Your Attention, Please: Optimizing for Humans
•
1 like•89 views
Your Attention, Please: Optimizing for Humans by John Feminella Operators of large distributed systems often try to optimize for scarce computational resources — memory, network bandwidth, and CPU cycles. However, research shows that if we want to maximize the health of our systems, there’s another scarce resource we should be prioritizing instead: the attention of our users. In this talk, I’ll show how we can use machine learning to improve how we present data and information and tell a story to our users, optimizing for their attention and holistic understanding, instead of overwhelming them with a barrage of low-level metrics. More intriguingly, we’ll see how this approach has produced massive improvements in several real-world systems at large Fortune 100 institutions. By the end of the talk, we’ll hopefully have convinced you that happier and more informed operators isn’t merely a good outcome, but rather is the most ethical thing to do if we don’t want to lead people astray.
Recommended
More Related Content
Similar to Your Attention, Please: Optimizing for Humans
Similar to Your Attention, Please: Optimizing for Humans (20)
More from J On The Beach
More from J On The Beach (20)
Recently uploaded
Recently uploaded (20)
Your Attention, Please: Optimizing for Humans
- 1. blockchains and distributed apps a detailed primer on the world’s most overhyped technology by: John Feminella at: Craft in: Budapest, Hungary on: May 10th, 2018
- 3. http://jxf.me · @jxxf Should I use a blockchain for my app? Probably not.
- 4. in the beginning http://jxf.me · @jxxf
- 5. http://jxf.me · @jxxf "A purely peer-to-peer version of electronic cash would allow online payments to be sent directly from one party to another without going through a financial institution. Digital signatures provide part of the solution, but the main benefits are lost if a trusted third party is still required to prevent double-spending. We propose a solution to the double-spending problem using a peer- to-peer network. The network timestamps transactions by hashing them into an ongoing chain of hash-based proof-of-work, forming a record that cannot be changed without redoing the proof-of-work. The longest chain not only serves as proof of the sequence of events witnessed, but proof that it came from the largest pool of CPU power. As long as a majority of CPU power is controlled by nodes that are not cooperating to attack the network, they'll generate the longest chain and outpace attackers."
- 6. http://jxf.me · @jxxf "A purely peer-to-peer version of electronic cash would allow online payments to be sent directly from one party to another without going through a financial institution. Digital signatures provide part of the solution, but the main benefits are lost if a trusted third party is still required to prevent double-spending. We propose a solution to the double-spending problem using a peer- to-peer network. The network timestamps transactions by hashing them into an ongoing chain of hash-based proof-of-work, forming a record that cannot be changed without redoing the proof-of-work. The longest chain not only serves as proof of the sequence of events witnessed, but proof that it came from the largest pool of CPU power. As long as a majority of CPU power is controlled by nodes that are not cooperating to attack the network, they'll generate the longest chain and outpace attackers."
- 7. http://jxf.me · @jxxf "A purely peer-to-peer version of electronic cash would allow online payments to be sent directly from one party to another without going through a financial institution. Digital signatures provide part of the solution, but the main benefits are lost if a trusted third party is still required to prevent double-spending. We propose a solution to the double-spending problem using a peer- to-peer network. The network timestamps transactions by hashing them into an ongoing chain of hash-based proof-of-work, forming a record that cannot be changed without redoing the proof-of-work. The longest chain not only serves as proof of the sequence of events witnessed, but proof that it came from the largest pool of CPU power. As long as a majority of CPU power is controlled by nodes that are not cooperating to attack the network, they'll generate the longest chain and outpace attackers."
- 8. http://jxf.me · @jxxf "A purely peer-to-peer version of electronic cash would allow online payments to be sent directly from one party to another without going through a financial institution. Digital signatures provide part of the solution, but the main benefits are lost if a trusted third party is still required to prevent double-spending. We propose a solution to the double-spending problem using a peer- to-peer network. The network timestamps transactions by hashing them into an ongoing chain of hash-based proof-of-work, forming a record that cannot be changed without redoing the proof-of-work. The longest chain not only serves as proof of the sequence of events witnessed, but proof that it came from the largest pool of CPU power. As long as a majority of CPU power is controlled by nodes that are not cooperating to attack the network, they'll generate the longest chain and outpace attackers."
