3 ms·

The premise before streaming tech matured at least was Netflix's data pipeline was so great and ML algorithms so good they'd recommend films so effectively and

The premise before streaming tech matured at least was Netflix's data pipeline was so great and ML algorithms so good they'd recommend films so effectively and produce films that users would crave. Feels like those two claims have failed, and at the end of the day content (IP) is king in the entertainment media world.

tyingq 4y ago
The recommender may be having trouble because of the large catalog of garbage content. That is, it's working right, but there's so much shite in the catalog that the recommendations are polluted.
- bdefore 4y ago
  Why is there a belief that the provider of content should also be the source of quality recommendations? I also see this disappointment with for example Steam and gaming. Or Spotify and music. But the days of walking into department stores and purchasing things based on a service rep's recommendations are behind us. We understand stores are incentivized to sell you the item most profitable to them. The Internet has a vast amount of places to inform your decisions independent of parties acting against your interests. Netflix is a gigantic warehouse of a lifetime of content. Some bad, some good. If you select something poorly, perhaps you're better served honing a better selection criteria.
  - tyingq 4y ago
    >Why is there a belief that the provider of content should also be the source of quality recommendations? I suppose because they should be able to do this well. They know what I have watched, what I bailed out early on, terms I search for and which results I looked at, and so on. And no shortage of tech talent to do the right thing with all of that.
pydry 4y ago
Has it failed? Youtube and tiktok's strategy of using top notch AI to hypertarget a tsunami of really cheap and mediocre content seems to work pretty well.
- fullshark 4y ago
  I think it failed for Netflix, those two sites have massive corpuses and a constantly pumping pipeline of fresh content. I guess they also figured out the economics re: paying content creators to ensure that's the case, Netflix I think dropped billions to establish IP to keep people hooked to Netflix and failed.
- polartx 4y ago
  I can’t speak for TikTok (as I’m not a user), but I haven’t met anyone that would consider YouTube’s recommendation algo any good at all. I believe most users share the experience of pausing their scrolling to deliberately say “Don’t recommend this channel”, or “I’m not interested in this”. Recommendations have to be pretty terrible for users to break the apathetic endless scrolling motion they’ve been conditioned to embrace, and make several distinct and deliberate clicks amounting to the provided feedback, “stop annoying me with how much you suck”

tyingq 4y ago

The recommender may be having trouble because of the large catalog of garbage content. That is, it's working right, but there's so much shite in the catalog that the recommendations are polluted.

bdefore 4y ago
Why is there a belief that the provider of content should also be the source of quality recommendations? I also see this disappointment with for example Steam and gaming. Or Spotify and music. But the days of walking into department stores and purchasing things based on a service rep's recommendations are behind us. We understand stores are incentivized to sell you the item most profitable to them. The Internet has a vast amount of places to inform your decisions independent of parties acting against your interests. Netflix is a gigantic warehouse of a lifetime of content. Some bad, some good. If you select something poorly, perhaps you're better served honing a better selection criteria.
- tyingq 4y ago
  >Why is there a belief that the provider of content should also be the source of quality recommendations? I suppose because they should be able to do this well. They know what I have watched, what I bailed out early on, terms I search for and which results I looked at, and so on. And no shortage of tech talent to do the right thing with all of that.

pydry 4y ago

Has it failed? Youtube and tiktok's strategy of using top notch AI to hypertarget a tsunami of really cheap and mediocre content seems to work pretty well.

fullshark 4y ago
I think it failed for Netflix, those two sites have massive corpuses and a constantly pumping pipeline of fresh content. I guess they also figured out the economics re: paying content creators to ensure that's the case, Netflix I think dropped billions to establish IP to keep people hooked to Netflix and failed.
polartx 4y ago
I can’t speak for TikTok (as I’m not a user), but I haven’t met anyone that would consider YouTube’s recommendation algo any good at all. I believe most users share the experience of pausing their scrolling to deliberately say “Don’t recommend this channel”, or “I’m not interested in this”. Recommendations have to be pretty terrible for users to break the apathetic endless scrolling motion they’ve been conditioned to embrace, and make several distinct and deliberate clicks amounting to the provided feedback, “stop annoying me with how much you suck”