5 Use Cases for Neural Networks You Might Not Know

What is currently more important to the success of the AI industry than Artificial Neural Networks? ANNs have already reached into just about every industry, due to their ability to improve just about every business process. With the help of this post, I hope to shed more light on how ANNs are currently being used in practice, as well as where they might be used in the future as technologies continue to develop.

ANNs may be able to improve medical diagnoses. Figuring out what is wrong with a patient and then prescribing an effective treatment for them is a complicated process, to say the least. As of today, numerous scientific papers have been published that claim ANNs can both streamline and speed-up diagnoses. At this point, most sources seem to agree that this is simply due to the recorded accuracy of ANNs related to diagnosing certain diseases. One scientific article claims that neural networks can diagnose five specific diseases, including chickenpox with between 90 and 97% accuracy. Another study found that when this sort of diagnostic ability is combined with that of a doctor, the number should rise to an almost perfect level(99.5% accuracy). With this in mind, before we have AI doctors, we’ll likely have AI assistants that augment just about every aspect of a doctor’s job.

2. GANs are a particular ANN framework that have been used for several striking use cases in different industries. For the most part, the media has spent a lot of time focusing on how this type of ANN can create art by studying the characteristics of hordes of artwork from the past. With this same structure however, a lot more is possible. Because of the precise way that a GAN looks at input data to constantly improve its’ outputs, it is also a useful structure in all of the areas on this list as well as other examples like the blockchain space, the cybersecurity industry, the 3D printing industry, and the pharmaceutical industry. To recap why this is the case, in a GAN, one or more neural networks are split in half. One half, called the Generator, works to create its’ desired outputs but the other, called the Discriminator constantly challenges the outputs that the Generator creates. Since these outputs can be just about anything depending on the industry, let’s take the 3D printing space as an example. One of the latest trends in 3D printing is “metal printing” or the usage of 3D printers to create metal-based products. Since this is a highly involved process to say the least, the usage of neural networks has been put forth by companies like Sculpteo and Ai Build. In both cases, they mention the possibility of AI ushering a new age of industry through being connected to and driving the evolution of 3D printing. For this to happen, however, networks like GANs would have to get 3D printing and as a result, manufacturing, to the point that everything could be done autonomously. Imagine a world in which factories created everything without any sort of human intervention. While we’re a long way off from this point, a good starting point would be for 3D Printing firms to integrate GANs into all of their machines to teach them to gradually work completely autonomously. From here, the next step would be for 3D printing to supplant traditional manufacturing across all verticals, which will be a tall task to say the least. If GANs can become a mainstay in factories, though, who knows what the future will hold.

3. ANNs are likely essential to the continued development and maintenance of any sort of autonomous vehicle. The easiest way to understand this is to think about how self-driving cars take in and analyze data to improve. In other words, if you’ve read any of our other posts, consider the reinforcement learning framework. In the simplest sense, structuring an AI system based on a reinforcement learning framework means making sure it can learn through repetition as a human does. It repeats a desired action until it achieves a number called its’ desired q-value. As the Towards Data Science blog points out, the q-value is not the reward that the system wants to achieve but the output value of the current action it has taken, which gives it an idea of how close it is to achieving its’ numerical reward. This reward is given to the AI when it does an action perfectly, as it was meant to do. If we then take all of this in the context of ANNs and autonomous cars, we can say that every action the vehicle chooses to take is processed as an input, which then pases through the hidden layer of its’ ANNs, and comes out as an output with an assigned q-value. A more specific example of how this might play out would be if a car chooses to turn right in 300 feet, when it should have done so in 350 feet to complete the turn effectively. The result of this action, therefore, would teach the car that it did not achieve its’ desired result, because it would be assigned a low q-value with no reward attached. With this, I hope you’re beginning to see that neural networks are integral to the success of autonomous vehicles over time. The AV space is still very early-stage, but the more you know about these topics, the better you will be prepared for the future that many believe is inevitable. Overall, keep in mind that this discussion of the importance of q-values and numerical rewards is also only the tip of the iceberg. For more on these subjects, check out the links related to these topics in our resource list below.

4. Last but not least, the future may be more connected than any of us can currently imagine. ANNs will one day likely be the main engine in all of our smart devices, including smart homes. As of last year, there were reportedly already 7 billion devices that could be considered to be IoT enabled in existence. One can only imagine where that number will stand by the end of this year, especially with the newest iPhones already having built-in neural networks and more, similar devices from competitors likely on the way. Now imagine when the IoT finally scales in a fashion that can support 7 billion or more devices without experiencing major slowdowns. New possibilities like having major neighborhoods or even cities full of smart homes may likely be truly possible. For projects that are working on bringing this era about with the help of ANNs, check out Fetch.ai and even Augur, both of which I have mentioned before in previous posts.

With these cases in mind, it will hopefully be easier for you to imagine future uses for ANNs or even how you might contribute to their evolution as a technical framework. In one of my next posts, I’ll dig a bit into how AI and ML can truly help a company drive its’ data strategy. Until then, if you’re interested in taking a deeper dive into any of these subjects, start with out list of resources below.