- 9. http://jxf.me · @jxxf "A purely peer-to-peer version of electronic cash would allow online payments to be sent directly from one party to another without going through a financial institution. Digital signatures provide part of the solution, but the main benefits are lost if a trusted third party is still required to prevent double-spending. We propose a solution to the double-spending problem using a peer- to-peer network. The network timestamps transactions by hashing them into an ongoing chain of hash-based proof-of-work, forming a record that cannot be changed without redoing the proof-of-work. The longest chain not only serves as proof of the sequence of events witnessed, but proof that it came from the largest pool of CPU power. As long as a majority of CPU power is controlled by nodes that are not cooperating to attack the network, they'll generate the longest chain and outpace attackers."
- 10. http://jxf.me · @jxxf "A purely peer-to-peer version of electronic cash would allow online payments to be sent directly from one party to another without going through a financial institution. Digital signatures provide part of the solution, but the main benefits are lost if a trusted third party is still required to prevent double-spending. We propose a solution to the double-spending problem using a peer- to-peer network. The network timestamps transactions by hashing them into an ongoing chain of hash-based proof-of-work, forming a record that cannot be changed without redoing the proof-of-work. The longest chain not only serves as proof of the sequence of events witnessed, but proof that it came from the largest pool of CPU power. As long as a majority of CPU power is controlled by nodes that are not cooperating to attack the network, they'll generate the longest chain and outpace attackers."
- 11. before money http://jxf.me · @jxxf 0
- 13. a transaction http://jxf.me · @jxxf
- 17. http://jxf.me · @jxxf 🍰? ⌚?
- 18. http://jxf.me · @jxxf 🍰? ⌚?
- 19. http://jxf.me · @jxxf 🍰? ⌚?
- 20. http://jxf.me · @jxxf 😃 😃 ✓
- 21. a failed transaction http://jxf.me · @jxxf
- 22. http://jxf.me · @jxxf 🍰? ⌚?
- 23. http://jxf.me · @jxxf 🍰? ⌚?
- 24. http://jxf.me · @jxxf 🍰? ⌚? ✗
- 25. fixing a failed transaction http://jxf.me · @jxxf
- 26. http://jxf.me · @jxxf 🍰? ⌚?🐌?
- 27. http://jxf.me · @jxxf 😃 😃😃
- 28. “double coincidence of wants” http://jxf.me · @jxxf
- 30. http://jxf.me · @jxxf 🍰? ✗ ✗ ✗ ✓
- 32. http://jxf.me · @jxxf // money.ts class Money { amount: number currency: string }
- 33. what’s not money? http://jxf.me · @jxxf
- 35. http://jxf.me · @jxxf { ? }
- 41. http://jxf.me · @jxxf poor disintermediation poor fungibility low resiliency not easily measurable not a debt instrument
- 42. can we do better? http://jxf.me · @jxxf
- 43. fiat currency advocates think so http://jxf.me · @jxxf
- 44. http://jxf.me · @jxxf fiat: “it shall be”
- 47. transactions with money http://jxf.me · @jxxf
- 48. http://jxf.me · @jxxf 🍰? ⌚/$?
- 49. http://jxf.me · @jxxf 🍰? ⌚/$?
- 50. http://jxf.me · @jxxf 😃 😃 ✓
- 51. but money requires trust http://jxf.me · @jxxf
- 55. a recipe for money http://jxf.me · @jxxf
- 56. http://jxf.me · @jxxf trust rules history
- 57. making fiat currency http://jxf.me · @jxxf
- 58. http://jxf.me · @jxxf trust rules history
- 59. http://jxf.me · @jxxf trust rules history
- 60. http://jxf.me · @jxxf trust rules history
- 61. http://jxf.me · @jxxf trust rules + + history
- 62. http://jxf.me · @jxxf these details are abstracted from users
- 63. can we do better? http://jxf.me · @jxxf
- 64. http://jxf.me · @jxxf "A purely peer-to-peer version of electronic cash would allow online payments to be sent directly from one party to another without going through a financial institution."
