Background:
Artificial intelligence (AI) is nothing new to today’s society. Having been around for many years, it has mainly operated behind the scenes with autonomous cars and your Instagram algorithm. Not until recent developments has AI become more consumer-facing in areas like copywriting, graphic design, video creation, and much more.
Contents:
In the AI, space many terms and ideas have been misconstrued or bunched together, so in order to clear the air for the rest of this deep dive, I wanted to clear up some stuff.
First, let’s broadly define AI. I find Packy McCormick is one of the best when it comes to dulling down advanced concepts. As he puts it, “AI is a broad bucket of “algorithms that mimic the intelligence of humans,” some of which exist today in machine learning and deep learning, and some of which still live only in the realm of sci-fi”.
Machine learning (ML) is a subset of AI that exists broadly throughout technology today. Put simply, ML works by feeding an AI a large number of data points about a certain topic. The AI then takes this data and puts it to work. Let’s take chess for example, by feeding an AI millions of different situations in chess games, the AI can apply these data points in a game and become unbeatable.
The funny thing is that if you take the same AI and play it in checkers, you will win 9/10. Why? Because most current AI applications run on Artificial Narrow Intelligence (ANI).
TLDR; Recent advancements in the machine learning space like deep learning, neural networks, and Transformers, have allowed for powerful AI’s like GPT-3 to be leveraged in ways that weren’t possible before. What we have been seeing recently in the market is a result of founders leveraging this new technology.
As mentioned above, most AI today runs on ANI. Tomorrow’s AI is being built on Artificial General Intelligence (AGI). If the same AI that you beat in checkers was built on AGI, you wouldn’t stand a chance. That is because AGI has the ability to search and learn anything that a human can. It doesn’t have a narrow view, it has a general view.
Artificial intelligence has been ignited in recent years by a few key factors. ML has been around since the 1990s but only recently has a subset of ML emerged that contains Deep Learning and Neural Networks. For the first time, these subsets of machine learning created scalable algorithms. Essentially, the more data you feed an ML algorithm, the better it became. This was a big spark.
Then in 2017, a few researchers created a new neural network called Transformers, which allowed AI and ML platforms to process more data than ever before. Now, advanced AI programs like OpenAI’s GPT-3, or DeepMind’s AlphaFold 2 run on transformers. This technology allows GPT-3 to have the capacity for 175 billion machine learning parameters.
All these developments occurred within the last decade and with the help of data infrastructure companies like Scale, founders and builders are now starting to leverage these developments to build workplace tools.