Thursday, November 01, 2018

07.11.2017 Andrew Ng - The State of Artificial Intelligence (English)

Tech company in AI are (according to Andrew Ng):

* Strategic data acquisition
* Unified data warehouse
* Pervasive automation
* New job descriptions

Ниже есть продолжение.

Rule of thumb:
Anything that typical person can do with less than a second of thinking we can assume automate.

It has false positive and false negatives.

For example, security guard watching a security video. What is their job? They are looking at the video,
figure out where the people, that’s one-second task,
figure out where the people standing, that takes a second,
figure out where the faces of these people, that's a second,
figure of any of these people that you see are a known bad guy.

This job is the sequence of the one second tasks and thus it is in significant risk of automation.

Why deep learning is taking off just now?

* We start to collect more data, because digitization of society.
* Having computers fast enough, rise of GPU computing, to train very large neural networks.

What is really most valuable today?

* Supervised learning (A->B)
* Transfer learning
* Unsupervised learning
* Reinforcement learning

What is driving economy today is supervised learning. There is rapid fall of as you go down this list
In terms of economic value created today.

No comments:

Post a Comment