Understanding the Brain with the Help of Artificial Intelligence

Neurobiologists aim to decode the brain’s circuitry with the help of artificial neural networks. NeuroscienceNews.com image is credited to Julia Kuhl.

Researchers have trained neural networks to accelerate the reconstruction of neural circuits.

How does consciousness arise? Researchers suspect that the answer to this question lies in the connections between neurons. Unfortunately, however, little is known about the wiring of the brain. This is due also to a problem of time: tracking down connections in collected data would require man-hours amounting to many lifetimes, as no computer has been able to identify the neural cell contacts reliably enough up to now. Scientists from the Max Planck Institute of Neurobiology in Martinsried plan to change this with the help of artificial intelligence. They have trained several artificial neural networks and thereby enabled the vastly accelerated reconstruction of neural circuits.

Neurons need company. Individually, these cells can achieve little, however when they join forces neurons form a powerful network which controls our behaviour, among other things. As part of this process, the cells exchange information via their contact points, the synapses. Information about which neurons are connected to each other when and where is crucial to our understanding of basic brain functions and superordinate processes like learning, memory, consciousness and disorders of the nervous system. Researchers suspect that the key to all of this lies in the wiring of the approximately 100 billion cells in the human brain.



To be able to use this key, the connectome, that is every single neuron in the brain with its thousands of contacts and partner cells, must be mapped. Only a few years ago, the prospect of achieving this seemed unattainable. However, the scientists in the Electrons – Photons – Neurons Department of the Max Planck Institute of Neurobiology refuse to be deterred by the notion that something seems “unattainable”. Hence, over the past few years, they have developed and improved staining and microscopy methods which can be used to transform brain tissue samples into high-resolution, three-dimensional electron microscope images. Their latest microscope, which is being used by the Department as a prototype, scans the surface of a sample with 91 electron beams in parallel before exposing the next sample level. Compared to the previous model, this increases the data acquisition rate by a factor of over 50. As a result an entire mouse brain could be mapped in just a few years rather than decades.

Although it is now possible to decompose a piece of brain tissue into billions of pixels, the analysis of these electron microscope images takes many years. This is due to the fact that the standard computer algorithms are often too inaccurate to reliably trace the neurons’ wafer-thin projections over long distances and to identify the synapses. For this reason, people still have to spend hours in front of computer screens identifying the synapses in the piles of images generated by the electron microscope.



Training for neural networks
However the Max Planck scientists led by Jörgen Kornfeld have now overcome this obstacle with the help of artificial neural networks. These algorithms can learn from examples and experience and make generalizations based on this knowledge. They are already applied very successfully in image process and pattern recognition today. “So it was not a big stretch to conceive of using an artificial network for the analysis of a real neural network,” says study leader Jörgen Kornfeld. Nonetheless, it was not quite as simple as it sounds. For months the scientists worked on training and testing so-called Convolutional Neural Networks to recognize cell extensions, cell components and synapses and to distinguish them from each other.

Following a brief training phase, the resulting SyConn network can now identify these structures autonomously and extremely reliably. Its use on data from the songbird brain showed that SyConn is so reliable that there is no need for humans to check for errors. “This is absolutely fantastic as we did not expect to achieve such a low error rate,” says Kornfeld with obvious delight at the success of SyConn, which forms part of his doctoral study. And he has every reason to be delighted as the newly developed neural networks will relieve neurobiologists of many thousands of hours of monotonous work in the future. As a result, they will also reduce the time needed to decode the connectome and, perhaps also, the consciousness, by many years.
Source: Max Planck Institute / Neuroscience.news

YOUR INPUT IS MUCH APPRECIATED! LEAVE YOUR COMMENT BELOW.

How Companies are benefiting from lite artificial intelligence

Artificial intelligence is hot, but also daunting. The latest advances — known variously as cognitive computing, machine learning, and deep learning — sound complicated and expensive. And they are, despite the enormous potential they bring to the marketplace. For many companies, the price tag and the commitment of resources are too high a hurdle. And for other reasons, even a giant like Apple is facing challenges.

But the good news is that the early dividends from AI are already within reach of most midsize companies as they look for ways to expand their digital boundaries. In fact, the building blocks of AI can produce great results with fewer technical requirements and less time and money than many companies realize. What’s more, those that take this initial step are getting a leg up on AI’s future, since that step is going to be a prerequisite for everything that follows.

