Artificial Intelligence (AI) is here and here to stay. Beginning around 2012, the deep learning revolution brought advanced artificial intelligence into the mainstream. Leaps and bounds in improvements in hardware (GPUs), processing power and cloud services (AWS), and deep learning software frameworks (Tensorflow, Theano, Torch) allowed researchers to make strides at an unparalleled pace, inventing new AI models that could match human performance on many tasks.
These days AI is ubiquitous and is used everywhere from financial services (automated credit scoring), automotives (self-driving vehicles), pharmaceuticals (computational drug discovery), biotech (precision medicine and genetics), robotics process automation, and so many other areas. With this, the number of publicly traded companies using AI in some way in their products has also increased exponentially. Today, you’ll be hard pressed to find a company which isn’t using AI in at least some small way within their pipeline. The future of society as it relates to AI is compelling, but also much different from what we currently face. It’s difficult to anticipate all the ways in which AI might someday be used or influence the ever shifting landscape of our communities and culture, but one thing’s for certain: it’s not going anywhere and we are the ones who’ll need to adjust – not the machines!
1. Autonomous Vehicles
Unless you’ve been buried under a rock the past couple of years, you’ve probably heard about the strides that are being made with self-driving vehicles. Self-driving technology uses machine learning and AI at its very core, from mapping the visual environment a car sees on the road, to detecting pedestrians and obstacles, to route planning. Autonomous driving capabilities will likely revolutionize not only personal transportation, but public transit, taxi services, trucking, and anything that uses navigation.
Companies are already investing in autonomous trucks for delivery of consumer packages as well as for shipping materials cross country. This could potentially save them billions of dollars that would otherwise have gone to paying drivers. In the personal transportation market, Google and Lyft have already launched autonomous taxi services in select markets with the goal of expanding after further market validation.
Crazily enough, although the autonomous driving tech is far from being perfected at the moment, there exist publicly traded companies which are looking to develop autonomous flying vehicles. This is actually not a huge leap of the imagination, as AI is already being used for self- piloting drones for package delivery and military reconnaissance.
2. AI and Language
Many of the recent gains in deep learning have been in the field of Natural Language Processing (NLP). NLP concerns itself with machine learning models that deal with human language. The places where NLP technology is used range from chatbots, to machine translation systems, personal assistant devices, and more.
One of the key players in NLP these days is Open AI. Originally founded by Elon Musk, Sam Altman, and other tech visionaries as a company to promote the safe development of artificial general intelligence (AGI). Open AI has created a series of GPT models which can produce realistic natural language text. It can also write full articles, and the outputs are so uncannily accurate that many are worried that it will be used to create fake news articles for purposes such as manipulating elections.
AI is being used with language for many other purposes. It is facilitating much more rapid and accurate translation services than were ever possible in the past. This is a godsend for foreign service workers, military overseas, or even just the everyday traveler who can now rely on AI powered translation to help them get around and interact with locals.
AI is also being used for analyzing textual data found on social media for purposes such as marketing and advertising. By inferring sentiment around consumer discussions of products, AI can help businesses better target their marketing efforts and gain insight into their customer demographics that were not possible in the past.
3. Internet of Things
The internet of things is a rapidly growing set of common and household devices connected to one another via the internet. The reason that AI is so important here is that the internet connectivity allows data to be gathered and shared across multiple devices. More data always means more opportunity for AI to be effectively utilized. This superfluity of data will allow new AI models to be trained on more diverse data sources, creating new predictive value out of our devices.
For example, while you can use a voice assistant device like Amazon’s Alexa and say “Hey Alexa, turn on my office lights so that I can read,” wouldn’t it be better if Alexa could anticipate your daily reading schedule, combined with your location data, and turn on your lights without you even having to ask? That’s a contrived example, but one that’s not too far off from what the IOT allows.
Although the pharma industry is often thought of as stodgy and sluggish due to both heavy regulations from the FDA as well as slow-to-move executives holding on to antiquated business philosophies, in reality, this is not quite accurate. Many pharmaceutical companies are fostering internal groups that operate much like cutting-edge Silicon Valley startups. They are pouring millions of dollars of investment into some of the latest technologies including machine learning and artificial intelligence.
ML and AI are primarily used within the pharma industry for speeding up or streamlining existing development processes. One of the primary issues plaguing pharma is the fact that drugs are taking longer, and becoming far more expensive, to develop than was once the case. This is due to a combination of factors, some of which are related more to regulation than to the actual technical difficulty of developing a new drug (although this is not to be underestimated).
However, the way in which AI can help is via machine learning methods for computational (sometimes called in-silico) drug discovery. The fact is that the space of synthesizable molecules are so large (10^10 or more by many estimates) that it is infeasible to create and test even a small fraction of this possible search space. Thus, many companies are using machine learning methods on molecular structures to optimize for the properties which make a molecule most likely to bind to a target protein or receptor (affinity) and also be tolerable to the human body (solubility, efficacy, minimization of side effects, etc.). These days you can pretty much guarantee that almost every major pharma company has at least dipped its toes into AI methodologies.
Another smaller, but no less important, way in which pharma companies are using AI is to better select participants for clinical trials. Using machine learning, companies could select a more statistically robust sampling of individuals by training on features such as their individual health history, conditions, and even their genome.
