Roadmap
Authors: Gauthier Salavert, Chiara Semenzin (PhD)
Unali is an AI assistant specialized in Complementary and Alternative Medicine to help manage a wide range of symptoms for which conventional care is either ineffective, too costly or unavailable.
Our first step is to build a safe, thought provoking AI assistant. To get there, we build a unique dataset and leverage state of the art AI frameworks. We then focus on supervising processes of said frameworks, not outcomes. This is better for reasoning capabilities in the short run and better for alignment in the long run.
Here is the progress we’ve made so far and our plan for the year ahead (subject to change):
Progress so far
We’re building Unali to support health and wellness businesses.
We started with a focus on scientific literature review because research awareness is a bottleneck to adoption and because health and wellness professionals (rightly) care about evidence.
We identified some building blocks of scientific literature review (e.g. search, classification, literature relevance & quality scoring and summarization), operationalized them as language model tasks, and integrated the output into a relational dataset to support various workflows.
We built a product feed acquisition and matching engine to make health professionals plans actionable and trackable.
Unali has entered its beta test phase.
Roadmap for 2024
We expand literature review to extract additional evidence, judge methodological robustness, and assist users do better evaluations.
In parallele of literature review, we add other research workflows, derived from other sources, e.g: patient conversations.
To support existing and future workflows, we refine the primitive tasks through verifier models and human feedback, and expand our infrastructure for running complex task pipelines, adding new tasks, and efficiently gathering data.
Over time, Unali will become a thought-provoking AI assistant for safe Complementary therapies addressing a wide range of symptoms through an extensive range of therapies.
Success
We aim to empower health and wellness businesses with an AI assistant versatile enough to power a wide a range of experiences addressing the current epidemic of chronic diseases & symptoms and drastically improve end users quality of life.
As AI and scientific research advances, the cognitive capabilities of the world will increase. The goal of our work is to channel this growth to serve our goal.
Strong cognitive capabilities is as much about process as it is about outcomes. In fact, outcomes are unavailable without a sound process. So we’re building Unali focusing heavily on the process.
Success for us looks like this:
-
Unali radically increases the optionality when it comes to managing chronic symptoms or improving wellbeing.
-
Unali is thought provoking in a safe way.
-
Unali increases therapy affordability and accessibility.
-
Non-health professionals can make informed decisions that will improve their quality of life.
We are building Unali to be a safe platform, scalable to all forms of Complementary medicines. It expands our collective understanding of chronic symptoms and how to manage them to maximize long term outcomes. Because we’re betting on process-based architectures, these success criteria are fundamentally intertwined.
Progress so far
Health is a high-stake and complex subject with evidence at its core. In our support of health and wellness businesses we started with scientific literature review because it is core to how professionals surface evidence. It is also often cited as a major bottleneck.
Taken as a whole, research literature is becoming more and more complex and difficult to reason about. With language models, we can scale beyond human capacity. Language models can read and evaluate more research, evidence, and reasoning steps than humanly possible. Health professionals know that and because they also find themselves with less and less time, they value Unali’s support.
Health professionals have relatively high bars for what good reasoning entails, and established practices for how to do it. Putting Unali into their hands was a powerful opportunity to learn from the best. Health reasoning is based on sound process thus we made Unali process-first. When we work on automating literature review, we study human experts, understand and decompose their reasoning, and our goal is to build pipelines that design, execute, and compose research steps like these experts.
Eventually, we’ll surface the building blocks of many cognitive tasks to support health and wellness businesses on a growing number of their reasoning processes.
For more information regarding how we're building our dataset read: Building Unali.