First, let’s get our bearings.
At the high end of artificial intelligence are systems like cognitive computing that are allowing driverless cars and other machines to develop the capacity to learn from their experiences in the world — by incorporating nuances, remembering outcomes, and adapting to mistakes. (A recent accident involving a Tesla “autonomous car” has raised questions about AI’s current limits.)

At the less-expensive end is a knowledge-based approach that organizes data and language into highly malleable and helpful blocks of information. These “AI Lite” systems don’t learn new tricks — unless their human minders use new code instructions to “teach” them. But they can become very smart indeed about sorting and distributing their information in extremely fast ways.



What follows are some guideposts to help you put AI Lite to work.

Find the right places to use it. Even in this new information age, not everything requires the razzle-dazzle of AI. But companies and government agencies are starting to find plenty of places where knowledge-based tools can make a huge difference. These include improving data-mining operations, helping with training, and making structured, repeatable tasks and processes far more efficient and less costly. And they are finding the tools increasingly useful, of course, in dealing with online customers.

For example, Allstate Business Insurance, a division of Allstate Insurance, used the tools to develop a virtual assistant known as ABIe (pronounced “Abby”) to answer questions from its 12,000 agents. It was a bit like hiring Apple’s Siri at a sliver of the cost. Mike Barton, the division’s president, put it this way: “We think of ABIe as our precursor to cognitive computing on a shoestring.”

When the company upgraded its commercial insurance line for small businesses a few years ago, the agents jammed internal call centers with questions about how the policies worked and how to set sales quotes. The cost of simply expanding the call centers was prohibitive.

Enter ABIe (shorthand for the Allstate Business Insurance Expert), which my firm helped develop. Employing a rigorous approach to the words and phrases at the heart of the company’s products, ABIe’s avatar-driven interface offers accurate answers to policy questions while streamlining the quote process. From just a few thousand queries a month in 2013, ABIe now handles 100,000 — from all of the company’s employees, and not just from agents. A new version of ABIe will soon be taking queries directly from the customers. Best of all, ABIe paid for itself the first year, so almost all of the ongoing savings now drop to the bottom line.



Roll up your sleeves. AI Lite is far less complicated and less expensive than the high end of the spectrum, but that doesn’t mean it is off the shelf. One size does not fit all — and there are no plug-and-play magic bullets.

For example, it took nearly a year to design, build, and implement ABIe. Allstate Business assigned a team of managers and subject-matter experts to the project to figure out the “taxonomy” of the words, phrases, and data that ABIe would require to answer all of those questions. ABIe pulls its answers from a data warehouse, where all the knowledge relating to the company’s products and processes has been organized. In AI speak, that meant all of the company’s numbers, charts, words, and phrases had to be chopped, chunked, tagged, and HR-optimized to give ABIe the ingredients for those answers.

In short, companies that want to get into this game will have to roll up their sleeves and do some old-fashioned blocking and tackling. But the prize is worth pursuing: If the right data is married to the right vocabulary and terminology, a company’s information capabilities will soar.

Don’t expect everything to be perfect. Given all the moving parts, however, mistakes, and the need for midcourse corrections, are inevitable, so prepare for them. In ABIe’s case, the team started out by offering some answers that were far too thorough, essentially telling the questioner how to build a clock when all that was sought was the correct time. It took trial and error — including detailed debriefs of those using ABIe — to arrive at very specific and actionable answers and to put in place the governance, metrics, and change-management processes to make controlled, methodical updates.



Don’t make your AI too lite. When using AI to interact with online customers, keep in mind that piecemeal approaches won’t work. Most organizations are deploying department-level solutions and standalone tools without sufficient funding. The results are typically inconsistent and haphazard, forcing time-consuming and costly fixes.

One department at a company, for example, may see the customer through a transactional lens, while another puts the emphasis on promotions. The differences in their data models will slow, and perhaps hobble, the entire program.

Getting your money’s worth from AI — whether at the knowledge-based end of the spectrum or later on, with more extensive, and expensive, applications — requires three things. The effort must involve careful analysis and preparation, which takes into account each department but keeps the focus on the full enterprise. It must have a formal (and nuanced) governance structure. And someone at the most senior level of the company must sponsor it.

Without those foundational elements in place, you will fall short in deploying the incremental power of AI Lite and be ill prepared for the revolutionary changes that AI promises to bring.



Source: Seth Earley

YOUR INPUT IS MUCH APPRECIATED! LEAVE YOUR COMMENT BELOW.