AI in Media & Social: A nail begging for a hammer

First things first. Let’s get the generic disclaimers out of the way.  

Yes, we can’t predict how AI will change the world. Yes, AI is wondrous and magical and scary. Yes, we don’t want to be “strong and wrong” in our opinions. Etc etc etc. It’s still fun to run thought experiments. Today we’re going to put our imagination hats on and deep dive into the world of ‘Media and Social’. We’ll also uncover a few opportunities for large value creation for new companies.  

‘Media and Social’? You mean Social Media?

Yes… and no. Let’s unpack what we, at Matrix, classify as ‘Media and Social’ (MS). It’s a broad term that includes the following

1. Creators/ Influencers

2. Social Media & Social Networking platforms (Youtube, Josh, Clubhouse, Tinder, Linkedin, Blind)

3. OTT platforms and Content Aggregators (Netflix, Spotify, Inshorts, Dailyhunt)

4. Media and Publishing companies (ESPN, Disney, Universal, Sony, Jio)

What’s common you ask? Content.  

Basically it’s a broad bucket of everything that you waste your time on J. This category, a mixture of text, audio, images and videos, demands BOTH quality and quantity. It also requires speed to capture our fickle attention spans. AI, with its multi-modal capabilities (text, image, video, audio etc) is not a hammer looking for a nail in MS.  

This nail is begging to be hammered.  

AI = Mjölnir (not your regular hammer)

Andrej Karpathy captures the sentiment best in his Software 2.0 article. There are some (all?) things that AI just does better – but it’s superiority in handling modalities (text to image, image to video, video to video etc) is already proven. Our internal view on maturity across modalities is captured in the image below (shoutout to Ankur Edkie at Murf.ai for his insights here).

The green arrows on its own represent Billion $ value pools, but unlocking the yellow and the red ones will be game-changers. The primary reason - video is just a tough format to work with. Editing videos is among the most frustrating jobs you can do in a creative setting. Compile raw footage, sync audio, clip & stitch videos, adjust light – let’s just say it’s not your average, “I-made-it-on-Canva-in-5min” experience. Now imagine “Canva for Video powered by text prompts”. Before you start prompting “Nolan’s Tenet in linear plotline”, let’s remember that Canva is a 25B$ company solving for just images.  

Does that mean all value will be captured by a few large companies? In our view, while some companies will capture especially large value pools (especially those that solve for modalities), there will be multiple winners who solve deeply for the stakeholders they serve. Before we look at stakeholder-specific opportunities let’s make some sweeping observations/predictions.

1. AI will push the boundaries on one of the great content dichotomies – Quality vs Quantity. Making high quality content at much higher volumes will now be possible.  

2. AI promises hyper-personalization at scale. If you thought Bandersnatch was cool, wait till your social media or OTT experience is fully personalized for you. It may start with curation, but real-time creation may not be far off. Content designed for you in real time basis predicted experience value.  

3. High velocity of content will lead to innovations in content management. We are already attention starved – now imagine 100X content volume. This has a few implications

a. Moderation and compliance verticals will see significant solutioning.  

b. Bar on curation of content and personalization of feed will go up drastically (creating virtuous loop with pt 2)

So, how to build AI for Media and Social? Just like every business – thinking Customer backwards! Let’s look at our stakeholders.  

1. Creatorus Maximus

Our Creatorus Maximus is a simple animal. He/she wants to do 2 things -  create better content at higher volumes and increase viewership. AI can help unlock both.

Create better content at higher volumes

A typical 20min YT video includes research, scripting (1-2hrs), set-up (1-2 days), shooting (3-6hrs) and post-production (1-2 days). While scripting itself can get replaced by a ChatGPT plugin (“Tarantino plot + Sorkin dialogues”), tailored solutions could learn from the Creator’s previous work to auto-create plots/scripts. The bigger unlock, however, will be in set-up, shooting and post-production.

