How Artificial Intelligence is radically transforming your TV experience

Published in

Predict

7 min readAug 19, 2018

This story is part of my series on How Artificial Intelligence transforms the world. The following article is How Artificial Intelligence is transforming Asset Management.

How emerging is a technology if it’s already in use. We often see artificial intelligence and machine learning as something of the future that “will” change the world. Look at our television experience.

I grew up with 3 TV channels. Our first TV set was in black and white. Our second TV set was a color Telefunken that we were so proud of that I still remember it. Our access to programing was for a long time, limited to the 3 government channels. And the tv set did not have a remote control.

Today when you log in your Netflix account I have access to shows related to my previous viewings, same on Amazon prime and Hulu. Additionally TV sets are also becoming “smart” and some have started to incorporate AI for various features.
So where do we stand?

AITV

Imagine using an Intelligent TV. It could be equipped with its own AI or connected to your Alexa or Echo. It’s now possible. LG has launched the Thinq brand in which one can use Alexa, Echo and Echo dot to change channel. There is a skill to watch TV, adjust volume, play, pause, start, stop, fast forward controls, select a channel and search for content. And, The new TVs can also function as smart home hubs. So you’ll be able to control other smart home devices like robot vacuum cleaners, air conditioners, smart lights, and others that can connect to the TV via Wi-Fi or Bluetooth. You can now do all of that by speaking to your TV while watching TV.

Source: IOTGadgets

The change also means better targeting

Imagine if the ds you were viewing were relevant to you. They would be catered to your age, gender, and preferences and instead of driving away would capture your attention. They would be aligned with your professional interests and your hobbies. It’s now possible. As of 2017, there are nearly 133 million connected TV users in the US and will grow to at least 181 million in this year. Now 55% of the population can receive ads tied to online and offline behavior. Machine learning excels at sifting through enormous amounts of data.
Programmatic companies now allow marketers to simultaneously advertise to a particular through various channels simultaneously. In short you may receive convergent signals from a brand or group of brands through your tv, laptop, tablet, smartphone and desktop at once.
Where machine learning excels is in predictive analytics, a search for a disease may precede a search for a treatment, drug or prescription. AI can now infer from a search for swim wear or hiking gear that you will soon search for travel prizes and accommodations. The machine now gives the marketers the ability to get in front of their client at the right time.

Source: Genesis media

AI is driving changes in advertising
Before the rise of AI it was virtually impossible for advertisers to know if the customers watch the spot, did the sports resulted in an action, a buy? Did their behavior change? Today consumers use multiple devices, all the time. While watching TVs the fizzle on their smartphones, keep their laptop on the coffee table or on their lap, and keep their tablet at hand. This new framework has created a complex landscape in which attribution is almost impossible. The technology solution that solves for this issue is called multi channel attribution across multiple screens (MCA — AMS). The objective here is to map the customer journey. By connecting multiple devices to the same individual, or household, AI allows to measure second screening while watching TV. By aggregating data on mobile presence in stores and mobile advertising, data scientist can infer the impact of the ad on the customer behavior.

Source: Accuracast.com

TV Programming

Like any other industries broadcasting is an industry in and on itself ready for digital transformation. Channel scheduling includes repetitive tasks that can now be done through robotic process automation. According to Doug Clark, IBM’s global cognitive solution leader, IBM Watson is already helping to speed up many of the processes involved in program production. “Watson is analyzing text, audio and video to create enhanced metadata and new ways of presenting content, with opportunities to mine and monetize it, and service the content more effectively” he said.

One of their recent projects is helping Fox Sports deliver a seamless experience for the 2018 world cup. The 2018 coverage includes 320 hours of total original broadcast content, including broadcasts of all 64 matches on FOX and FOX Sports channels. Aspera technology was used to send all of the feeds from the 12 sporting venues across Russia to Fox content management system and creative teams in LA.
For example Watson was able to build a highlight reel of any Cristiano Ronaldo or Lionel Messi shots on goal, from every World Cup they’ve played in.

At BBC4 Channel, machine learning is used to leverage some of the BBC’s 50 year archive of A-V content. The technology is used to find the best content, help explore the archive and spot content that would resonate with today’s audience.

Source: IBM

Technical applications

AI is also being applied in video compression and network elasticity. In video compression, AI is used to detect the quality of an image and is taught to compress the video thanks to hundreds of thousands of shots rated by viewers for their quality. Network elasticity helps streaming companies to manage their server load. AI is used to anticipate high loads for example when launching a new show and manage the allocation between the cloud and the company on premises servers.

Video recommendations

We are living the age of Peak TV. There were 487 programs aired in 2017.

There are a number of apps that aggregate all my subscriptions and offer me suggestions regardless of the platform based on my preferences and past views.
Some known players are Movix, Gowatchit, reelgood, JustWatch, and Yidio.
DeepSystems has launched Movix, a movie recommendation service based on artificial intelligence and Deep Learning. Users click movies and tag their likes and the system adapts preferences. They designed their own custom Neural Network architecture based on LSTM . It’s built on top of Ternsorflow framework from Google and runs on GPU. Unlike other engines, there are no pre-calculated recommendations — every click is being processed on the servers.

Arguably, Netflix is the most known video platform. Its recommendation engine is a key part of its success. The company has even published a research paper on the subject. It seems the platform uses the following techniques Features & techniques Netflix use to deliver the best possible experience to their users:

· Personalized Video Ranker: PVR: orders the entire catalog of videos (or subsets selected by genre or other filtering) for each member profile in a personalized way. The resulting ordering is used to select the order of the videos in genre and other rows, and is the reason why the same genre row shown to different members often has completely different videos.

· Top-N Video Ranker: produces the recommendations in the Top Picks row. The goal of this Algorithm is to find the best few personalized recommendations in the entire catalog for each member, that is, focusing only on the head of the ranking.

· Trending Now: which shows the videos that are trending in Netflix infused with some personalization for members. Data Collection for developing this algorithm happens with the following two streams: Viewing History (captures all the videos that are played by members) and Beacon (service that captures all impression events and user activities within Netflix).

· Continue Watching: sorts the subset of recently viewed titles based on the best estimate of whether the member intends to resume watching or re-watch, or whether the member has abandoned something not as interesting as anticipated.

· Video-Video Similarity: This one uses the movies you have just watched

· Page Generation: Row Selection and Ranking

· Evidence Selection

· Search Experience

· Statistical & Machine Learning Techniques for all of the above

Source: Redwoodalgorithms