Blog: Barney Finucane Subscribe to this blog's RSS feed!

Barney Finucane

Welcome to my BeyeNETWORK blog. My main goal here is to address hype issues that come up in the Internet, not to provide any overview of the BI market itself. I look forward to any questions or comments you may have.

About the author >

Barney Finucane has extensive experience in the BI industry. As a consultant, he has supported companies in the chemical, energy and manufacturing sector with the introduction of BI software. As product manager for the company MIS, he was responsible for the front-end products Plain and onVision, and kept a keen eye on projects and tools from other vendors. His areas of speciality include tool selection, quality assurance for BI, data warehouse strategies and their architectures.

Since renewable energy has started to become popular, an odd problem has appeared in wholesale energy markets: negative prices.

In other words, energy plants sometimes pay their customers to take energy off their hands. Usually older, less flexible plants that can't shut down without incurring costs are affected.
One solution to this problem is batteries. The idea is to store the energy when it is overabundant, and use it later when it is expensive. This is sometimes called "peak shaving".
Batteries are a great idea, but not the only solution. Another is to simply find an application that is energy hungry and can be run intermittently.

One possible application for soaking up excess energy is desalination. For example, a desert region near an ocean could build solar plants to desalinate water during the day only. The question is whether building a desalination plant that only runs 12 hours a day is worth the savings in energy. 
Another way to make use of energy that might go to waste is using it to power computers that perform analytics. The energy demand of data centers is growing quickly.

One source of energy needs is Bitcoin. Bitcoin mining consumes huge amounts of energy, so it is a great example of a use for negative energy prices. In fact there are already a lot of bitcoin miners in Western China, where solar and wind installations have outstripped grid upgrades. In these areas renewable energy is often curtailed because the grid can't keep up. So the energy is basically free to the miners.

Extremely cheap bitcoin mining arguably undermines the whole concept, but here is a more productive idea: Training artificial intelligence. For example, have a look at this link to gcp leela, a clone of Google Deepmind Alphago zero:
The entire source code is free, and it's not a lot of code. But that free code is just the learning model, and its based on well known principles. It's probably just as good as Deepmind Alphago Zero when trained, but they figure it would take them 1700 years to train -- unless of course they could harness other resources. This is partly because they don't have access to Google's specialized TPU hardware. Whatever the reason, training it is going to burn through a lot of energy.
This would be a great application for negatively priced energy. Game playing is more a stunt than a commercial application, but when they are paying you to use the energy, why not? And as time passes, more useful AI apps will need training.
So it gets down to whether the business model of peak shaving with batteries makes more economic sense than banks of custom chips for training neural networks for AI in batches. The advantage of batteries is that you can sell the energy later for more, but it's not terribly efficient, and using it directly is a better idea. Cheap computer hardware and a growing demand for AI may fit this niche very well.
This puts a whole new twist on the idea that big tech companies are investing in renewables. These companies make extensive used of AI, which is trained in batch processes. 

Posted January 4, 2018 8:33 AM
Permalink | No Comments |
Artificial neural networks are computer programs that learn a subject matter of their own accord. So an artificial neural network is a method of machine learning. Most software is created by programmers painstakingly detailing exactly how the program is expected to behave. But in machine learning systems, the programmers create a learning algorithm and feed it sample data, allowing the software to learn to solve a specific problem by itself.

Artificial neural networks were inspired by animal brains. They are a network of interconnected nodes that represent neurons, and the thinking is spread throughout the network. 

But information doesn't fly around in all directions in the network. Instead it flows in one direction through multiple layers of nodes from an input layer to an output layer. Each layer gets inputs from the previous layer and then sends calculation results to the next layer. In an image classification system, the initial input would be the pixels of the image, and the final output would be the list of classes.

The processing in each layer is simple: Each node get numbers from multiple nodes in the previous layer, and adds them up. If the sum is big enough, it sends a signal to the nodes in the layer below it. Otherwise it does nothing. But there is a trick: The connections between the nodes are weighted. So if node A sends a 1 to nodes B and C, it might arrive at B as 0.5, and a C as 3, depending on the weights in the connections. 

The system learns by adjusting the weights of the connections between the nodes. To stay with visual classification, it gets a picture and guesses which class it belongs to, for example "cat" or "fire truck". If it guesses wrong, the weights are adjusted.This is repeated until the system can identify pictures.

To make all this work, the programmer has to design the network correctly. This is more an art than a science, and in many cases, copying someone else's design and tweaking it is the best bet.

In practice, neural network calculations boil down to lots and lots of matrix math operations as well at the threshold operation the neurons use to decide whether to fire. It's fairly easy to imagine all this as a bunch of interconnected nodes sending each other signals, but fairly painful to implement in code. 

