We've finally entered the world of HAL, where voice activated services like Siri, Cortana and Alexa attempt to cater to our needs. Within a few decades, synthetic "brains" will be able to execute 100 trillion instructions per second--an activity level that rivals the human brain. That brings the prospect of helpful servants like Cerebro and Lieutenant Commander Data--or perhaps malevolent entities like Brainiac and Skynet. Our panelists will discuss the A.I. capabilities soon to arrive and the opportunities and threats they will bring to humanity. Speakers include Greg Corrado (Senior Research Scientist, Google), Lucas Ives (Engineering Lead, PullString Inc.), Brandon Wirtz (CEO and founder, Recognant), Sarah Austin (CEO and founder, Peak Energies), Nigel Duffy (CTO, Sentient Technologies), and moderator Danny Sullivan (founding Editor of Search Engine Land). Room 25ABC
1. Beyond Human:
The Rise Of Machine Intelligence
Lucas Ives
PullString
@Lri - @PullStringInc
Nigel Duffy
Sentient Technologies
n/a - @sentientdai
Sarah Austin
Broad Listening
@sarahaustin
@BroadListening
Brandon Wirtz
Recognant
@bwops - n/a
Greg Corrado
Google
n/a - @googleresearch
Danny Sullivan
Search Engine Land
@dannysullivan
@sengineland
#SDCCAI
Thanks: @Sbroback & @Dent
2. •Industry leading platform for building computer conversation
•Author-driven
•Leverages advantages of AI, ML, NLP, ASR
•Direct support for team-based workflows
•Write once, publish anywhere
•Scalability, analytics, enterprise security
31. List of artificial intelligence films
https://en.wikipedia.org/wiki/List_of_artificial_intelligence_films
Which movies get artificial
intelligence right?
http://www.sciencemag.org/news/2015/07/which-movies-get-artificial-intelligence-
right
How Machine Learning Works,
As Explained By Google
http://marketingland.com/how-machine-learning-works-150366
Image credits: Wikipedia & Google
32. Beyond Human:
The Rise Of Machine Intelligence
Greg Corrado
Google
n/a - @googleresearch
Lucas Ives
PullString
@Lri - @PullStringInc
Sarah Austin
Broad Listening
@sarahaustin
@BroadListening
Nigel Duffy
Sentient Technologies
n/a - @sentientdai
Brandon Wirtz
Recognant
@bwops - n/a
Danny Sullivan
Search Engine Land
@dannysullivan
@sengineland
#SDCCAI
Thanks: @Sbroback & @Dent
Hinweis der Redaktion
find photos using text search, or face clustering
quick, context-sensitive replies to email
dedupe cleans noisy/bad data
Google Photos - Back in the day of film cameras, or even digital cameras, we took a few pictures could afford the time to organize our photos. With the rise of mobile phones and higher capacity memory cards, we take far more pictures than we can curate. Image clustering (e.g. each family member) and search for untagged photos allows us to discover images in our collection.
Smart Reply - Typing on a phone is hard, and we are often rushed by the deluge of email. Having prepared answers allows to tackle email more efficiently.
Dedupe - Our data might have a lot of duplicate locations, for example two adjacent Starbucks locations. Except if you’re in Seattle, you may not want to dedupe as there really are that many Starbucks!
find photos using text search, or face clustering
quick, context-sensitive replies to email
dedupe cleans noisy/bad data
Google Photos - Back in the day of film cameras, or even digital cameras, we took a few pictures could afford the time to organize our photos. With the rise of mobile phones and higher capacity memory cards, we take far more pictures than we can curate. Image clustering (e.g. each family member) and search for untagged photos allows us to discover images in our collection.
Smart Reply - Typing on a phone is hard, and we are often rushed by the deluge of email. Having prepared answers allows to tackle email more efficiently.
Dedupe - Our data might have a lot of duplicate locations, for example two adjacent Starbucks locations. Except if you’re in Seattle, you may not want to dedupe as there really are that many Starbucks!
Pedestrian detection isn’t the only component of a self-driving car, but it is critical to understand what is a pedestrian. This photo was taken around Halloween 2015 on the top level of the parking garage at the X office, where they test some of the cars. Understanding that pedestrians may not all have humanoid forms is important when the streets flood with kids every Halloween.
We have build programs so that machine can learn how to play Atari games from the raw pixels and a score. And now, we’re tackled the most complex game - Go. In October 2015, AlphaGo beat the European champion, Fan Hui (whom we subsequently hired as advisor). Everyone thought that the “solving” Go would take many more decades. In March 2016, we took on the leading world player Lee Sedol. AlphaGo won 4-1. This represents a significant achievement in AGI.
Go is a ~3000 year old game from China. Rules are very simple, but the number of game configurations is vast, ~641 orders of magnitude more than chess (10e761 vs 10e120). It is impossible to compute all possible configurations - there are more than the number of atoms in the universe.
Chess - Deep Blue - 1997
Jeopardy - Watson - 2011