## Fin World Tour 2017 --- ###### Agenda Who are we / why did we start Fin? -- What is Fin? -- How does Fin work under the hood? -- --- ###### Why? The internet is broken... -- as a knowledge machine. -- It's a very efficient entertainment machine. --- ###### What is Fin? --- Fin is a **'work engine'** capable of performing **heterogeneous knowledge work**.
🤔 --- ###### (1) What is heterogeneous knowledge work? --- _Can you setup a dinner at my apt every Thu and invite 2 people from the team? Tell them to bring an SO if they want._ --- _Can you pay this electric / gas bill?_ --- _Can you restock the dry erase markers for the office?_ --- _Can you get me and Rob a dinner res for 2 for Mon night?_ --- ###### ... any work that can be done with phone + email + internet connection. --- ###### (b) What is a 'work engine'? --- ###### A Search Engine... → Query -- → Search Index -- → Result --- ###### A Work Engine... → Request -- → Perform Actions -- → Result --- ###### An example... --- _Can you get me and Rob a dinner res for 2 for Mon night?_ --- ###### What would Siri do?
Pretty good, but far from human-level work. --- ###### This is kind of like when you text a friend... You: -- _Hey Jenny, what's the formula for computing compounding interest?_ --
Jenny: -- _JFGI https://www.google.com/search?q=formula+for+compound+interest_ --- ###### What would Fin do? _Can you get me and Rob a dinner res for 2 for Mon night?_ --
Fin: -- _I got you and Rob a reservation at your usual spot, Nopa, for 10pm. I've added to both of your calendars and let Rob know via email._ --
😲 --- ###### How people think Fin works → Request 🤔 -- → AI + Deep Learning Black Box -- → Result 🎉 -- --- class: smaller ###### How Fin Actually Works → Request -- → Categorize -- → Estimate & Prioritize -- → Understand Context -- → Plan -- → Do 'Real' World Work ☎️ 📩 📅 -- → Respond -- → Learn --- ###### Fin combines human and machine intelligence _meh, you use humans?_ --
_how does that scale?_ --- ###### But, let's consider what's inside the ⬛️... - Human Knowledge Workers 👩👱 -- - Lots of Measurement ⚖️⏰📊 -- - Operations + Logistics! 🤓👵 -- - Well Typed, Semantic Knowledge Graph 📕📈 -- - Contextual Search Layer 🔦 -- - ML 🤖 --- ###### Back to Our Example --- ###### → Request _Can you get me and Rob a dinner res for 2 for Mon night?_ --
_From:_ -- _andrew.kortina@gmail.com_ --- ###### → Categorize, Estimate, Prioritize (🤖) Restaurant, Dinner, Reservation -- (👱) In-Person -- (🤖) Should Take: ~6 Minutes -- (🤖) Should Be Done By: EOD --- ###### → Understand Context (🤖) Rob = Rob Cheung -- (🤖) Best email for Rob: robc12345@gmail.com -- (🤖) Favorite Restaurant = Nopa -- (🤖) Eats dinner late, after 10pm -- ... --- ###### Do 'Real' World Work ☎️ 📩 📅 (👩) Call Nopa -- --- ###### → Respond _(👵) I got you and Rob a reservation at your usual spot, Nopa, for 10pm. I've added to both of your calendars and let Rob know via email._ --- ###### → Learn (🤖) Improve Models: Context, Prioritization, Prediction -- (🎓) Add World & Personal Knowledge to Graph -- (👩) Help People Continuously Improve --- We believe all work will be performed by human + software hybrid systems in the near future.
It's hard tho! --- ## Fin kortina@finxpc.com