Hey, my name is Sammy Hansali 👋.
I am currently a researcher at the Levin Lab at Tufts University and a 3rd year Chemistry undergraduate student at Rutgers University.
Today I’m going to attempt to explain my research in 3 different levels of sophistication:
As Richard Feynman once said, “If you cannot explain something in simple terms, you don't understand it.”
To make sure I really understand, for each level I will limit my explanation to:
20 words
200 words
500 words
Let’s start!
Level I: 5 year old
Work in progress!
Level II: STEM college student
II. 20 words
What enables agents to work together to produce a new agent with higher cognition? Forced stress (and other information) sharing?
II. 200 words
We want to know:
What enables the scaling of cognition?
Cognition = describes the spatiotemporal scale of the goals an agent can pursue (see the “Cognitive lightcones” below for semi-quantitative comparisons of cognition between diverse intelligences)
For now, we will discuss this similar question:
How did evolution enable individual cells to work together to create a new multicellular agent with higher cognition than its constituent parts?
Our hypothesis:
Multicellularity happened when cells that were bound together started sharing stress. When one cell got stressed, the other cells did too. This forced these cells to team-up to minimize the sum of everyone’s stress.
Stress = the biological response to not reaching a goal.
Multicellularity increases stress capacity. Increasing stress capacity increases cognition.
The Research Project
Our project will quantify the evolutionary dynamics behind the scaling of cognition.
We’ve built a tool to study the conditions that are required for individual cells to work together to solve different body patterning problems.
Implications
We hope that the results of this work will aid 1) treating diseases using the intelligence of biology, 2) the safety of the fields of AI and synthetic biology, and 3) more.
II. 500 words
We want to know:
What enables the scaling of cognition?
Cognition = describes the spatiotemporal scale of the goals an agent can pursue (see the “Cognitive lightcones” below for semi-quantitative comparisons of cognition between diverse intelligences)
For now, we will discuss this similar question:
How did evolution enable individual cells to work together to create a new multicellular agent with higher cognition than its constituent parts?
Our hypothesis:
Multicellularity happened when cells that were bound together started sharing stress. When one cell got stressed, the other cells did too. This forced these cells to team-up to minimize the sum of everyone’s stress.
Stress = the biological response to not reaching a goal.
But what does multicellularity have to do with increased cognition?
Imagine a bunch of people stranded on a deserted island.
They are far from each other, so each person must be self-sufficient to survive.
Many of these people will die out.
Now imagine if they grouped together. Idealistically, individuals will specialize in what they are best at to maximize the survival of the group. The athletic will mostly hunt, the observant will mostly forage, the smart will mostly invent, etc.
The group will likely survive and may eventually live comfortably.
In the individuals scenario, everyone is stressed about themselves.
They pursue goals on a small timeframe (trying to survive one day at a time), and can only be one place at a time.
The “unicellular” groups have lower cognitive capacity.
In the group scenario, everyone is stressed about the group as well as themselves. If one dies, gets hurt, or gets lazy, the group’s stress increases. To minimize it, they will take action (work harder, trial, cannibalism).
They pursue goals on a small timeframe at first, but that expands once they’re live comfortably (pursuing children, fitness, philosophy, etc). They can also pursue goals farther in space than an individual.
The “multicellular” group has higher cognitive capacity.
The Research Project
Our project is taking the first step in quantifying the evolutionary dynamics behind the scaling of cognition.
We’ve built a tool that lets us study the conditions that are required for individual cells to work together to solve different body patterning problems.
Implications
We hope that pursuing this work will help advance the fields of AI, synthetic biology, biomedicine, and more.
Biomedicine
Each level of biological organization has its own cognitive capacity… therefore each level also has traits like memory and learning capacity, to different extents.
That means you can train biology to do what you want (with the right reward mechanism at the right scale and organization)!
Train: lungs to prevent cancer, limbs to regenerate, and organ systems to prevent aging. Thats the next era of biomedicine.
AI and Synthetic Biology
The fields of AI and Synthetic Biology are ticking time bombs. They openly admit 0 ability to predict the full extent of what their creations are capable of, a priori.
