AI drive spawns new takes on chip design

As soon as we solve machine
learning we will fix printer.

It is has been interesting to see a re-mergence in interest in new chip architectures. Just when you think it is all been done and there's nothing new. Bam! For sure.

The driver these days is A.I. but more particularly the machine learning aspect of AI. GPUs jumped out of the gamer console and onto the Google and Facebook data center. But there was more in the way of hardware tricks to come. The effort is to get around the tableau I here repeatedly cited: the scene is the data scientist sitting there thumb twiddling while the neural machine slowly does its learning.

I know when I saw that Google had created a custom ASIC for Tensor Flow processing, I was taken aback. If new chips are what is needed to succeed in this racket, it will be a rich man's game.

Turns out a slew of startups are on the case. This article by Cade Metz suggests that at least 45 startups are working on chips for AI type applications such as speech recognition and self-driving cars. It seems the Nvidia GPU that has gotten us to where we are, may not be enough going forward. Co processors for co processors, chips that shuttle data about in I/O roles for GPUs, may be the next frontier.

Metz names a number of AI chip startups: Cerbras, Graphcore, Wave Computing, Mythic, Nervana (now part of Intel). - Jack Vaughan

Related
https://www.nytimes.com/2018/01/14/technology/artificial-intelligence-chip-start-ups.html

What is the risk of AI?

Happy New Year from all of us at the DataDataData blog. Let's start the year out with a look at Artificial Intelligence - actually a story thereof. That is, "Leave Artificial Intelligence Alone" by Andrew Burt, appearing in last Friday's NYTimes' Op-Ed section.

Would that people could leave AI alone! You cant pick up a the supermarket sales flyer without hearing someone's bit on the subject. As Burt points out, a lot of the discussion is unnecessarily - and unhelpfully - doomy and gloomy. Burt points out that AI lacks definition. You can see the effect in much of the criticism, which lashes out with haymakers at a phantom - one that really comprises very many tributary technologies - quite various ones at that.

Some definition, some narrowing of the problem scope is in order.

If you study the history of consumer data privacy you discover, as Burt reminds, the Equal Credit Opportunity Act of 1974. Consider it as a pathway for data privacy that still can be followed.

Burt also points to SR 11-7 regulations that are intended to provide breadcrumbs back to how and why trading models were constructed, so that there is good understanding of risk involved in the automated pits of Wall Street.

Within the United States’ vast framework of laws and regulatory agencies already lie answers to some of the most vexing challenges created by A.I. In the financial sector, for example, the Federal Reserve enforces a regulation called SR 11-7, which addresses the risks created by the complex algorithms used by today’s banks. SR 11-7’s solution to those challenges is called “effective challenge,” which seeks to embed critical analysis into every stage of an algorithm’s life cycle — from thoroughly examining the data used to train the algorithm to explicitly outlining the assumptions underlying the model, and more. While SR 11-7 is among the most detailed attempts at governing the challenges of complex algorithms, it’s also one of the most overlooked.

Burt sees such staged algorithm analysis as a path for understanding AI and machine learning going forward.

It is good to see there may be previous experience that can be tapped when looking at how to handle AI decision making - as opposed to jumping up and down and yelling 'the sky is falling.'

As he says, it is better to distinguish the elements of AI application according to use cases, and look at regulation specifically in verticals - where needed. 

Spoke with Andrew Burt last year as part of my work for SearchDataManagement - linked to here: Machine learning meets Data Governance.  - Jack Vaughan