- 65. cryptocurrency advocates think so http://jxf.me · @jxxf 🔒
- 66. http://jxf.me · @jxxf trust rules history 🔒
- 67. http://jxf.me · @jxxf trust rules history 𝐻 𝑡 ≡ ⋯ 🔒
- 68. http://jxf.me · @jxxf 𝐻 𝑡 ≡ ⋯ 🔒 0101 0010 1010 1110 0110 0101 1000 0011 trust rules history
- 69. http://jxf.me · @jxxf 0101 0010 1010 1110 0110 0101 1000 0011 𝐻 𝑡 ≡ ⋯ 🔒 trust rules history
- 70. http://jxf.me · @jxxf 0101 0010 1010 1110 0110 0101 1000 0011 trust rules history + +𝐻 𝑡 ≡ ⋯ 🔒
- 71. http://jxf.me · @jxxf these details are abstracted from users (?) 🔒
- 72. http://jxf.me · @jxxf each currency’s rules and history make it unique 0101 0010 1010 1110 0110 0101 1000 0011 rules history +
- 73. http://jxf.me · @jxxf blockchain as history is not very interesting 0101 0010 1010 1110 0110 0101 1000 0011 rules history +
- 74. http://jxf.me · @jxxf rules about how things work is what’s interesting 0101 0010 1010 1110 0110 0101 1000 0011 rules history +
- 75. a very different kind of money http://jxf.me · @jxxf
- 76. http://jxf.me · @jxxf 0101 0010 1010 1110 0110 0101 1000 0011 trust rules history 𝐻 𝑡 ≡ ⋯🔒
- 77. a very different kind of money http://jxf.me · @jxxf is it really money at all?
- 78. a very different kind of money http://jxf.me · @jxxf is it really money at all? we’ll come back to this
- 79. let’s build a cryptocurrency http://jxf.me · @jxxf
- 80. http://jxf.me · @jxxf // money.ts class Money { amount: number currency: string }
- 84. http://jxf.me · @jxxf A B C Alice paid Carol $15
- 85. http://jxf.me · @jxxf A B C Bob paid Alice $20 Bob paid Carol $30
- 86. shuffling cash is cumbersome http://jxf.me · @jxxf
- 87. can we do better? http://jxf.me · @jxxf
- 91. http://jxf.me · @jxxf our ledger records our transactions Ledger
- 92. Ledger http://jxf.me · @jxxf our ledger doesn’t run out of room
- 93. http://jxf.me · @jxxf Alice gets $100 Bob gets $100 Carol gets $100 (...) Ledger
- 94. http://jxf.me · @jxxf A B C Alice paid Carol $15
- 95. http://jxf.me · @jxxf our protocol describes how we use the ledger
- 96. http://jxf.me · @jxxf protocol: 1. anyone can add valid lines
- 97. http://jxf.me · @jxxf Alice gets $100 Bob gets $100 Carol gets $100 Alice paid Carol $15 (...)Ledger
- 98. http://jxf.me · @jxxf A B C Alice paid Carol $15 Ledger
- 99. http://jxf.me · @jxxf A B C Bob paid Alice $20 Bob paid Carol $30
- 100. http://jxf.me · @jxxf Alice gets $100 Bob gets $100 Carol gets $100 Alice paid Carol $15 Bob paid Alice $20 Bob paid Carol $30
- 101. http://jxf.me · @jxxf A B C Bob paid Alice $20 Bob paid Carol $30 Ledger
- 102. what about big transactions? http://jxf.me · @jxxf
- 103. http://jxf.me · @jxxf A B C Alice paid Carol $1,500 Ledger
- 104. http://jxf.me · @jxxf A B C Ledger Alice paid Carol $1,500 overspending!
