I am in the process of building a software company. I'm writing the back end code with a GenZ friend of a friend who is more in touch with his generation's preferences than I, the older and wiser millenial. It's a tedious process of GitHub iterations, Zoom calls, turning visions into reality, and further revisions.
What if humans didn't have to program software and could rely on the abilities of deep learning to automate the creation of software? What if we didn't have to learn coding languages anymore and could let machine-driven algorithms write our software for us? Turns out we are headed in that direction. ARK Invest, a thematic fund manager that manages active ETFs, believes deep learning will add $30 trillion to the global equity market capitalization during the next 15-20 years.
Deep learning techniques, when combined with human ingenuity, can power the way forward towards greater technological innovation. Software behind self-driving cars, drug discovery, consumer-focused apps, and conversational computers (think Siri) already use varying degrees of artificial intelligence. Waymo's autonomous vehicles have collected more than 20 million real world driving miles across 25 cities, including San Francisco, Detroit, and Phoenix. TikTok uses deep learning for video recommendations and its daily active user base has outgrown that of Snapchat and Pinterest combined. Smart speakers answered 100 billion voice commands in 2020, 75% more than in 2019.
While hardware and software advances continue to drive down costs of AI training, the size of the AI models is growing much faster. At 10x growth per year, AI training costs are likely to continue climbing as tasks grow in complexity. As a result, spending on specialized processors that can handle the increased model size are likely to account for the majority of the growth.
Most impressively, artificial intelligence is expanding from vision to language. 2020 was the first time that AI systems could understand and generate language with human-like accuracy. Conversational AI requires 10x the computing resources of computer vision and is likely to spur massive investments in the coming years.
Nonetheless, the deployment phase for deep learning as it relates to software programming will proliferate throughout the economy and benefit a large number of industries. Greater efficiency will likely be the result, as will disinflationary pressure as less human services are required. Who knows, maybe I won't have to write my software code, after all.
No comments:
Post a Comment