2020: Hypozooming

You know the anxiety of the blank page? Well the opposite is just as scary. At the beginning of 2020, we were armed with a lot of assembly diagrams, architectural plans, independent modules and so on. In a word, we were preparing a labyrinth. Let a young physicist for 2 years with means to produce proto in an independent lab on a university campus full of students of all sections and you get a revolutionary technology but too raw to be adapted to the market.

Hypozooming is the process by which a raw idea must go through to be revealed like a diamond to the world. It is about cutting into the material, removing pieces and redesigning the cut. This is not done without the help of people who come to bring their experience. You should see the reaction of people when they are told: “We are building a general AI”. That’s when the “you shouldn’t…what if you do instead…” music starts. And here you have two choices:

1) Respond politely, and incorporate the thought into the development.
2) Attempt to fervently explain your vision.

In both cases, the result is not guaranteed.

An AI trained on all the meetings we have done for the project (about 3 trillion) will present us with a result clustered in two categories, unrelated to the strategy of the meeting:

after 11am: great
before 11am: no-go

So the management made a “data-driven” decision by setting the important meetings after 11am.

Not sure that an AI trained on all the data of the world, is really efficient to bring out a pattern adapted to our particular workflow. With this in mind, our hypozooming came from the observation of our own way of working. Let’s reduce the amount of data needed by algorithmic optimization approaches, increase the quality of the training dataset with specialized hardware and the rough diamond “general AI” becomes a “smartcomputer”.

Next step: 2021: Fundraising!

Stay tuned!

#ai #innovation #smartcomputer

2019: Pretotype

First of all, Pretotype is not a typo. We’re talking about a pre-prototype here. The concept is quite simple, before investing heavily in R&D to create a prototype, you need to create a princeps that “looks-like” and “works-like”. In tech industry, this step is known to save a lot of funds allocated to the development of a new technology.

An inventor’s nightmare is to be too far ahead of his time, which results in extraordinary products that do not make it to market because they do not meet the market’s expectations. A pretotype allows to go further than a POC and to confront it from the beginning with potential interested parties. Thanks to such an approach, we can quickly obtain validations, but also “no-gos”. Any such information is precious and needs to be taken in consideration ! There is a « right it », but there is also a « right time », and as Gandalf would say « A wizard is never late, nor is he early. He arrives precisely when he means to ».

After 1 year of blue sky research, building a pretotype also took time. Finding the right skills, meeting the right people and from the beginning…negotiating! There is no secret, when it comes to R&D, every penny counts! Whether it’s walking through the labs at EPFL, discovering the excitment of “la-semaine-de-tous-les-possibles” at HEAD-Genève and discussing the non-existence of “zero” at ETH Zurich in Switzerland, reaching out to l’X Ecole Polytechnique and cooling systems with nitrogen at La Sorbonne Université in France, talking about GDPR at TUM in Germany or visiting the industrial design workshops at Politecnico di Milano in Italy, all of this is only possible today because a significant number of people were there when it mattered!

ANTS A.I. Systems is not only about AI and IT. It’s also about design, biology, sociology and much more. Physics is the study of natural phenomenons and in the case of AI, our physics paradigm is the only thing that we, humans, deeply share with AI. Human-machine interaction requires so much more than algorithms, it requires style, sustainability and un-compromised privacy.

In the coming weeks you will discover more about our adventure! Stay tuned!

#ai #innovation #smartcomputer

2018: Blue Sky Research

Blue Sky Research is an approach of innovation that can be described as « science for science ». The goal is to bring together creative minds with materials and means, without any apparent goal. This is one of the approaches that led to the creation of the JPL, which in the 1930s, despite two explosions on the CalTech campus, received even more resources to continue prototyping rocket engines…. a little further away from the campus. This notion has been widely popularized in pop culture with the recurring reference to “Mr. Blue Sky”, a 1978 Electric Light Orchestra song. Indeed, when it comes to showing the evolution of a slightly crazy scientific project in the cinema, we will systematically find this reference.

Far from the “mad scientist” side that culture popularizes, there is a real logic of construction in such an approach to innovation. When researchers are brought together, when a workshop is created with equipment and when exchanges are encouraged, the laws of entropy apply as in any physical system. Time, contextual events and a control of the environment (temperature, pressure, elements of the whole) allow a small laboratory of “blue sky research” to set up a disruptive approach of technology.

If the word « disruptive » is a bit of a buzz-word, the real physics concept that this term represents is “chaos”. Far from being the scary notion of chaos as we can understand it in the western literature, chaos, from a physics point of view, is a sensitive dependency on initial conditions. This dependence, in adequacy with the most fundamental determinism, allows us to appreciate the beauty of the turbulence of the world which allows us, humans, to fly up in the sky or dive deep in the oceans.

No need for physics to feel the effervescence of an innovative environment and to intrinsically adopt an “anthropomorphic” vision of chaos. To create this effervescence and drive it towards convincing results, we must not consider investing in innovation as project management with KPIs to reach. We need another metric. A temperature metric, a pressure metric but from an anthropomorphic point of view.

ANTS A.I. Systems is the result of more than a year of Blue Sky Research back in 2018, the beginning of an exciting journey! Investment in innovation does not have a linear return. I’ll conclude by quoting one of my favorite Physics professors, who, after a discussion on ANTS’ evolution, calmly leaned over, smiled, and said, “You know, an exponential…it starts very low”.

Next chapter; 2019: Pretotype! Stay tuned!

#ai #innovation #smartcomputer