- 105. http://jxf.me · @jxxf protocol: 1. anyone can add valid lines 2. settle up every week
- 109. http://jxf.me · @jxxf protocol: 1. anyone can add valid lines 2. no overspending
- 110. ledger is now its own currency (🔒) http://jxf.me · @jxxf
- 112. http://jxf.me · @jxxf Alice gets 🔒100 Bob gets 🔒100 Carol gets 🔒100 Alice paid Carol 🔒15 Bob paid Alice 🔒20 Bob paid Carol 🔒30 (...)
- 113. problem: malicious actors http://jxf.me · @jxxf
- 114. http://jxf.me · @jxxf Alice gets 🔒100 Bob gets 🔒100 Carol gets 🔒100 B
- 115. http://jxf.me · @jxxf Alice gets 🔒100 Bob gets 🔒100 Carol gets 🔒100 Carol paid Bob 🔒100 B
- 116. problem: can’t trust the ledger forgery: actors can add lines that aren’t valid erasure: actors can remove lines that are valid http://jxf.me · @jxxf
- 117. how do we stop forgery? http://jxf.me · @jxxf
- 118. securing the ledger http://jxf.me · @jxxf 3
- 119. http://jxf.me · @jxxf "Digital signatures provide part of the solution."
- 120. http://jxf.me · @jxxf Alice paid Carol 🔒15 Bob paid Alice 🔒20 Bob paid Carol 🔒30
- 121. http://jxf.me · @jxxf Alice paid Carol 🔒15 Alice Bob paid Alice 🔒20 Bob Bob paid Carol 🔒30 Bob
- 122. physical signatures can be copied http://jxf.me · @jxxf
- 123. http://jxf.me · @jxxf Alice paid Carol 🔒15 Alice Bob paid Alice 🔒20 Bob Bob paid Carol 🔒30 Bob Alice paid Bob 🔒50 Alice B
- 124. let’s use digital signatures instead http://jxf.me · @jxxf
- 125. http://jxf.me · @jxxf sign(⋯) sign lines to authorize transactions verify(⋯) verify signed lines to ensure signature is valid
- 127. http://jxf.me · @jxxf A public key (pk) secret key (sk)
- 128. http://jxf.me · @jxxf A public key (pk) secret key (sk)
- 129. http://jxf.me · @jxxf sign(m,sk) sign lines to authorize transactions secret key (sk)message (m) Alice paid Carol 🔒15
- 130. http://jxf.me · @jxxf sign(m,sk) sign lines to authorize transactions secret key (sk)message (m) Alice paid Carol 🔒15
- 131. http://jxf.me · @jxxf sign(m,sk) sign lines to authorize transactions
- 132. http://jxf.me · @jxxf sign(m,sk) sign lines to authorize transactions signature 0101 1101 1110 1010 0001 0111 [⋯] =
- 133. http://jxf.me · @jxxf changes in m produce unpredictable signatures sign(m’,sk) sign lines to authorize transactions signature 1010 1101 0010 0110 0111 0101 [⋯] =
- 134. http://jxf.me · @jxxf sign(m,sk) sign lines to authorize transactions verify(m,s,pk) sign lines to authorize transactions = signature 0101 1101 1110 1010 0001 0111 [⋯]
- 135. http://jxf.me · @jxxf sign(m,sk) sign lines to authorize transactions verify(m,s,pk) sign lines to authorize transactions = signature 0101 1101 1110 1010 0001 0111 [⋯]
- 136. http://jxf.me · @jxxf sign(m,sk) sign lines to authorize transactions verify(m,s,pk) sign lines to authorize transactions = signature 0101 1101 1110 1010 0001 0111 [⋯]
- 137. http://jxf.me · @jxxf sign(m,sk) sign lines to authorize transactions verify(m,s,pk) sign lines to authorize transactions = signature 0101 1101 1110 1010 0001 0111 [⋯]
- 138. http://jxf.me · @jxxf sign(m,sk) sign lines to authorize transactions signature 0101 1101 1110 1010 0001 0111 [⋯] verify(m,s,pk) sign lines to authorize transactions = = validation {true, false}
- 139. http://jxf.me · @jxxf sign(m,sk) sign lines to authorize transactions = signature 0101 1101 1110 1010 0001 0111 [⋯] virtually impossible very easy!
