Many tasks can be automated and made simpler with AI. As the technology continues to grow and become more precise, the use cases have expanded from something as simple as turning off the lights in a house to the very complex, such as autonomous driving at scale.
The most widely used AI functions today include conversational (chatbots), recommendation engines (recommended items from past purchases), fraud detection (automated credit card alerts), image recognition (face scanning) and robotic-process automation (factory robots).
A basic example is online search. Let’s suppose you want to cook a meal for six people with filet mignon as the main course. You could page through search results and recipes to come up with a menu.
Instead, you could ask a chatbot to prepare a menu for six people with filets as the main course. You also could exclude peanut ingredients if a guest has allergies. The chatbot would perform the heavy lifting of searching the internet, bringing back results in seconds. You could even follow up by asking for a grocery list.
This is a basic example of the efficiency Al can bring to various markets, saving time and reducing costs along the way.
Al has the potential to disrupt many industries and unlock significant efficiency improvements across the economy. Given Al’s wide reach and potential impact, it could potentially become not only one of the largest markets in technology, but also one of the fastest growing.
Two primary opportunities for AI investors include:
- Infrastructure — This refers to the nuts and bolts of Al — the chips needed to power complex algorithms, such as the most advanced central and graphics processing units. We think demand should grow as companies that serve as data centers continue to expand and upgrade their hardware.
- Large cloud service providers — These give customers access to data centers with the most advanced hardware so they can run Al algorithms. Renting computing power tends to be cheaper and more efficient than building it from scratch, especially with the need for frequent upgrades and maintenance. Providers of cloud services should benefit as Al adoption accelerates.
Cost
AI comes with a hefty price tag. The most complex models require a significant amount of computing power: thousands of central processing units and processing accelerators, such as graphics processing units, and the accompanying software to run the algorithms. Training a model is much more expensive and time-consuming than using it.
Even if a company has access to the appropriate hardware and software, there is a tremendous need for data scientists and engineers who can develop complex algorithms. Given the high demand for this type of skilled talent, the costs to maintain even a small, dedicated Al team can come to millions of dollars a year or more at larger companies.
Data management and ethical issues
While Al is an exciting technology, it brings ethical questions and dilemmas around data security and privacy, potential discrimination and biases within models, and the misuse of the technology.
The ability to understand and explain how a model arrives at an answer is another important consideration. Al algorithms usually provide little insight into the understanding or process of how they arrive at a conclusion. Making the process more transparent is critical to building trust.