• Set-up & Shoot: Green screens have revolutionised both cost and quality of production - it also led to significant reduction in time to shoot (except for James Cameroon). AI can reduce this even more drastically. Shoot a scene with actors in different locations and no props. AI stitches everything together including props, background, music and lighting. Heck, you’ll just create an AI superstar and reduce your cost of cast/crew (20-50% of budget). You won’t need 60M$ to make Zack Snyder’s 300 – just download a virtual Gerard Butler with 6-pack abs.
• Post production: “AI for X” (X= Adobe Premiere, FL studios, Moviemaker etc). Professional level editing through simple Ui gestures / text inputs will unlock hours of productivity. Simple tasks are already close to being automated (remove background, add music, extend image, edit podcast, add thumbnail etc), but the future will create entire video workflows from start to end. Post production for a YT video might take less than an hour compared to ~2 days currently. Key to also note that Large players will not be far behind in innovation – check out Adobe Firefly integration with Photoshop.

• Re-purpose content: Lex Fridman typically does a 3-5hr podcast with among the smartest people in the world discussing the most abstract of topics. What do you do with a marathon 4hr podcast? Repurpose it of course. Using existing content to create new is a no-brainer – what is a manual process will soon be automated with a few clicks.  

By the way, the modern creator might have a fully virtual presence. AI can create viral content with minimal effort. Why take a leap of faith when we can point you to two live examples?

1. Check out the Balenciaga viral videos – The Harry Potter version has >9M views. Fully AI generated.

2. 23yr old Snapchat Influencer made waves with her AI girlfriend avatar. CarynAI (which made >$70k in one week) could be your girlfriend for 1$ per minute. The model was trained on 20k hours of her Youtube content.  

Increase viewership

AI allows the creator to increase distribution in interesting ways fully utilizing capabilities around modality and transformation. Three fundamental ways in which Creators can increase viewership – Increase formats, broaden distribution or grow audience (plus some quick “future prompts”)

Size of prize and what it will take to win

The Influencer marketing industry is estimated to be >20B$ in 2023, estimated to be ~150B$ in 2030. Capturing just 1-5% of the spends would mean a 1.5-7.5B$ revenue opportunity in 2030. To capture a meaningful share companies, in our view, will need to plug themselves directly into the Influencer workflow. Deeper the solution, higher the value capture and stickiness. A full stack solution where the creator is fully dependent on your solution would mean servicing the creator across many of the work-blocks outlined below. While all blocks are critical  Multi-platform, Library and Audience Management can drive stickiness through 10X product value.  

2. Social Platforms (Media and networking)

Social platforms love AI – how else can you implement a system that uniquely engages with millions of users. Meta has potentially been using AI/ML to personalize your feed since 2015. Social platforms can benefit from AI across 4 key dimensions.  

1. Increase engagement (get you even more hooked)

a. Sorting and personalization (already live)

b. AI generated content

2. Decrease content costs (fatten bottom-line)

a. AI generated content

b. AI generated music

3. Optimize Ad targeting (get you to buy more)

a. Personalized targeting (already live)

b. Personalized ad creatives and messaging

4. Increase safety and compliance (do the above legally)

a. Removing harmful content (already live)

b. Complying with copyright laws

Most of the above are self-explanatory, but it pays to look at AI generated content and Personalized Ad creatives.  

Why Social Platforms will leverage AI for content.

Social platforms spend a lot of money on creators. Youtube reportedly has paid out 30B$ between 2018 and 2021 to across 2 Million video creators. The Tiktok Creator fund announced in 2020 set aside 200M$ for creators that year – estimated to spend close to 1B$ in 3 years. While it’s unlikely that big platforms will risk angering creators, it is possible that the newer platforms embrace AI to generate engaging content. This especially applies for social platforms struggling for monetization – unable to break the vicious cycle of “less users = less creators = less users”.  

AI also allows for quick generation of content - imagine a short video of Dhoni hitting a 6 off the last ball 10s after the shot. The time lapse between “live” and “non-live” will reduce considerably – creating infinite pieces of content flavoured with the gossip of the day. Yes, we know you enjoyed that Johnny Depp trial!

Music is another interesting application of AI based content. Platforms share more than a pound of flesh with Music labels for the songs used in creator videos – they’d love to get rid of this expense item. More about this here.

Ad personalization on steroids!

If you’re worried that Meta is reading your whatsapp messages, wait till it shows you an Ad creative you can’t ignore. More benign applications might involve showing ad creatives tailored to what was searched, your (predicted) mood or current location. Before you get diverted to that “Black Mirror” episode, we’d like to point out that some element of this happens already. It will still hit you hard when the next ad you see features… You. Maybe you’ll like it?  