- 140. http://jxf.me · @jxxf sign(m,sk) sign lines to authorize transactions = signature 0101 1101 1110 1010 0001 0111 [⋯] virtually impossible very easy!
- 141. http://jxf.me · @jxxf sign(m,sk) sign lines to authorize transactions = signature 0101 1101 1110 1010 0001 0111 [⋯] computationally infeasible very easy!
- 142. http://jxf.me · @jxxf protocol: 1. anyone can add valid lines 2. no overspending 3. must sign lines to be valid
- 143. http://jxf.me · @jxxf Alice paid Bob 🔒50 0111⋯ B
- 144. http://jxf.me · @jxxf Alice paid Bob 🔒50 0111⋯ B
- 145. can no longer forge lines… http://jxf.me · @jxxf
- 146. http://jxf.me · @jxxf Alice paid Bob 🔒50 1011⋯ B
- 147. http://jxf.me · @jxxf Alice paid Bob 🔒50 1011⋯ Alice paid Bob 🔒50 1011⋯ B
- 148. http://jxf.me · @jxxf Alice paid Bob 🔒50 1011⋯ Alice paid Bob 🔒50 1011⋯ B
- 149. but can duplicate legitimate lines! http://jxf.me · @jxxf
- 150. http://jxf.me · @jxxf sign(m,sk) sign lines to authorize transactions secret key (sk) 1 Alice paid Carol 🔒15 message (m) add id to message
- 151. http://jxf.me · @jxxf 1 Alice paid Bob 🔒50 1011⋯ 2 Alice paid Bob 🔒50 1011⋯ B
- 152. http://jxf.me · @jxxf B 1 Alice paid Bob 🔒50 1011⋯ 2 Alice paid Bob 🔒50 1011⋯
- 153. http://jxf.me · @jxxf protocol: 1. anyone can add valid lines 2. no overspending 3. must sign lines to be valid 4. lines have unique identifiers
- 154. problem: ledger is centralized http://jxf.me · @jxxf
- 156. distributing the ledger http://jxf.me · @jxxf 4
- 159. broadcast our ledger updates http://jxf.me · @jxxf
- 162. http://jxf.me · @jxxf Ledger Ledger Ledger 2 Bob paid Alice 🔒50 1011⋯ B
- 163. http://jxf.me · @jxxf Ledger Ledger 2 Bob paid Alice 🔒50 1011⋯2 Bob paid Alice 🔒50 1011⋯2 Bob paid Alice 🔒50 1011⋯ LedgerB
- 164. http://jxf.me · @jxxf Ledger Ledger 2 Bob paid Alice 🔒50 1011⋯ 2 Bob paid Alice 🔒50 1011⋯ 2 Bob paid Alice 🔒50 1011⋯ LedgerB
- 165. http://jxf.me · @jxxf protocol: 1. anyone can add valid lines 2. no overspending 3. must sign lines to be valid 4. lines have unique identifiers 5. distribute the ledger
- 166. problem: what order to use? http://jxf.me · @jxxf
- 167. http://jxf.me · @jxxf 2 Alice paid Dave 🔒30 1001⋯ 2 Alice paid Carol 🔒30 0101⋯ LedgerB
- 168. http://jxf.me · @jxxf 2 Alice paid Dave 🔒30 1001⋯ 2 Alice paid Carol 🔒30 0101⋯ LedgerB double spending!
- 169. http://jxf.me · @jxxf protocol: 1. anyone can add valid lines 2. no overspending 3. must sign lines to be valid 4. lines have unique identifiers 5. distribute the ledger (how?!)
