Artificial intelligence (AI) and nanotechnology are among the most hyped emerging technologies today. But in many ways, they are also the least understood. While unusual use cases of either technology—such as the ability to fake your own voice with AI—grab the headlines, the truth is that both AI and nanotech are already with us and being used in mundane, everyday circumstances.
In this article, we want to get beyond the hype. Here, we won’t dwell on speculative, far-future use cases. Instead, we’ll look at real-world, actually-existing situations in which AI and nanotech are already being used. And by doing that, we’ll immediately see that there is a natural overlap between the technologies, and one that can drive the development of both.
What AI and nanotech are (and what they aren’t)
First, a word about what AI and nanotech are and what they are not. The novelty of both technologies today is such that many people still think of them as science fiction. This impression isn’t helped by high-profile scientists claiming that AIs could eventually destroy humanity or nanotechnology could take our bodies away from us.
These are exciting, spectacular scenarios, but the reality of both technologies is much less spectacular and closely tied to the needs of the contemporary economy. In fact, the most widespread use of AI has been in the form of chatbots; rather than taking over the world, at the moment AIs are mainly focused on customer service.
Similarly, though the term “nanotechnology” still sounds like one from science fiction, for researchers working in the field, it has a precise and somewhat less sensational definition—ANT technology—that makes use of the nanometer scale. And today, this is actually pretty common. More than 300 products are already nano-based, according to a database maintained by the Woodrow Wilson International Center, in Washington, D.C.
With these definitions in mind, let’s take a look at three ways in which both technologies are converging.
Atomic force microscopy (AFM) might seem like an arcane place to start with this list, but it’s one of the clearest examples of how nanotechnology and AI can work together. Put simply, AFM is a technique for imaging objects at the nanoscale. This is useful for quality assurance when making microchips and in looking at cells within the human body.
The problem is that, at this scale, the materials that make up the microscope itself have a significant effect on the data it returns. If you are using tiny atomic forces to investigate materials, in other words, you need to be prepared for an extremely noisy signal. This is an inherent property of nano-scale microscopy, and also happens with electron microscopes. And while there are ways of filtering out signal noise, they are computationally expensive.
This is where AI comes in. An AI approach known as functional recognition imaging (FR-SPM) addresses this issue through the direct identification of local actions from measured spectroscopic reactions. This process uses artificial neural networks (ANNs) with principal component analysis (PCA) to streamline the input data to the neural network.
These models are trained with data sets produced from exhaustive analysis of microscopy signals. It is possible, in other words, to filter out signal noise manually, but doing this requires that the same sample be analyzed many times. AI models can greatly reduce this requirement, because they are able to spot the principal component of a data signal much more quickly than a human can.
This approach allows researchers to identify their target signal from the surrounding noise and, therefore, work with materials at the nanoscale in a more efficient way. And best of all, some of these models are available as open-source projects, which is likely to speed their adoption across the scientific community.
A similar revolution has been quietly occurring in the world of chemical modeling. Chemical modeling simulates how molecules will interact with each other. It’s used widely in bioscience and drug development. More recently, however, scientists have begun using the same modeling techniques to better understand the behavior of materials at the nanoscale and thereby have been able to improve their efficiency and efficacy.
Neural networks have been used for chemical modeling for years, but it’s only recently that they’ve been applied to nanotechnologies specifically towards understanding how nanotech materials behave under real-world conditions. AI is being used, for instance, to understand carbon nanotube structures by quantifying structural qualities like alignment and curvature.
There are many factors that must be considered in order to generate an image or a dynamic depiction of a chemical system. Up until recently, and as with the example of microscopy above, isolating these factors from the surrounding noise has been very difficult. AIs, however, are effective at this task.
Using AIs, scientists and engineers can now minimise the degree of error related to the geometry or size of a system or particle. The most popular approach to doing this is to train an AI model on data emerging from systems whose behaviour is already well understood.
Techniques like this are especially useful for nanomaterials as the several effects and phenomena seen with materials like graphene can often be difficult to recreate. This application is one that has a huge potential. It promises, in fact, the ability to integrate machine learning into production techniques—and therefore catalyze the future development of both AI and nanotechnologies.
Finally, nanocomputing. This is arguably the area in which there is the closest correspondence between the two technologies, and potentially the most productive overlap.
The central promise of nanocomputing is that it can greatly increase the computing power available to researchers and engineers alike. In recent years, some have worried that Moore’s Law—that transistors per chip and therefore computing power doubles over a predictable timeframe—is no longer holding true, because as we build ever smaller computers, we are encountering strange quantum effects that limit our ability to work at this scale.
Nanocomputing is one approach to overcoming this problem. Nanocomputers use a variety of novel media to perform calculations—anything from organic chemical reactions to nano-MOSFETs. However, most of these devices depend on intricate physical systems to allow for intricate computational algorithms and machine learning procedures that can be used to generate novel information representations for a broad range of uses.
In simple terms, AIs can help us to understand the ways in which materials work at the nanoscale. This might allow us to build computers at this scale that are not dependent on the transistor-based architecture that most computers today are based on. This, in turn, will allow the creation of ever more sophisticated AIs, which will allow us to probe this behavior still further. In the same way that neural networks could help computers code themselves, nanocomputing technologies could allow computers to build themselves.
Of course, both AI and nanotechnology are emerging technologies, and it remains to be seen how each will develop. However, the ways in which these technologies are already being used mean that it’s possible to see an emerging syncretism. Advances in AI are allowing us to understand the behavior of materials at the nano scale, and this is turn might allow us to create ever more powerful AIs. In this sense, the two technologies are closely intertwined.