Building Jypsor: the world’s most advanced research interface.

In this blog, we discuss how Jypsor is set to assist researchers in securing the most important technological breakthroughs of this century & beyond. We started by asking the pivotal question - “how should an ideal research software behave?”, and the answer became increasingly obvious after some drudgery.

A vision: the ideal research software.

I’d like us to imagine a choronaut from Type 1 future human society. An impactful question to ask from such a being is: “how did we solve our grand challenges?”, though still second to just being smart and ask for solutions. “We believe that human advancement goes on an exponential curve, but universe’s increasing complexity flattens out uncontrolled progress.”, “Our best minds communicate necessary actions, via language and with sufficient experimentally gathered data to achieve ineffable productivity.”, “We essentially removed all bottlenecks. Ex. writing code manually.”, “Our tool that can use its own reasoning to autonomously conduct experiments and deduce important information” are arguably the most logical answer predicts. Now, so what ‘thing’ fits this analogy the best? For a modern researcher whether in academia pursing their PhD or doing their daily job at some recognized model lab - I’d hope the evidence is overwhelming in support of, ‘an eerily obedient, super smart colleague’ that sits at your desk all day, waiting for instructions to labor on, not showing external tiredness or lack of motivation. Now we incorporate silicon systems to remove any ethical boundaries & human constraints, and the conclusion is we’re looking for a tool powered with artifical general intelligence.

The synergy: a step in the right direction.

We’re proud that humans have far progressed since the original inception of this idea. As Sir Demis Hassabis, leader of one of the most consequential AI labs in human history puts it: “step one, solve intelligence; step two, use it to solve everything else.” We’re aware that great minds are on it. As a significant breakthrough in 2017, researchers at Google discovered the transformer architecture. They had found the solution to training models that can find complex non-linear patterns over any distance, which was previously a huge issue with RNNs/LSTMs.

On another note, humans have been studying the universe for over 30,000 years and developed complex thoughts as back as 135,000 years ago. Sometime in between, around 50,000 years ago, we had the first versions of modern language in the form of speech. And it wasn’t until around 3,800 years ago that the traces of our activities indicate we built a functional writing system, wherein individual letters represented single phonetic sounds. We have since structured our thoughts using this system, primarily for communicative/expressive or archival reasons.

When the synergy between our these two creations truly showed itself, it felt magical. The first use of language models was as incredibly smart, and context aware autocomplete bots. Our computers finally understood text, and meanings deeply engraved as it was displayed in the material they wrote. Therefore, an obvious next step was to make it conversational. ‘Role-play’ between a human user, and an ‘AI-chatbot’ was introduced.

The mission: forging reality with work.

Looking back at all this, we’re are now at a stage where communication between humans & software in natural language is possible. But the bigger challenge of AGI remains, and so do other important ones in diverse fields - e.g. cancer cure, climate control, quantum sciences etc. Crux is: “are we giving our researchers the best software for their jobs?”. The answer to this was dissapointing. Data scientists still write repetitive code when their task should be to express fundamental math naturally, via language. Biologists hit gridlock in best using analysis tools, because current notebooks prevent them to. Students often cannot directly contribute ideas because of restricted access to latest hardware.

Jypsor’s mission is to fix this mess. Our goal is to build the world’s most advanced, multidisciplinary, language-driven research notebook. And democratize accelerated compute access to everyone on the planet. With us, a newer more pleasant study format will be common for all. Everyone can express their ideas, and build a civillization with highest levels of productivity. Last and most crucial, we want to accelerate AGI research and benefit all.