Google I/O 2021: horizontal integration and machine learning

Google I/O 2021 comes in light of one of the strongest quarters in the company’s history. And, to no surprise, the threads we already saw in the Q1 earnings call continue throughout I/O: retail and productivity.

Google I/O 2021 comes in light of one of the strongest quarters in the company’s history. And, to no surprise, the threads we already saw in the Q1 earnings call continue throughout I/O: retail and productivity.

The keynote took roughly 115 minutes. When we look at how much time each topic occupied, we get a glance into Alphabet’s priorities:

  • Intro: 3.5%
  • Workspace: 18%
  • Safer with Google: 6.5%
  • Helpful information: 19.3%
  • Design & Android: 18%
  • Health: 6%
  • Sustainability: 3%

This agenda highlights Google’s focal areas: Search, productivity, privacy & security.

Search: E-A-T, MUM, and the Shopping Graph

Google Search innovation seemed a bit thin during I/O. It feels a bit like Search is working as intended and only needs polishing here and there. It’s still the company’s cash cow, but not where Google finds exponential growth.

Only three points stood out: E-A-T, MUM, and the Shopping Graph.

Since the 2016 election, when the term “fake news” gained rapid traction, trustworthy results and expertise have become important for Google. In SEO, we know this concept as E-A-T.

The only true change is coming to “about this result”, where Google will also factor in what other sources are saying about a site. E-A-T exists on the site level, driven by the entity graph and enriched by reviews from trustworthy sources.

MUM, the Multitask Unified Model, is a multimodal, transformer-based NLU engine that gives users answers from different search verticals and harvests information from over +75 languages. It can understand implicit meaning and context 1000x better than BERT.

The impressive part about MUM is not the performance improvement but multimodality. In fact, this term was used several times throughout the keynote, also in the context of the Shopping Graph. Google pulls information from many different sources together: Lens, Photos, Youtube, Search, Gmail, Youtube, Maps, and Shopping.

Thinking back about the essential user intents informational, navigation, transactional - though, I’m not a fan of this simplistic view - what Google is building is a vertical product for each intent:

Informational = Web search, Youtube,

Navigational = Google maps

Transactional = Shopping Graph

Shopping Graph, a reflection of products across all Alphabet properties, is a perfect example of integration: Shopping will be connected to Google Lens, Youtube, Images, Photos, and Chrome. Google will connect loyalty programs, display shopping carts on the Chrome homepage, and identify products in your photos.

Google Workspace: integration and machine learning

“Work is no longer a place”. What a perfect reflection of reality and caption of the Zeitgeist!

The pandemic accelerated home office work and retail by lightyears. Google has a stake in both. “G Suite” is now “Workspace” and integration of

Productivity is a fast-growing market. By stringing many products like Gmail, Meet, Calendar, Drive, Docs, Slides, Sheets, Keep, and more together, Google now directly competes with:

  • Atlassian
  • Notion
  • Coda
  • Microsoft 365
  • Airtable
  • Zoom
  • Asana
  • Dropbox
  • ...

Once again, Google pulls its freemium leverage:

Besides a lower barrier to entry, freemium products push data network effects. Putting a part of the product in front of the paywall allows you to learn and understand what users love and how they use the product. It gives product teams a chance to iterate faster, and Marketing teams to learn what works. Some companies go as far as launching a completely free product to quickly iterate or Product/Market-Fit and then start charging.

The network effect Google leverages is integrating all these products with each other, a classic aggregator playbook used by Apple (iPhone + iMessage + iWatch + Mac + iPad),  Amazon (Prime + Video + Delivery + Music), or Facebook (Facebook + IG + Whatsapp + Messenger).

I call this “Platform Confluence”, a powerful network effect:

Google, Facebook, Amazon, and Co started out as single platform (search, social, online retail), then built or bought more products (WhatsApp, Gmail, Twitch), and now tie them together into ecosystems. The power of such an ecosystem consists of tracking user signals – whether intent or behavioral – to show better ads and use the returns to create more ways to track user signals.”

But Platform Confluence can also lead to antitrust issues.

Privacy & Security: the aggregator’s Achilles Heel

Google is an aggregator that provides a free, superior user experience to channel demand. As such, the two natural predators of aggregators are government regulation and mistrust.

If users believe Google’s Search results are fake or that the company “sells their data” (whatever that means) and Google loses trust, it will lose its network effects. It’s the demand that gives aggregators power over supply.

As a result, it wasn’t surprising that I/O contained many displays of trust and safety: in the part about Android, Shopping Graph, and Search.

Google in 2021: all about integration, made possible by machine learning

Two threads that pulled through almost every section of the Google I/O keynote: integration and machine learning.

Google wants to bring its product closer together and harvest synergies (1+1=3), not just on the software and cloud level but also in hardware. Android phones are your remote control, unlock your car, and are the bridge between Google Lens and your personal Cloud.

All that is powered by AI: LaMDA, a new conversational technology built on transformers and GPT-3, and MUM, a multi-task NLU engine that flawlessly understands images, videos, text, and audio.

I/O 2021 was exciting. I’m personally pumped about Google’s productivity track. At the same time, shadows of antitrust and mistrust were lingering over the Shoreline Theater.