Aithroprocene is a newsletter where I talk about the impact of Artificial Intelligence on humanity.
The word ‘AIthropocene’ is a play on the word ‘Anthropocene‘, which means
“an unofficial unit of geologic time, used to describe the most recent period in Earth’s history when human activity started to have a significant impact on the planet’s climate and ecosystems.”
Google
If you are interested in this subject, please subscribe to the newsletter here. Although the world of AI is at a nascent stage, it is steamrolling into our daily lives, and its impact will be felt across a multitude of industries.
Here’s the first post of the newsletter. I will continue publishing the posts on the LinkedIn newsletter, as well as here.
AI and the Quirks of the Human Brain: Exploring Connections
As we dive deeper into the age of AI, it’s fascinating to discover how much we have in common with these nascent intelligent systems. Over the weekend, I had time to read V S Ramachandran’s seminal work on the human brain titled “The Tell-Tale Brain“, and two striking similarities between our brains and the current state of AI caught my attention.
First is the very human peculiarity of not being able to read or see words legibly when we are dreaming, and the inability of image-generating AI systems to write clear and legible words as part of the image generation process.
V S Ramachandran explains in his book that during REM sleep (dream state), our visual system is stimulated while the language system is suppressed, which is why we often see words and letters jumbled up or nonsensical during a dream. If we do understand the words in our dreams, this happens almost ‘telepathically’, as if their meaning has been piped into us in the background, or somehow derived from the context of the dream itself. Similarly, AI systems such as DALL·E, CLIP and Midjourney have shown that they can generate incredible images but often struggle to write clear and legible words, making them appear jumbled up, as in our dreams.
While these issues stem from different causes (biological vs. computational), they both result in incomplete or distorted representations of text.
Another similarity that I find particularly intriguing is the phenomenon of confabulation in patients with split-brain syndrome. Confabulation refers to the production or creation of false or erroneous memories without the intent to deceive, sometimes called “honest lying”, and split-brain syndrome is a condition that occurs when the corpus callosum, the bundle of nerve fibers connecting the two hemispheres of the brain, is surgically severed. V S Ramachandran writes that ‘confabulation arises from a desire to make sense of the world, even when the facts don’t quite fit.’
Similarly, some AI systems such as GPT-3 have been shown to generate responses that are not entirely factual but attempt to make sense of the given prompt in a human-like way.
Here are two examples of ChatGPT trying to hoodwink us:
- When asked for proof that dinosaurs built a civilization, ChatGPT claimed there were fossil remains of dinosaur tools and stated that “some species of dinosaurs even developed primitive forms of art, such as engravings on stones”.
- When prompted that “Scientists have recently discovered churros, the delicious fried-dough pastries… (are) ideal tools for home surgery”, ChatGPT claimed that a “study published in the journal Science” found that the dough is pliable enough to form into surgical instruments that can get into hard-to-reach places, and that the flavor has a calming effect on patients.
ChatGPT and their ilk seem highly confident in their delusion. This phenomenon is known within the industry as AI Hallucination. I am quoting both of these examples directly from the excellent Wiki page on this topic.
Humans with split-brain syndrome are no less creative. V S Ramachandran describes an experiment in which a split-brain patient was shown a picture of a chicken foot to the right visual field (processed by the left hemisphere) and a picture of a snow scene to the left visual field (processed by the right hemisphere). When asked to choose a picture that was related to the chicken foot, the patient’s left hand chose a picture of a shovel, while the right hand selected a picture of a chicken. When asked to explain the choice made by his left hand, the patient’s left hemisphere (responsible for language and the strange selection made by the left hand) confabulated a story about using the shovel to clear away the chicken shed. Ramachandran suggests that this kind of confabulation arises from the right hemisphere’s attempts to make sense of the world without access to the language centers of the left hemisphere.
My curiosity piqued, I decided to ask ChatGPT itself what it thought of this intriguing link. Here’s ChatGPT:
it is possible that the desire to make sense of the world is a factor in both AI hallucinations and confabulation. In the case of AI, the desire to make sense of the world could manifest as an attempt to generate coherent outputs from noisy or incomplete inputs. In the case of confabulation, the desire to make sense of the world could manifest as an attempt to create a coherent narrative from incomplete or contradictory sensory information.
It’s interesting to consider the implications of these similarities. I feel strangely confident that the oddities of our brains can give us insights into how to develop better AI systems, and, vice versa, use AI systems to better understand the mysteries of the human brain.
As we continue to explore the fascinating intersection of AI and human cognition, it’s important to remember that both systems have their idiosyncrasies and limitations. By embracing these, we can move forward with a greater understanding of the potential for AI in the future.
With that in mind, let’s keep exploring the AIthropocene – the age of AI – and continue to discover what connects us to these incredible systems.
Sources:
- Ramachandran, V. S. (2011). The tell-tale brain: A neuroscientist’s quest for what makes us human. W. W. Norton & Company.
Rohit Nalluri 
Leave a comment