If you are using commercial skip you already trusted Nuvyyo with your name, address, and payment information.
I guess I don’t have to worry since the whole article was about WiFi vulnerabilities and I don’t use WiFi. Everything is wired.
Commercial skip seems to be working quite well here. It does often show a second of the commercial before the skip happens, but I don’t have an issue with that. I have not yet seen it cut out part of a show and I’ve not yet had it fail to process on eligible programs. Still on the fence as to how I feel about it ever becoming a paid add-on.
Perfect results for me on multiple shows now, even with occasional pixilation due to poor signal. I continue to be impressed but I also don’t think I would pay for it separately.
Which existing customer is actually going to come on here and say “I love the feature that is currently free, please charge me more for it”? lol
At this point I’m just happy they decided to do an open beta, primarily because it does give us each the opportunity to see how it will work for us and whether it brings enough value to actually pay extra for it should they choose to do that.
It also gives them the opportunity to see decent load on their system and use that data to determine whether it does indeed need to be an additional cost.
It gives them (Tablo) the opportunity to have a large sample pool to refine their algorithms, procedures and services before they start charging for it. Basically all our Tablo units are just rats in the maze, Pavlov’s dogs. Give it away for free, get the community used to it and then slowly introduce a fee.
A one hour show requires up to 200 mb of data uploaded for analysis. So a 140 minute movie can wind up transferring half a gig to their servers, not an insignificant amount. It then need to be processed and an edl-type file downloaded back to your device. I can only imagine the cost in infrastructure and bandwidth to Nuvyyo.
The rats may be running around the maze for a while. Are they improving the algorithms or just expecting the machine learning to acquire enough samples to become better?
I no longer report total misses where long (2 minute) commercials were totally missed. And for the type of show I recorded it might be expected to be hard to analyze. So for those I don’t think ML will get smarter.
It’s true, each time I hear a dinner bell I now have an insatiable desire to test commercial skip on yet another show
@Spunky03 when something is free, you’re the product.
Though I’m not sure that’s the case now…
Sure it is.
If tablo’s DSP/ML reached nirvana they would license it to others.
Darn it, I knew that barcode tattoo was a bad idea.
ML algorithms are like belly buttons…
I would assume they’re using a supervised learning approach, so in order for what you’re proposing to happen, they need a massive amount of data, from all different regions, and employees sifting through the recordings to manually tag the start and end of commercials. Then they can test how well their algorithm is doing, using us the users.
Now, if they found a non-supervised approach (like if you’re looking for errors, you throw in a bunch of unlabeled data and assume that 0.001% or some chosen number is bad, then let the algorithm find out which 0.001% is bad. Anything that looks like that will be flagged. you’re likely to get some false negatives) – they don’t need as much human resources, but I don’t see how this would work – then again, I didn’t put that much thought into it either.
Do you think Tablo or someone like Tivo would have more resources to accomplish this?
I think it’s simpler then that to get a patent.
You’re right, here’s a patent to skip commercials on MPEG from 1999 https://patents.google.com/patent/US7272295
I’m sure there are others that are more applicable to Tablo.
<— not a patent lawyer
Since I don’t know anything about tablo’s commercial skip Digital Signal Processing or what they consider Machine Learning, it’s not clear if any other patent applies.
You have to remember that Digital Signal Processing is over 70 years old. I don’t think there is just one patent covering all Digital Signal Processing.
So now we’re back to Tablo making the best ML algorithm? While it’s possible, I find it unlikely for reasons I stated above.
I get what you’re saying, they need to patent ML + DSP + commercial skip. But if a competitor uses a different flavor of ML, they just went around the patent.
Thankfully that’s not how patent approval works. You just have to prove it’s unique. You don’t have to get all the way down into the weeds. If you look at some of Apple’s and Google’s patents their hand drawn pictures could apply to just about anything.
I’ve had approved patents where the patent attorney remove some of the more detailed formulas.
Lol, you make them sound like drug dealers.