Size of prize and what it will take to win

Solving for the big Social platforms is always a challenge – they typically have large, AI capable teams and hence will look to in-source most of the value creation (potential M&A action?). Value might be best captured in the following pools

1. AI content generation at speed and scale – especially if tailored to their audience

2. Smaller social networks (vs social media). Think Tinder before Meta.

3. Ad personalization

Building for the Social platforms (as well as OTT and Content aggregators) could require an Enterprise first bespoke solution where the plumbing is customized for their specific use-case.  

Comments from Josh/Verse

Comments from Glance

3. OTT platforms and Content Aggregators

A lot of what was discussed in Social platforms is relevant for OTT platforms and Content aggregators as well. Content generation, personalization and Ad optimization can be a significant unlock. In fact AI based content generation can significantly reduce content curation and production costs. Generating personalized marketing content at scale is an additional lever that Content aggregators can pull for better acquisition and retention.  

Comments from Hotstar/ Inshorts

4. Media and publishing companies

While everything we’ve mentioned so far are applicable to Media and Publishing companies, we might miss interesting vertical specific plays if we use broad brush-strokes. Media and publishing (think Disney or ESPN) companies are not just sitting on archives of IP material but also create it at regular frequency (live sports, GOT). This can lead to very specific and yet large use-cases that will need bespoke solutioning and proprietary data access. Priding themselves on creative talent more than tech talent, these firms could create deep value pools for the right company. Best to delve deeper with a few sample categories.  

Size of prize and what it will take to win

These are very vertical (and company specific) solutions we are talking about. The best companies will be customer first, co-creating solutions with the Media/IP company. Once basic AI capabilities are developed, the GTM will include flying down to your customer and building out a workflow that not just works for them, but becomes an integral part of the way they operate.  

Shoutout to a few companies building deep vertical specific solutions

• Rizzle.com looks to work deeply with news and textual content generators to help them navigate the video first world.  
• Checkout Magnifi.ai for some inspiration on deep vertical solution for the Sports broadcasting vertical

What won’t change?

Good fun to let our imaginations run wild, but to build for the domain its equally important to think about what won’t change.  

• GenAI, in our view, will NOT make the creator redundant. It will empower her to create better, forge deeper audience relationships and potentially increase monetization.  
• Principles on core network loops and retention will remain relevant. If anything consumer relevance and engagement can now become even sharper.  
• Social platforms will remain a dominant force – however there may emerge newer platforms that leverage AI first ui/ux and consumption.  

• Concerns on safety, security and moderation will continue to be important to solve for.  

Conclusion: Waiting for Thor and his Mjölnir

AI and the modality support it offers is not an “alternative” for companies in Media and Social. It is quickly becoming a reality that can make or break a company. What can help you win?

1. Understand your customer and their workflows – the best solution combines the following

a. Part of their daily life – immersed fully into their workflows – ideally both in creation and distribution

b. Keeps ease of use central to all product design

c. Accesses proprietary data to build a moat / familiarity, thereby maximising chance to create cohesive, brand-first content

2. Solving for Sales - GTM is king

a. Whether you choose a distributed/fragmented customer set (creators/prosumers) or deep vertical solves (enterprise first) you are sure to enter a red ocean. Multiple teams across the world are working on modality and the team that wins will not just have the best tech, it will also create unfair customer access.

b. Evaluate ideas basis your ability to open doors and win.  

c. It isn’t winner-take-all… yet. Build your networks and create personal pull in the ecosystem of customers and developers to ensure you are top of mind for your customer set.  

3. Be flexible in building vs customizing models – Modality transformations are a work-in-progress

a. Create differentiated tech, but also leverage the best tech available. No points for proprietary model if it doesn’t work (or is too late to the market).

If you are building in AI first solutions in the space of Media and Social, feel free to reach us at chandrasekhar@matrixpartners.in tarun@matrixpartners.in aakash@matrixpartners.in . We’d love to exchange notes and whiteboard with you – share our learnings in the space and hopefully create clarity for the both of us J. We also crowdsource and maintain a repo of GenAI startups from India.

As a bonus: Here’s “Thor, the carpenter, hammering a nail” (courtesy Stable diffusion).  

Cheers,  

CV