Reality happens a single time and the distance between idea and implementation is collapsing so the stakes for imagining and selecting options for what to do has increased by magnitudes. Creation of artificial recursive self improvement(RSI) is contextualized by the broader RSI loop that we are in. The collective intelligence of civilization has been exponentiating for the last 10,000 or so years. Language models (LM) display traits of artificial general intelligence (AGI) and automating LM research could unlock artificial RSI. Large deltas in civilization are typically planned and executed over months and years but the distance between idea and implementation is collapsing. Change taking time and being incremental is frustrating because of the delayed gratification of seeing your results, but also it provides a smooth ramp to think about what is being built and observe the progress.
Civilization's collective ability to communicate needs to scale with our technology's intelligence to reach a desirable future destination. The last 1,000 years of collective intelligence scaling has been a result of the improving ability for people to communicate with each other. Some examples include the printing press, the internet and smartphones. The more people can exchange new ideas, analyze them and have meaningful discussion the more we can combine our efforts in the pursuit of meaningful deltas. Reviewing work or ideas with a fresh set of eyes provides analysis from a different perspective and sufficiently qualified discussion leads to meaningful insights that allows the work to be improved upon and for the loop to continue. The efficiency of this loop is determined by the iteration speed and the discussion quality.
More communication alone is not a good thing, because there is incentive to game it and systems that improve communication efficiency will follow up advancements in communication. For example: publishers, theatres, stadiums and now social media algorithms. Publishers gate the established distribution channels whether for books, newspapers, video games, music, etc; Theatres were established channels to enjoy content like plays, performances and movies. Stadiums in a similar vein for live events like sports or musical performances. Social media algorithms capture our attention with content from our communities and around the world. It allows people to watch and discuss content – recorded or live. The peers you interact with could be from any background and they share the common instinct to engage with the content being discussed. These systems are how we allocate the creation, experience and discussion of media, and while not perfect they have been necessary to incentivize participation in our collective communication.
Art connects to people by invoking their imagination and emotions and it is impactful when it influences how they think going forward. Media is inherently social for example movies, shows, literature that impact people are discussed and critiqued. Art can rehearse a future for technology by imagining a trajectory with state transitions that can’t be confirmed as possible. These rehearsals influence the planned trajectories and thereby execution of civilization. A few examples are: flying objects (which became planes), universal knowledge (google), instant messaging like hermes (which become telecommunications), personal assistants like HAL 9000 or J.A.R.V.I.S (chat gpt or llm assistants). Art, specifically that about the future, is the medium for collective imagination about tools useful to humans and the intangible problems civilization faces even in our next several magnitudes of technological jumps.
The representation of Jarvis in the Iron Man movies as an autonomous personal assistant that speaks to Tony Stark and does work for him captivated people by imagining a tool perfectly fit for the life that he lives. We can imagine through art what our life would be like with access to those technologies. Personal voice assistants have had concentrated innovation in recent years and products like WisprFlow directly mention Jarvis as a north star1sourceWispr Flow — Our vision is to build Jarvis: the voice interface for AI. People who are impacted enough by voice assistants in media begin to imagine its use in day to day life and may even be inspired to plan or contribute to a trajectory that realizes this technology. In their day to day actions their imagination of how this technology is useful shapes the action and states they move through. The idea of autonomous voice assistants has been around for decades; HAL 9000 came out in a move in 1968. Technology with respect to voice assistants has trudged along for the last 60 years with massive advancements in intelligence, voice synthesis, data throughput over the internet, latency, voice understanding to the point now where we can now honestly claim that we have voice agents that could feasibly use tools and take actions in the real world.
As the loop of idea to reality goes to zero our ability to imagine the next states to move reality to and choosing the right ones is the bottleneck for a broadly flourishing future for humans. There needs to be more impactful and higher quality art created that depicts what we want the future to look like, what we are scared of and how we solve the challenges that we face. There needs to be better search and recommendation algorithms that put art that is high quality, relevant, and contributes to some broader objective that has personal benefits and benefits to the community in front of the user. There needs to be better discussion mediums that let people from shared and differing perspectives engage in productive conversation that respect the freedom of communication while also respecting the importance of your attention span. As the loop for creation and feedback for art and imagination get better and faster the ideas that the collective gravitates to should bubble up.
Aligning an imagined future for humans is one of the largest friction points in civilization right now. Societal issues stem from imagined futures and how policy actions could change them. One part of society imagines a future with RSI and AGI benefitting them and civilization while the rest isn’t quite sure where they fit into this imagined future. It’s harder to imagine the future when the size of deltas are exponentiated and we don’t even know which states are even options to choose from. Exponentiating human communication is the only way for civilization to catch up and imagine futures that include flourishing for all of us because right now no individual or organization can lead us through imagining and selecting which single path of reality we should go down.
Generative models compress an output to our desired intent; this is similar to how the printing press compresses an output to the intent of making an intended number of copies or how a video camera can compress the experience of watching something into film or a video that's intended to be replayed. This new technology won’t advance our communication ability using our preexisting systems alone. To exponentiate quality of human communication with generative models new systems of curating, creating, consuming and discussing is needed.
Discovering what these new systems should be, how they should function and building then distributing them is as existential a problem as the intelligence and capabilities of artificial intelligence. This new system should be adaptive and flexible to the creator and the audience. Below are two examples of how AI could enhance the creation and consumption of art.
A world you can spend days in. An artist wants to paint a picture depicting a street in Europe. They use AI to choose the materials they need and get objective feedback and advice while they paint. The AI asynchronously prepares the steps for printing the painting to reduce the cost and lead time. A person who wants to see the painting can go see it physically or go to an immersive experience generated by AI and tuned by the artist to allow the user to experience the painting in full view, but also allow them to step into the painting and have their own journey with it. If it's depicting a street in Europe it may be stepping into the painting as a person on the street walking into a coffee shop and experiencing their life based on your own interpretation of the art. This experience could allow you to be more impacted by the art and have a deeper analysis or it could simply be more enjoyable.
Rehearsing a live performance. A dance company is preparing for a 90 minute performance where they are telling a story, but also presenting their art of dancing. The choreography is built in a 3d simulation where the different movements of the dancers are visualized. Teaching and learning the dance moves happen in person with live recordings giving nudges and feedback based on its analysis. Dancers can practice at home simulating the other dancers moving around them to build spatial awareness. The ramp up and down of practice and pieces is aligned with performance dates to minimize wear and tear and have peak performance.