The reason it is so hard is that there can be many layers that are hard to tell apart, making it easy to get confused about which is doing what. The programmer also has to keep in mind how to orient the matrices the right way to make the math work, and other technical details. 

It is possible to do all this from scratch in a programming language like Python, and recommended for beginner systems. But fortunately there is a better way to do advanced systems: In recent years a number of libraries such as Tensorflow have become available that greatly simplify the task. These libraries take a bit of fiddling to understand at first, and learning how to deal with them is key to learning how to create neural networks. But they are a huge improvement over hand coded systems. Not only do they greatly reduce programming effort, they also provide better performance.

Posted January 3, 2018 3:21 PM
Permalink | No Comments |
There has been quite a bit of discussion about whether Alphago can learn from the games it plays against Lee Sedol. I think not. At least, not directly. 
The heart of the program is the ”policy network” a convolutional neural network (CNN) that was designed for image processing. CNNs return a probability that a given image belongs to each of a predefined set of classifications, like ”cat”, ”horse”, etc. CNNs work astonishingly well, but have the weakness that they can only be used with a fix size image to estimate a fixed set of classifications.
The policy network views go positions as 19í—19 images and returns probabilities that human players would make one of 361 possible moves. This probability drives with the Monte Carlo tree search for good moves that has been used for some time in go computers.The policy network is trained on 30 million positions (or moves) initially. 
CNN (aka ”deep learning”) behavior is pretty well understood. The number of samples required for learning depends on the complexity of the model. A model of this complexity probably requires tes of thousands of example positions before it changes much. 
The number of samples required to train any machine learning program depends on the complexity of the strategy, not on the number of possible positions. For example, Gomoku ("five in a row", also called goban) on a 19í—19 board would take many fewer examples to train than go would, even though the number of possible positions is also very large.
Another point: Any machine learning algorithm will eventually hit a training limit, after which it won’t be able to improve itself by more training. After that, a new algorithm based on a new model of game play would be required to improve the play. It is interesting that the Alphago team seems to be actively seeking ideas in this area. Maybe that is because they are starting to  hit a limit, but maybe it's just because they are looking into the future.
So Alphago probably can’t improve its play measurably by playing any single player five times, no matter how strong. That would be ”overfitting”. The team will be learning from the comments of the pro players and modifying the program to improve it instead.
Interesting tidbit: Alphago said the chances of a human playing move 37 in game 2 was 1 in 10,000. So the policy network doesn’t decide everything.

Posted March 13, 2016 1:13 PM
Permalink | No Comments |
The key part of Alphago is a convolutional neural network. These are usually used for recognizing cat pictures and other visual tasks, and progress in the last five years has been incredible.
Alphago went from the level of a novice pro last October to world champion level for this match. It did so by playing itself over and over again.
Chess programs are well understood because they are programmed by humans. Alphago is uses an algorithm to pick a winning move in a given go position. But the heart of the program is a learning program to find that algorithm, not the algorithm itself.
Go programs made steady progress for a decade with improved tree pruning methods, which reduce the total number of positions the program has to evaluate. The cleverest method is Monte Carlo pruning, which simply prunes at random. 

Posted March 13, 2016 6:00 AM
Permalink | No Comments |

Paraccel was founded in 2005 by former Netezza executives as an alternative to Netezza's DW appliance idea. Paraccel runs on a commodity platform. It got lots of VC money (I've heard estimates as high as $90m) but has comparatively few customers. 

Recently Paraccel apparently hoped to generate a little revenue with hosting deals, which are always low priced. Or maybe they were just in it for the publicity. The technology is used for MicroStrategy Wisdom and I doubt MicroStrategy paid them very much. Shortly thereafter, they entered a deal with Amazon to license their technology for Redshift, which Amazon is reselling at very low rates.

Paraccel's struggles aren’t very surprising, since ALL the other vendors in the space except Teradata lost their independence in 2010/2011. Sybase, Netezza, Kickfire, Greenplum, Vertica and Aster Data were all acquired. Also HP killed Neoview in the same time frame. 

Actian is the newish name for Ingres, and controls the open source database by the same name. It also has several other databases including VectorWise, and there seems to me to be a good deal of overlap between the products. It is not a high profile company but it will be interesting to see what strategy they adopt to squeeze cash out of this highly funded and presumably expensive product.

Posted April 27, 2013 1:42 PM
Permalink | No Comments |
Have a look at this TIME article. The headline is "Americans Are Eating Fewer Calories, So Why Are We Still Obese?".

It goes on to cite two studies. One study shows that old people eat less fast food then young adults, and that in general food consumption is declining. This comes from the aging population I guess. The other shows a small decline (about 7%) in the calorie consumption of children. 

