Hey, and welcome to the third update of the Brave New Teams newsletter.
Nothing takes place in the world whose meaning is not that of some maximum or minimum.
Leonhard Euler, 1744 in De Curvis Elasticis
Leonhard Euler was a Swiss mathematician, physicist, astronomer, geographer, logician, and engineer who made significant and influential discoveries in many branches of mathematics, such as infinitesimal calculus and graph theory. Euler was one of the most eminent mathematicians of the 18th century and is one of the greatest in history. He grew up in Riehen, a village near Basel, and he spent most of his adult life in Saint Petersburg, Russia, and in Berlin, then the capital of Prussia.
Euler is the father of graph theory, which helped him solve the problem of the âSeven Bridges of Königsbergâ and which is the basis for todayâs network analysis (here is more on the story of networks). And for those of you interested in what Leonhard Euler’s followers did with the foundation he laid, here is a gentle introduction to Graph Neural Networks.
A quick update from SuperScript
January provided for an exciting start to the new year as we released the application for the agent-based model and a game to illustrate the power of algorithmic team composition. In addition, we developed the NLP-powered skills classification tool and the associated skills hierarchy.
Launching the agent-based model for teamwork
The agent-based model (ABM) is deployed as a Streamlit app to make it easy to run simulations and analyze data in your web browser (it works well on large screens and iPad, less so on a mobile phone).
Run the model! (in your browser)
Launching the game
The game’s objective is to compose teams of five (out of nine) to solve problems (thatâs 126 possible combinations to assemble the team). The idea is to illustrate the superiority of an algorithm (a ârobotâ) to form teams. The value created by the machine is usually about twice as high as that of a human team assembler. So, how close can you get to ârobot modeâ?
We released the game in public beta in early January. Since then, the game has been played over 260 times. Thank you for all your feedback and suggestions. We have implemented some improvements (e.g., we increased the time you have to pick your team from 5 to 10 seconds) and added more explanations.
Run the game! (in your browser)
NLP-powered skills classification
We use Natural Language Processing (NLP) to generate the relevant skills from a user-defined problem description. The NLP model needs a skills taxonomy as an input. Quite surprisingly, no good skills hierarchies were available. We, therefore, built a three-level skills hierarchy from scratch that âalmostâ meets the MESE principle (mutually exclusive and collectively exhaustive). The individual skills are primarily sourced from EMSI.
We employed various embedding models that were trained using transfer learning, starting with a pre-trained model and using a small dataset of tagged project descriptions to fine-tune the model. The âSentenceTransformersâ library was used, which is built on top of the Hugging Face library.
We expect to share a link to a prototype of the tool (deployed as an API) in one of the upcoming newsletters.
Status
The purpose of SuperScript remains unchanged. It is to build optimal teams to solve complex problems. The optimization algorithm – step 3 – has been developed (part of the ABM). The optimization step assumes that problems are well defined and that the required skill mix and the worker skill portfolios are known. To create the inputs for the optimizer, we built the missing puzzle pieces, namely steps 1 and 2.
The diagram below illustrates the high-level model setup of SuperScript. Step 3, the optimization algorithm is developed (and will later be enhanced with reinforcement learning to make it run faster and deliver solutions that converge faster to the global optimum). Step 2 is about finding the optimal skill mix for a given portfolio. To do this, pattern recognition techniques are employed, i.e., required skills of past projects are used to predict the needed skill mix for the current project. Step 1 is about the classification of the problems and about understanding their (deep) structure.
Road to the MVP (aka the roadmap)
Now that prototypes exist for both the NLP-powered skills classification model and the agent-based model (ABM, with algorithms to compose teams optimally), the focus is now on integrating the two across a logical user flow. As a consequence, the priority for the next few months until summer 2022 is the creation of a holistic user flow that combines NLP and ABM (or, in other words, the focus is on the âarrowsâ rather than the âboxesâ) :
- Defining the MVP in terms of features, user flow (UI/UX), and technology stack (thereby combining the NLP and ABM models) to prepare for backend and front end development
- Applying fast reinforcement learning methods to complement (or replace) the optimization algorithm behind team assembly
- Improving the Streamlit application for the agent-based model to make its usage more intuitive (and self-explanatory)
- Updating pitch materials (narrative, deck, and video) to be prepared for pitching
- Participating in pitching competitions (and maybe in an accelerator program) to gather feedback (and maybe find a co-founder)
- Continuing (informal) discussions with representatives of medium-to-large-sized organizations to gather feedback
An experiment on the Blockchain
Thereâs a lot of talk (and hype!) these days about cryptocurrencies, the underlying infrastructure like Ethereum, and applications built on top of it. While many activities in this space are flawed and sometimes outright ridiculous, we believe that the technology and the fundamental beliefs are sound. (A good article and podcast on why cryptocurrencies are here to stay and how they will likely disrupt many areas of finance). Since a crypto-based future of work (article from Dror Poleg, his âhype-free cryptoâ course is highly recommended) is very close to SuperScriptâs vision of a more open, less credentialist, and more meritocratic future (see here, here, and here), we plan to experiment with setting up a DAO.
What is a DAO? DAO stands for decentralized autonomous organization. It is an organization represented by rules encoded as a computer program that is transparent, controlled by the organization members, and not influenced by a central government. As the rules are embedded into the code, no managers are needed, thus removing any bureaucracy or hierarchy hurdles.
As part of this experiment, we want to achieve the following:
- Developing the game-theoretical aspects of collaboration and coordination as part of a reputation staking mechanism
- Improving our understanding of novel (less hierarchical) governance frameworks
- Testing (if and) how roles, teams, and hierarchies, as well as tasks, can be reconfigured dynamically in a version-control-inspired framework
The DAO, named MetaScript, should ultimately deliver a novel and open-source version for collaboration and coordination that seamlessly integrates with the other elements of SuperScript (see diagram below).
Endnote
Thank you for subscribing to Brave New Teams. There is a lot of talk about the future of work, the role of automation, and robots. But, as discussed in the post Automatic for the People, there is plenty of reason to remain optimistic about an augmented (and still human) future of knowledge work.
The next newsletter will discuss NLP-powered text classification, the Blockchain experiment and introduce SuperScriptâs user flow for an MVP. And much more!
We are committed to building in public and will accompany the journey with newsletter updates in a (more or less) monthly rhythm.
Stay safe, and donât lose the script.
