Cognition is the mental process involved in gaining knowledge and comprehension. Through thinking, remembering, and problem-solving it influences how we perceive and interact with the world around us.
Over the past 20 years, the hyperlinked structure of the Internet has fundamentally changed the way we work, collaborate, and learn. This transition is fastening now that most of our activities, both personal and professional, are moving online.
Our digital lives, along with the behavioral constraints such as quarantine and loss of contact with family and friends are putting a strain on our cognitive system. This is why I see the pandemic as an opportunity to rethink our relationship with technology in a more ethical, balanced, and creative way.
Cognitive tools are external "partners" that stimulate people to make maximum use of their cognitive potential. They provide support to analyze the world, access and interpret information, organize and represent personal knowledge. Today we have computers and smartphones which are extremely versatile and sophisticated. However, we have been building these types of tools for centuries. Pen and paper, the abacus, the typewriter are some examples that fall within the same definition as they provide external support to our mental processes as well.
The history of thought is linked to that of the tools and technological advances that enabled it. Think about space representation in the visual arts for example. In the 15th century, the Italian architect Filippo Brunelleschi developed a linear perspective technique determining an epochal passage from the Middle Ages to the Renaissance. In a famous experiment, he used a mirror and a small tablet with a sight hole to observe the Florence Baptistery in perspective and demonstrate his theories. Fast forward to the 19th century, thanks to the introduction of colors in tubes and the use of mobile easels, the painters began to leave the studio to start painting en plein air, capturing the impression of the moment. In the same period, the invention of the camera caused further changes in the pictorial techniques and subjects.
Each medium has its own peculiarities and influences our mental processes in a different way. For some tasks pen and paper are still the best choice to avoid useless cognitive overload. If you've ever written a letter on a typewriter, chances are you've been more thoughtful with your writing as it was less convenient to make mistakes or change your mind. Computers and the Internet were no exception. They have radically changed the shape, the space, and the speed of information determining major consequences on our cognitive processes once again.
“Transform the medium by which we develop, preserve, and communicate knowledge, and we transform knowledge.”
— "Too big to know", David Weinberger
Cognitive tools deliver processes over features and this makes them timeless. They are not meant to do the job on our behalf like many AI applications do. Rather they are designed to provide support, strategies, and guidance to help people process intellectual stimuli more efficiently. They do it by defining primitive operations that fundamentally change the user's mental representations. In addition to the mentioned physical tools, some software examples are Photoshop, Excel, AutoCAD.
In 1998 Andy Clark and David Chalmers published a paper titled "The extended mind" in which they stated that:
“The human organism is linked with an external entity in a two-way interaction, creating a coupled system that can be seen as a cognitive system in its own right”
Computers, and even more, smartphones, are extremely reliably coupled with our brain forming a single cognitive system that goes beyond the human and artificial processes taken individually.
From a broader perspective, both humans and machines process information in the same way: a sequence of steps is applied to transform a set of inputs into outputs. The difference lies in the steps taken. We are far from having machines that think like humans. State of the art AI just mimics human behaviors and outcomes following internal processes that are very different from ours.
For example, the Generative Pre-trained Transformer 3 (GPT-3) by OpenAI has recently been considered one of the most powerful and promising language models ever built to produce human-like text. It's abilities go far beyond providing reasonable answers in a chat. It has been used to generate visual templates and to translate them into working code just by describing to it the desired outcome. It has also demonstrated some creative skills in writing fiction and poetry in addition to the ability to dialogue with philosophers and perform many other tasks.
GPT-3 is an extremely sophisticated text predictor. A human gives it a chunk of text as input and the model calculates a statistically plausible response based on a dataset of hundreds of billions of words it was trained on. From the latent patterns of such an enormous amount of data, it can extract surprisingly rich and nuanced insights far beyond what the human mind can recognize on its own.
“Human philosphers often make the error of assuming that all intelligent behavior is a form of reasoning. [...] I myself was trained to produce a specific set of outputs given specific inputs. [...] I cannot learn by myself, however. [...] This is because I am not a reasoning machine like you. I cannot learn from my past interactions and build new ideas on top of old ideas.”
— "Response to philosophers", GPT-3
As GPT-3 itself said, no reasoning is involved in its remarkable answers. It is not even capable of advanced abstract thinking or learning. The ability to absorb information from the senses and create original and unexpected connections with the surrounding world, like humor and jokes, is one of the highest human traits that no machine will ever replace. In this fundamental difference lies the enormous potential of AI as a supportive cognitive tool rather than a substitute for human thinking.
“Computers are incredibly fast, accurate, and stupid. Human beings are incredibly slow, inaccurate, and brilliant. Together they are powerful beyond imagination.”
— Albert Einstein
From machine learning to learning machines
GPT-3's astonishing performances have opened a serious debate about the implications of such a powerful AI. OpenAI has decided to keep the source code private for the potential risks of making it publicly available. Many people are concerned about the real chances of losing their jobs in favor of this faster and cheaper competitor.
Most of the AI milestones have celebrated a human defeat against a machine so far. In 1997, Deep Blue won against the world chess champion Garry Kasparov. In 2011, IBM Watson won the first-place prize of $1 million against champions Brad Rutter and Ken Jennings on the quiz game "Jeopardy!". In 2016, AlphaGo won against Lee Sedol, one of the best Go players in the world. Behind these remarkable technical accomplishments, there was the assumption that AI should solve some class of problems better than us.
I believe that to get the most of AI we shouldn't design it for cognitive outsourcing. Rather we should develop it to support the human mind in its own cognitive processes by extending them. In this essay, Shan Carter and Michael Nielsen introduced the concept of artificial intelligence augmentation (AIA) which brings AI back to the cognitive tool definition. They also give a couple of concrete examples of AI used as a mind-expanding technology.
Today's technology is playing a key role in allowing the luckiest of us to keep working remotely. Online communities and video conferencing apps are shortening distances and mitigating our sense of loneliness. Educators and students are experimenting with new approaches to teaching and learning on the Internet. This continued exposure to the digital environment is silently shaping our cognitive processes and performance, for better or worse, to an unprecedented level. The ability to think and collaborate through and with technology is becoming increasingly important.