Neither study showed a decrease in calorie consumption by adult Americans, as far as I can tell. The naive assumption would be that consuming fewer calories would result in less obesity, but no data on the topic is presented. The headline is a complete misdirection.

The moral of the story is you should limit your observations to what the data actually says.

Posted March 27, 2013 1:56 PM
Permalink | No Comments |
Analysts like to provide a list of predictions for the coming year. I've never been a big fan because I think that real tipping points are pretty rare in any industry. People tend to overestimate the short-term change and underestimate long-term change.

Here are some trend ideas that are floating around the internet. There are basically three kinds:

New data sources: This mostly means Big Data, whatever that means. (Personally, I like Amazon's definition, which is "too much data to handle on a single server". Other definitions strive to include appliances like Teradata and HANA.)

New applications: Sentiment analysis, predictive analytics, collaboration

New UIs and functions: Better dashboards and visualizations, more self service and agility, voice interfaces, mobile.

New Platforms: This is mostly in memory, cloud and SaaS.

My take on all this is simple:

New data sources: I think big data is still mostly for online and mobile providers. It's true that manufacturers and retailers are trying to figure out how to make better use of the huge amounts of data their business directly and indirectly generates. But this business is still heavily dependent on boutique providers that bring a lot of domain knowledge and deep understanding of statistics with into the deal. I do not think it will have much impact on existing BI business. It's something different.

New applications: The same remarks apply to sentiment analysis as to big data. Predictive analytics is a more interesting market, but to find large, non-specialist audiences, vendors need to prove that their "black box" predictions are as reliable as the expertise of business users without explaining the math behind them.

New user interfaces and functions continue to appear, but I believe that as long as most BI companies stick to solving the easy problems, like making software look cool in a demo, and ignore the harder problems, like user-friendly data governance, there will be no big changes here. Mobile has surprised me, but it still hasn't made a big difference in the BI business.

New platforms. It is good to remember that business users don't care what platform is used, and that the most successful projects are controlled by business users. A platform is only good in the sense it delivers things like speed and convenience. It doesn't add any value per se.

Posted March 14, 2013 9:13 AM
Permalink | No Comments |
Have a look at this TIME article. The headline is "Americans Are Eating Fewer Calories, So Why Are We Still Obese?".

It goes on to cite two studies. One study shows that old people eat less fast food then young adults, and that in general food food consumption is declining. This comes from the aging population I guess. The other shows a small decline (about 7%) in the calorie consumption of children. 

Neither study showed a decrease in calorie consumption by adult Americans, as far as I can tell. The naive assumption would be that consuming fewer calories would result in less obesity, but no data on the topic is presented. The headline is a complete misdirection.

The moral of the story is you should limit your observations to what the data actually says.

Posted February 22, 2013 9:20 AM
Permalink | No Comments |
Christmas shopping this year really drove home to me how completely the electronics industry has made itself obsolete in recent years. The electronics stores are emptying out. All the gadgets that have been so popular in recent decades -- cameras, camcorders, VCRs, tape recorders, CD players, portable music boxes, dictation machines, game consoles, pagers, PCs, notebook computers, TVs, radios, pocket calculators, GPS navigation devices, synthesizers, mixing consoles and of course telephones (mobile and land based) have all disappeared into smart phones.

But smart phones aren't really phones at all, they are just palmtop computers that include an interface for cellular networks and a phone app. They are called smart phones for marketing reasons -- because the phone companies use them to lock consumers into overpriced network contracts. If photography were the killer app, they would be called smart cameras, which is just as appropriate. The only thing that seems to be keeping the entire electronics industry from being swallowed up by the black hole of Moore's Law is this kind of marketing wheeze and the (rapidly falling) price of screens.

I'm old enough to remember when Radio Shack was a national treasure. Now it's just a place to buy batteries while being harassed by a hard selling salesman from some telecoms oligopolist best known for hating its customers. And in this age of smart hearing aids, robotic assembly lines and fly-by-wire airplanes, its not just consumer electronics that is being computerized.

Posted December 30, 2011 2:17 PM
Permalink | No Comments |
There is a discussion of some remarks I made to Ann All at IT Business Edge here, together with some opnions by Howard Dresner.

The interview reflects my opinion pretty well. I see mobile BI as a way to find new types of customers for BI more than as a way to replace existing installations. Too bad I didn't mention salespeople on the road, I think they are an important potential market as well.

Another point is that I think most people who said they expected mobile BI to be in use within 12 months were being too optimistic. The BI Survey 11 will address this question.

Posted December 14, 2011 1:24 PM
Permalink | No Comments |