5. Materials Design
Closely related to drug discovery is the field of materials design. The idea here is simple – for a given application, we want to be able to design the best materials, those that are the cheapest, most easily synthesizable, strongest, and most resistant to elements such as stress, force, rust, and heat. This is a big deal in, for example, spacecraft design where the materials of the rocket’s outer shell can determine whether it returns safely to earth or burns up into a fiery ball upon reentering the atmosphere.
Materials engineering is similar to computational drug discovery in that both involve searching through a large molecular space for hits that optimize certain chemical properties. In drug discovery, these properties have to do with things such as bioavailability and how the body processes a drug. In materials design, they include things such as predicting crystal structures, modeling interatomic potentials, or assessing something like the melting point of a given Material.
The utility of materials engineering is not limited solely to the aerospace sector. Better materials are needed almost everywhere from biotech, where you might want to engineer nanoparticles for vaccine delivery or artificial tissues for organ replacement, to 3D printing where you would want to engineer new and better plastics.
6. Computer Vision
Computer vision is currently one of the best-developed subfields of AI as research on algorithms that help computers to “see” has been around for years. In many ways, this research sparked the deep learning revolution of the past decade. Computer vision methodologies are likely to be used by any company which has data of a visual nature, be that images, videos, or VR models.
Smartphone products need computer vision algorithms to correct and post-process camera photos. Similarly, photo sharing platforms use AI and computer vision to tag, classify, categorize, and identify individuals in uploaded photos. AI and computer vision can also be used in a more industrial setting. For example, computer vision algorithms are being developed that allow for automating factory and manufacturing processes. These algorithms can take as input images from assembly lines and perform tasks, such as identifying manufacturing defects in products.
7. Financial and Insurance
Some other classically stodgy industries that are turning over a new leaf by rapidly adopting AI technology include the financial and insurance sectors. Of course, quantitative trading firms such as Citadel, Goldman Sachs, and Renaissance Technologies have always been on the cutting edge, using machine learning and deep learning techniques to predict market performance and engage in high-frequency trading in order to generate outsized profits for investors.
Yet these are mostly private institutions, closed off to all but the wealthiest insiders. Yet, machine learning is at the point where it is now making its way into regular banking. Large organizations are using machine learning for everything from fighting transaction fraud to approving auto, home, and business loans. The use of AI in determining eligibility or worthiness for a loan or other financial service is a key theme that is repeating across industries.
Organizations are now routinely incorporating AI into their credit risk scoring models, and startups are using AI to identify credit-worthy individuals and grant them access to affordable financing plans for items they would not have otherwise been able to purchase. Of course, the insurance industry has long been known for its complex, statistical underwriting models, but they’re not limiting the use of AI to just that. AI can be used for everything from estimating vehicle repair costs by adjusting drivers’ premiums based on their driving habits.
8. Robotics Process Automation
You’ve probably heard people voicing the concern that AI will automate a way a lot of the jobs we currently take for granted, and it is a valid one. Already, many companies exist that take highly repetitive and manual tasks that would previously have been performed by a human and automate them using robotics and machine learning technology. Repetitive jobs such as file processing, emailing, appointment scheduling, and many others are increasingly being outsourced to algorithms. AI can also be used to automate more complex tasks such as document analysis, process improvement via mining of system logs, and inventory management. It is even being used by legal firms for heavy manpower tasks such as document Discovery.
9. Genomics and Precision Medicine
This section relates back to the one on pharmaceuticals, but is actually somewhat separate, although pharmaceutical companies are deploying efforts within this domain as well. It is only in the last 10 years or so that getting someone’s genome sequence has become an affordable and timely process. Now, genetic data is proving to be a treasure trove for researchers who want to investigate more closely the human genome, our risk for disease, and how our genetic code predicts who we are and who we might become.
AI is primarily being used in this space to predict disease onset and severity. The technology can be applied to virtually any disease for which enough data is available, but it has attracted a lot of interest lately for its use in Alzheimer’s diagnosis. This is important because the etiology of Alzheimer’s is not yet known, and treatment is not that ameliorative. The best way to address Alzheimer’s currently is to identify it early so that pharmaceuticals can be administered and disease progression slowed.
AI is also being used to predict which drugs you might respond to best. For example, the metabolism of many psychiatric drugs is driven by an enzyme called CYP2D6. Now, certain People secrete low amounts of this enzyme and thus are slow metabolizers of these drugs. They require a smaller dosage or a different drug entirely. Others, however, produce this enzyme in higher quantities and thus can metabolize these drugs more effectively, and at larger doses. For them, a drug that’s processed by CYP2D6 might be a good fit. How our bodies produce this enzyme is coded into our DNA. Thus, genetic testing and AI can be used to target the drugs that suit our own unique biochemistry the best.
AI is being adopted across a wide variety of industries, and its influence is only spreading. There will be a point in the near future where AI will touch every field in some way and automate a large swath of the jobs that our society is employed in today. The progression of AI is likely to be exponential as new researchers flood the field, and demand for AI services is not likely to slow anytime soon. Now is the best time to get in on the ground floor and begin investing in companies utilising AI across key areas of their businesses.