- 170. trusting the ledger http://jxf.me · @jxxf 5
- 175. how do we know it’s valid? http://jxf.me · @jxxf
- 177. http://jxf.me · @jxxf sign(m ⋯)
- 178. http://jxf.me · @jxxf sign(m + n) nonce
- 179. http://jxf.me · @jxxf sign(m + n) 01011101101⋯nonce
- 180. http://jxf.me · @jxxf sign(m + n) 01011101101⋯nonce
- 181. http://jxf.me · @jxxf sign(m + n) 01011101101⋯ ~1 2nonce
- 182. http://jxf.me · @jxxf sign(m + n) 00101101101⋯ ~1 4nonce
- 183. http://jxf.me · @jxxf sign(m + n) 00001101101⋯ ~ 1 16nonce
- 184. http://jxf.me · @jxxf sign(m + n) 000000000000⋯ ~ 1 2 𝑛nonce
- 188. http://jxf.me · @jxxf signature nonce 175 <network> paid Dave 🔒50 1001⋯
- 191. http://jxf.me · @jxxf signature nonce previous signature nonce previous signature nonce previous
- 192. http://jxf.me · @jxxf signature nonce previous signature nonce previous signature nonce previous
- 193. consensus: trust longest chain http://jxf.me · @jxxf
- 199. building with blockchains http://jxf.me · @jxxf 6
- 200. http://jxf.me · @jxxf Should I use a blockchain for my app? Probably not.
- 201. can I wait a while for consensus? http://jxf.me · @jxxf
- 202. is this better for me than a database? http://jxf.me · @jxxf
- 203. does my app need to be distributed? http://jxf.me · @jxxf
- 204. do I need to trust people? http://jxf.me · @jxxf
- 205. blockchain challenges http://jxf.me · @jxxf 7
- 206. a very different kind of money http://jxf.me · @jxxf is it a good kind of money?
- 207. cryptocurrencies aren’t currency… http://jxf.me · @jxxf
- 208. cryptocurrencies aren’t currency… right now http://jxf.me · @jxxf
- 209. http://jxf.me · @jxxfblockchain.info price graph [source]
- 210. http://jxf.me · @jxxfblockchain.info price graph [source] $7,000 $1,000 2014 2017
- 211. deflation discourages production http://jxf.me · @jxxf
- 216. http://jxf.me · @jxxf $ $$$
- 217. http://jxf.me · @jxxf $ $$$ 🔒 🔒
- 218. http://jxf.me · @jxxf $ $$$ 🔒 🔒
- 219. PoW is ecologically problematic http://jxf.me · @jxxf
- 220. http://jxf.me · @jxxfVice Motherboard [source]
- 221. the future http://jxf.me · @jxxf 7
- 222. http://jxf.me · @jxxf 0101 0010 1010 1110 0110 0101 1000 0011 trust rules history + +𝐻 𝑡 ≡ ⋯ 🔒
- 223. http://jxf.me · @jxxf 0101 0010 1010 1110 0110 0101 1000 0011 trust rules history + +𝐻 𝑡 ≡ ⋯ 🔒
- 224. http://jxf.me · @jxxf 0101 0010 1010 1110 0110 0101 1000 0011 trust rules history + +𝐻 𝑡 ≡ ⋯ 🔒
- 225. takeaways blockchain applications offer trustless, permissionless rules cryptocurrency lets us trustlessly exchange value many protocols have some serious problems blockchains have enormous potential … if we get it right http://jxf.me · @jxxf
- 226. takeaways blockchain applications offer trustless, permissionless rules cryptocurrency lets us trustlessly exchange value many protocols have some serious problems blockchains have enormous potential … if we get it right http://jxf.me · @jxxf
- 227. takeaways blockchain applications offer trustless, permissionless rules cryptocurrency lets us trustlessly exchange value many protocols have some serious problems blockchains have enormous potential … if we get it right http://jxf.me · @jxxf
- 228. takeaways blockchain applications offer trustless, permissionless rules cryptocurrency lets us trustlessly exchange value many protocols have some serious problems blockchains have enormous potential … if we get it right http://jxf.me · @jxxf
- 230. 3px shadow should be visible above shadow test ⇖ bleed ⇗ bleed ⇘ bleed⇙ bleed 28: λx.t(λt).x 36: λx.t(λt).x 44: λx.t(λt).x http://jxf.me · @jxxf Sample purple text. Sample blue text. Sample green text. Sample yellow text. Sample orange text. Sample red text.