Connect with us

Artificial Intelligence: The Journey to a Thinking Machine



The following content is sponsored by VERSES AI

Artificial Intelligence: The Journey to a Thinking Machine

When the latest iteration of generative artificial intelligence dropped in late 2022, it was clear that something significant had changed.

The language model ChatGPT reached 100 million active monthly users in just two months, making it the fastest-growing consumer application in history. Meanwhile, Goldman Sachs predicted that AI could add 7% to global GDP over a 10-year period, almost $7 trillion, but also replace 300 million jobs in the process.

But even as AI continues to disrupt every aspect of life and work, it’s worth taking a step back. 

In this visualization, the first in a three-part series called The AI Revolution for sponsor VERSES AI, we ask how we got here, where we’re going, and how close are we to achieving a truly thinking machine?

Milestones to Mainstream

The term “artificial Intelligence” was coined by computer scientist John McCarthy in 1955 in a conference proposal. Along with Alan Turing, Marvin Minsky, and many others, he is often referred to as one of the fathers of AI. 

Since then, AI has grown in leaps and bounds. AI has mastered chess, beating Russian grandmaster and former World Chess Champion Garry Kasparov in 1997. In 2016, Google’s AlphaGo beat South Korean Go champion Lee Sedol, 4-1. The nine-year gap in achievements is explained by the complexity of Go, which has 10360 possible moves compared to chess’ paltry 10123 combinations.

DALL-E arrived in 2021 and ChatGPT-4 in early 2023, which brings us to today.

But What is Artificial Intelligence?

There’s a big difference between the Roomba that vacuums your condo and HAL from 2001: Space Odyssey. This is why researchers working in the field have come up with the following ways to classify AI:

AI Category DefinitionExample
Artificial Narrow Intelligence (ANI)An AI that is good at a particular task at a level equal or better than a human beingVirtual assistants, such as Siri or Alexa
Artificial General Intelligence (AGI)
An AI that can perform any task that a human being can. This is what most of us think of when we think of AIMarvel’s J.A.R.V.I.S. (Just A Rather Very Intelligent System)
Artificial Super Intelligence (ASI)
An AI that can surpass anything that human beings can do
The Minds from Iain M. Banks' sci-fi Culture series

Despite a false alarm by one Google software engineer in 2022 and a paper by early GPT-4 boosters, no one really believes that recent generative AIs qualify as thinking machines, however you define it. ChatGPT, for all its capabilities, is still just a souped-up version of autocomplete. 

Do Androids Dream of Electric Sheep?

That was the title of Philip K. Dick’s science fiction classic and basis for the movie Blade Runner. In it, Harrison Ford plays a blade runner, a kind of private investigator who used a version of the Turing Test to ferret out life-like androids. But we’re not Harrison Ford and this isn’t science fiction, so how could we tell?

People working in the field have proposed various tests over the years. Cognitive scientist Ben Goertzel thought that if an AI could enroll in college, do the coursework and graduate, then it would pass. Steve Wozniak, co-founder of Apple, suggested that if an AI could enter a strange house, find the kitchen, and then make a cup of coffee, then it would meet the threshold. 

A common thread that runs through many of them, however, is the ability to perform at one thing that humans do without effort: generalize, adapt, and problem solve. And this is something that AI has traditionally struggled at, even as it continues to excel on other tasks. 

Can Current State-of-Art AI Achieve Thinking Machines?

And it may be that the current approach, which has shown incredible results, is running out of road.

Researchers have created thousands of benchmarks to test the performance of AI models on a range of human tasks, from image classification to natural language inference. According to Stanford University’s AI Index, AI scores on standard benchmarks have begun to plateau, with median improvement in 2022 limited to just 4%.

New comprehensive benchmark suites have begun to appear in response, like BIG-Bench and HELM, but will these share the same fate as their predecessors? Quickly surpassed, but still no closer to an AI like J.A.R.V.I.S. that could pass the Wozniak Coffee Test?

Imagine a Smarter World

VERSES AI, a cognitive computing company specializing in next generation AI and the sponsor of this piece, may have an answer.

The company recently released research that shows how to build an AI that can not only think, but also introspect and explain its “thought processes.” Catch the next part of The AI Revolution series to learn more.

Visual Capitalist Logo

Learn more about how VERSES AI is building a smarter world.

Click for Comments

You may also like