Skip to main content

Artificial Intelligence - Expert Perspectives

Sebastain Thrun:
  • Artificial Intelligence represents an opportunity for lessons learned to be applied on all systems and including systems not made yet.
  • Business leaders often want to go a certain direction but available toolsets including education did not keep up.
  • Anything repetitive would likely be done by AI within 20 years.

 Erik Brynjolfsson. Director, MIT Initative on Digital Economy:

  • Book: The Second Machine Age
  • Three trends:
    • Power
    • Data
    • Algorithms
  • Machines with very narrow specialty.
  • Machine learning - when to use and when not to use. Find the problem then find the solution. Identify the questions.
  • Image recognition - machines are way better.
  • Lots of churn under the surface so you have to be working on opportunities even though it looks like nothing is happening.
  • Examples:
    • A team looked flight patterns and was able to predict merger and acquisitions. 
    • A team used logs from locomotives to identify improvements to be made.
    • Finance showed massive improvements - 360,000 hours into few seconds (JP Morgan).
  • Need to watch ethics and privacy issues.

Gautam Khera. AI Executive Leader in Fortune 500 companies:

  • Every organization within a company should be thrilled about the AI journey.

Popular posts from this blog

CPU Model in Linux

 Ever wondered what kind of CPU model you are using in Linux? This command will pull it up: gcc -march=native -Q --help=target|grep march
Read more

AI Key Terms and Concepts Notes

Mere notes Model: A mathematical representation of a problem, situation, phenomenon, or process Training: The process of using an algorithm to create a model from a set of data Training Set: A subset of our ground truth data that our model will learn from Algorithm: A procedure, or set of steps Machine learning (probabilistic approach) vs Rules-based system If the universe of possible outcomes is well delineated then stick with rules. Purpose of Ground Truth Data Ground truth data helps many learning systems to learn and evaluate their own performance by providing a gold standard on what the truth actually is. Source: Udacity AI for Business Leaders
Read more

Supervised, Unsupervised, Semi-supervised, and Reinforcement Learning

 https://blogs.nvidia.com/blog/2018/08/02/supervised-unsupervised-learning/  Source: Udacity AI for Business Leaders Supervised Learning - A category of machine learning which relies on the proper values, or labels , being present for the output data in the ground truth dataset, which the model can “learn” from during its training process. Labels - Values of the output variable/column/piece of data you are interested in Unsupervised Learning - A category of machine learning which uses the underlying characteristics of data itself, rather than already-supplied labels for outputs, to inform the model’s training process. Often uses segmenting or clustering algorithms to determine the output values. Reinforcement Learning - A category of machine learning which uses a series of simulations/cycles to reward optimal behaviors to learn policies for decision making in complex scenarios such as autonomous driving.
Read more