If we can predict how cognition scales, we can solve this problem (yay for Nick Bostrom)!
Level III: Expert in my field
III. 20 words
What dynamics enable agents to work together to produce a new agent with higher cognition? Forced stress (and other information) sharing?
III. 200 words
We want to know:
What dynamics enable the scaling of cognition?
Cognition = describes the spatiotemporal scale of the goals an agent can pursue (see the “Cognitive lightcones” below for semi-quantitative comparisons of cognition between diverse intelligences)
To simplify this problem, we focus on a special case—the evolutionary transition to multicellularity:
What evolutionary dynamics enable cells to work together to produce a new multicellular agent with higher cognition than its constituent parts?
In biology, homeostatic set points can be modeled as goals. Cells therefore have goals. They reach goals by minimizing stress.
Stress = the biological response to not reaching the setpoint.
Hypothesis:
Evolution scales up the homeostatic drive of stress minimization, by forcing agents to share stress and information, to produce agents with higher cognition and organization.
The Research Project
Our project will quantify the evolutionary dynamics behind the scaling of cognition.
We’ve built a tool to study the conditions that are required for individual cells to work together to solve different body patterning problems.
Implications
We hope that the results of this work will aid 1) treating diseases using the intelligence of biology, 2) the safety of the fields of AI and synthetic biology, and 3) more.
III. 500 words
We want to know:
What dynamics enable the scaling of cognition?
Cognition = describes the spatiotemporal scale of the goals an agent can pursue (see the “Cognitive lightcones” below for semi-quantitative comparisons of cognition between diverse intelligences)
Or more descriptively:
What dynamics enable agents to work together to produce a new agent with higher cognition than its constituent parts?
To simplify this problem, we focus on a special case—the evolutionary transition to multicellularity:
What evolutionary dynamics enable cells to work together to produce a new multicellular agent with higher cognition than its constituent parts?
In biology, homeostatic set points can be modeled as goals. Cells therefore have goals. They reach goals by minimizing stress.
Stress = the biological response to not reaching the setpoint.
Our hypothesis:
Evolution scales up the homeostatic drive of stress minimization to produce agents with higher cognition and organization.
Scaling up homeostasis = Increasing Care, or the number and type of states that an agent can be stressed about. This necessarily increases cognition.
Now I bet your wondering…
How, practically speaking, is homeostasis scaled up in the special case of multicellularity?
Imagine a bunch of seperated cells.
By connecting the cells (with gap junctions) and allowing free information flow, stress generated by one cell is automatically shared with the collective.
If a cell is stressed, each cell in the collective will feel its stress, mistake it as their own, and minimize it with the correct action.
This group coordination via stress sharing allows new multicellular behavior not observed in unicellular organisms — like apoptosis, differentiation, and morphogenesis — to emerge.
The Research Project
Our project is taking the first step in quantifying the evolutionary dynamics behind the scaling of cognition.
We’ve built a tool that lets us study the conditions that are required for individual cells to work together to solve different body patterning problems.
Implications
We hope that pursuing this work will help advance the fields of biomedicine, AI, synthetic biology, and more.
Biomedicine
By acknowledging that each level of biological organization has its own intelligence, goals, and abilities… you also acknowledge biology’s trainability. For example: Organs already know what a disease-free state looks like, so train them to stay there.
Training biology is a more powerful approach to solve cancer, aging, and regeneration than the prevalent molecular pathway approach.
AI and Synthetic Biology
The fields of AI and Synthetic Biology are ticking time bombs. They openly admit 0 ability to predict the full extent of what their creations are capable of, a priori (See GPT3).
If we can predict how cognition scales, we can solve this problem (yay for Nick Bostrom)!
—
After we answer the previous questions, we may be able to answer this one:
“How to give rise to an intelligence capable of doing an arbitrary set of desired tasks, and nothing else?”
Once we have an answer to this, we will become craftsmen of intelligence. AGI. Robots. Organisms. Hybrids. You name it. We can build it.