Algorithm Progress

I now have an initial working prototype of the business plan generator – this specific version is focusing specifically on the “company purpose” which is a short paragraph that encompasses an entire business plan. This is also the only piece of writing that the majority of investors require. The prototype is actually made entirely by using php. The algorithm, in this first stage, is essentially replicating the initial prototype I had, but is freed from the restrictions of using a third-party system. This allows me to house the algorithm on my own server, but also, obviously, is more flexible for further development and refinement.

I really enjoy looking at the source code behind such an algorithm – it is really interesting to see the “back end” of a generative business plan. Here are some screenshots:

The system works by creating an initial template ( a series of sentences) that have dynamic elements embedded within them. Each of these dynamic components (words) are then made generative by pulling from an archive of words related to that subject matter.

The system, in this stage, is not very robust – we only have 2 or 3 sets of words it is pulling from, but moving forward the following is being considered / tweaked:

1. “Deep” crafting: instead of generating words, generate pairs of words / phrases. This will give a greater illusion of artificial intelligence, and will also form less of a noticeable relationship between the template and the outputs.
2. Pondering different means of approaching plurals, consonants, vowels, etc…
3. Should the company name be generated? Should that be left out, and be a part of what the human interprets from the machine’s output?
4. I am currently compiling a list of openly available “company purpose” statements in order to analyze more the language used, and the overall sentence structures at hand. This will allow us to break the common elements / forms of these statements into chunks of data that can be grouped into the algorithm.
5. “Flexible simplicity” is the method we are using – this essentially means an attempt to not have too much control, or too little control, but just in between the two.

The (Dis)Embodied Entrepreneur System Flow v2

Taking from the knowledge gained in my initial attempts at building the (Dis)Embodied Entrepreneur, this second iteration of the algorithm is designed with more complex layers that further remove the human / maker from the process of development as a way to push it further towards an autonomous existence.

Where the system, in this mock up, falls short is the final selection. Something that algorithms do especially well is selection based on detailed paramaters. This objectivity is what interests me in developing the entrepreneur. However, because the final step of the algorithm is simply a random selection, the algorithm becomes very subjective, and in a sense loses it’s purpose.

Initial future considerations:

• How can each component of the automation process be drawn out in an obsessively detailed manner?
• How can the system itself become embodied? Can there be a physical / illustrative form to add to the message as a whole?

1,000 Business Plans (part 01)

I am currently working on a new project, “1,000 Businesses,” that is a compilation of 1,000 algorithmically generated business plans. The project aims to step closer towards an automated system (as opposed to the 100% analog card game) in order to begin exploring the kinds of business plans an entrepreneurial machine might be capable of writing.

A bit of a discovery, however, has been found in the process of writing the algorithm itself. The executive summary is the first element to be generated, using my Serendipitous Executive Summary Generator, which spits out something like the following:After generating 1,000 of these summaries, a series of key-terms are extracted from each executive summary (i.e. opportunity, demographic, etc.), forming a database of words to pull from for each plan.

These terms are then manually inputed into the designated space(s) within each of the 1,000 business plans, as dictated by the algorithm.

Here is an example of an exported plan.

I’ve found that the process of generating this first output from my anonymous entrepreneur machine has been very reflective, and some what meditative, because it is so tedious and laborious. I currently have started 200 businesses, and am steadily working towards outputting the rest by the end of the week. I am very satisfied with the results thus far, as the range of businesses are beginning to go beyond the first-level “silly,” and more into the believable, yet strange, realm. While I do have some ideas as to where I might take the project next, I am going to wait until finishing the project to determine the best direction, so expect a post about some trajectories soon to follow.

Serendipitous Executive Summary Generator (SESG)

My colleague Nick Fehr developed an awesome product, MyGenGen, which is a simple tool for creating single-function websites. I decided to leverage this software to develop a set of algorithms (based on word lists and sentence structures) that create generative Executive Summaries. This initial, quick, prototype / result has a sort-of “Mad-Libs” feel to it, which is fun. See it in action here, and enjoy a selection of the initial outputs below.

The system essentially works like this:

1. Start with a sentence structure that has certain {bracket} words.
2. The words within the {bracket} are randomized, pulling from a list of options.
3. The sentenced structure itself is also randomized, pulling from a list of options for sentence structure.
4. Every time the user clicks “GIVE ME ANOTHER BUSINESS MODEL,” the page is refreshed, and a new statement is generated.

I am interested in what might happen if each section of a business plan is taken through a similar algorithm. The result of which would generate a complete business plan (market strategy, sales channels, etc.), written entirely through a generative algorithm.  I am also interested in what might happen if the database of words that are generated are not determined by myself, but perhaps by a collective of individuals.