, most of

They have a metric ton of data about you. I think they just don’t use it much because they probably realized that unless you get super creepy about it, most of this data is worthless from an advertising point of view. Except on YouTube they don’t really know what you actually like - most of us don’t tend to express our real desires in the web but only in instagram or TikTok or YouTube. YT has ads but it doesn’t seem to have devolved into a cesspool of content only designed to sell you things as Insta and others have become. Your location could be useful but it becomes creepy if google used that to target you with ads so it seems to have bowed out of it (they don’t even seem to store location history in their servers anymore). Clearly they’re happy with the money they make from AdWords and andsense. So they’re leaving us alone with the rest of the data for now.

mnky9800n 2y ago
It could be that the wow factor of ai/ml/data/whatever came about because of these examples like Zappos increasing sales 1% because of changing the colour of the website or Netflix making better recommendations and ever since then that has kind of set the expectation. Both of those were great examples of what to do with data and algorithms but there's a catch, neither is really tailored to the individual in a way that it will consistently work over time. Netflix recommendations work because you will eventually pick something to watch and they increase the likelihood that you like what you pick. It works the other way too , knowing what people like helps them decide what new shows to make. But the only way they know this is that your profile matches a set of profiles. And this aggregate way of doing things will always have too much variance to be able to consistently be applied to an individual because different people are different even when they are similar. This is the ultimate conclusion that foundation model people have come to, at least implicitly, as they hope that by putting all data available such as the entire internet they will arrive at a combination of things that will consistently meet everyone's needs on demand whatever they may be. But the complexity of life is still higher than that and people are quick to notice when things are being averaged and then served to them.

mnky9800n 2y ago

It could be that the wow factor of ai/ml/data/whatever came about because of these examples like Zappos increasing sales 1% because of changing the colour of the website or Netflix making better recommendations and ever since then that has kind of set the expectation. Both of those were great examples of what to do with data and algorithms but there's a catch, neither is really tailored to the individual in a way that it will consistently work over time. Netflix recommendations work because you will eventually pick something to watch and they increase the likelihood that you like what you pick. It works the other way too , knowing what people like helps them decide what new shows to make. But the only way they know this is that your profile matches a set of profiles. And this aggregate way of doing things will always have too much variance to be able to consistently be applied to an individual because different people are different even when they are similar. This is the ultimate conclusion that foundation model people have come to, at least implicitly, as they hope that by putting all data available such as the entire internet they will arrive at a combination of things that will consistently meet everyone's needs on demand whatever they may be. But the complexity of life is still higher than that and people are quick to notice when things are being averaged and then served to them.