Christopher olston google research
Download
1 / 4

Christopher Olston Google Research - PowerPoint PPT Presentation


  • 47 Views
  • Uploaded on

We can be at the center of AI 2.0. Christopher Olston Google Research. AI is getting its groove back. ... l argely thanks to Big Data e.g. Watson, Siri , Google Translate Building Big-AI systems is easy , thanks to scalable data management building blocks

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about ' Christopher Olston Google Research' - mary


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
Christopher olston google research

We can be at the center of AI 2.0

Christopher Olston

Google Research


Ai is getting its groove back
AI is getting its groove back

  • ... largely thanks to Big Data

  • e.g. Watson,Siri, Google Translate

  • Building Big-AI systems is easy, thanks to scalable data management building blocks

    • BigTable, Map-Reduce, Pregel, …

  • Life is good


Not really
NOT REALLY …

  • Life of a Big-AI project:

    • Commit to an algorithm

    • Bust it up into map functions, co-processors, ...

    • Optimize the crap out of it:

      • Caching, batching

      • Indexing, clever encoding

      • “Stupid map-reduce tricks”

    • Never ever disband the project (who else could understand the debris field that is your code?)

    • To keep entertained while you maintain your ossified code:

      • read papers about new algorithms and muse “it would be cool if we could try that”


We need higher level programming a bstractions
We Need Higher-Level Programming Abstractions

  • But unlike SQL etc.:

    • Power: Turing complete

    • Syntax: Math should look like math

    • Control: Physical transparency

  • Declarative programs that “just work” on small data (for experimentation, debugging)

  • Target scalable platforms (e.g. map-reduce), and choose optimizations to apply, via operational-style annotations


ad