Among blockchain related projects, the two biggest are most probably, on the one hand, projects that want to develop the crypto ecosystem (and improve our “digital societies”) by creating decentralized crypto-economic models and, on the other hand, projects that want to use the crypto-economic aspect of the crypto ecosystem to resolve physical societies flaws/issues. Neuron, the project proposed by Doc.ai, may become a key player in the second category.
What is Doc.ai
Doc.ai is actually the name of the firm developing the Neuron network, which is a decentralized learning platform for machines. The CFO Anthea Chung describes Doc.ai on her LinkedIn profile as “a computational linguistics firm focused on using the latest AI technology to build medical dialog systems and intelligent conversational agents for the next-generation personalized healthcare”. The concept of Neuron, which is what the ICO is about, is that users (peers) send data they wish to share to the network to train artificial intelligence models. This data can be anything from textual to pictural content and the AI models can be anything developed by anyone.
Neuron is primarily designed with the medical assistance in mind, but when you read their documentation, you quickly understand that it can easily have a broader function. Just think 1 minute about it: what if all the developers in the world were incentivized to code AIs to learn from all the data anyone is willing to share? You got it, if Neuron succeeds, it may become a serious competitor to Google AIs.
How it works
Like most of this kind of projects, it is unclear to which extent Neuron is going to use the blockchain in order to be fully operational, but it is sure that the blockchain will be the core of Neuron’s economic incentivization system.
Technically, the network is meant to allow decentralized black boxes, built by a worldwide community of seeders, developers, curators and data donors, to compete in their marketplace. This way, you are not anymore limited to your doctor for your health or to your university for your education. Instead, you should be able to access very accurate diagnoses, advice or predictions thanks to the Neuron network.
Here is a typical use case that you can find in the Neuron’s whitepaper:
- Marcus wants to begin by constructing an actionable, iterative medical (quantified biological) profile for his wife and do his own clinical trial.He contacts NeuRoN and the AI welcomes him, listens to his story, and starts documenting
- He contacts NeuRoN and the AI welcomes him, listens to his story, and starts documenting and learning. He uses his laptop together with his wife to have long conversations in a text-based question and answer format.
- (a) her facebook login
(b) a picture of herself and her medicine cabinet
(c) her to use some cheap wellness devices available on Amazon
(d) her to fill in an online questionnaire
(e) her latest blood results from LabCorp or Quest
(f) a cheap genomic and microbiomic test (spit in a tube send in the mail)
- By doing this Marcus now has a robust real-time medical record and biological profile on his wife, and the AI begins to give targeted education and helps them develop an initial plan of action.
- Marcus has also been awarded tokens, which he can now use to buy a prediction study about her health.
- He proposes a problem statement and associated bounty to the network.
- Find me how I can improve my wife’s situation and a plan for improving the quality of her life.
- The proposal is structured that anyone seeking to learn about the same problem (and put in their own datasets) may participate by funding the project.
- He proposes a bounty of $100 per participant expecting to hit a target of $5,000 and 50 participants with similar datasets.
- He starts to recruit friends and acquaintances for the clinical study of one. The empowerment is exhilarating.
- He broadcasts the proposal on the network and 100 interested parties join in. That makes the total bounty $10,000.
- Now the bounty is communicated to the doc.ai data scientist network.
- Once an initial model is ready for broadcasting to the NeuRoN network, the network receives it and enhances the model using its own data by means of a distributed variant of the stochastic gradient algorithm.
- Several data scientists answer the challenge and a leaderboard is produced with F1 scores.
- The highest F1 score is selected, the prediction model sent to the participants and the prize is awarded.
|Total supply||860 000 000 tokens|
|Conversion rate||1 NRN = 0.10 $USD|
|Hard Cap||~ 30 000 0000 $USD (287 866 667 NRN)|
|Distribution||37% investors – 33% ecosystem incentivization – 30% doc.ai.|
|Use of funds||See below.|
Regarding the use of funds, they say in their Private Placement Memorandum p.16 that “DOC.AI expects that a substantial amount of all of the proceeds of the Offering will be used by DOC.AI to progress the development of the Neuron Network and the Ecosystem” and that’s it, which is a bit vague.
They also say p. 8 that “One-third (1/3) of the NRN Tokens retained by DOC.AI will be locked-up for one (1) year from the closing date of the Offering (expected to be October 12, 2017); one-third (1/3) of the NRN Tokens retained by DOC.AI will be locked-up for two (2) years from the closing date of the Offering; and one-third (1/3) of the NRN Tokens retained by DOC.AI will be locked-up for three (3) years from the closing date of the Offering”, which means that they do not intend to use all their tokens before the end of a 3 years long progressive lock-up. 3 years seems a bit short since this project looks like it is a long-term one.
We get a better idea of what they intend to do with the money in their brochure.
As you can see, there is no particular soft cap, since the project is expected to run even with a 5M $USD funding.
Now the not very shiny part. It seems that the team behind Doc.ai is not very fond of speculation, which is not necessarily a bad thing, but they show this by eventually releasing the tokens at their discretion between the expected beta launch of the Neuron network (spring 2018, as it is shown in the roadmap below) and the crazily far 31 December 2019, which is not appealing at all if you are in a flipping mood!
The team leading Doc.ai is relevant but is not the project’s strength either. It is composed of academic researchers in healthcare and deep learning fields as well as renown entrepreneurs. The MVP is the CEO Walter De Brouwer, an entrepreneur who studied linguistics at Ghent University and became a resident at the University of Cambridge (UK). However, the advisory board is neatly undercrowded and there is no one in the whole team seriously involved in cryptocurrencies, which could explain why the technical aspect is mostly absent from the documentation.
Interview with the CEO
Despite our enthusiasm for the concept of the Doc.ai project, we were really concerned by it’s technical and investment aspects. We had the chance to ask the CEO, Walter De Brouwer, a few questions.
The technical aspect of the Neuron network is quite absent from the documentation and also from the founding team! Is there something you can tell us that will reassure those who may be worried by this? By technical aspect, I mean anything related to the on-chain and off-chain development.
We have githubs of course but we started as a b2b company, developing for our channel partner Deloitte, so these are not accessible. When we started building for Neuron we took the applications and the algorithms and decided to make a new big algorithm: the onboarding of the ai. We are producing our first report on the CNNs next week at the end of the presale and the reports on the LSTMs when the crowdsale starts.
As you can read in the Tokenomics paper we will start with the easy problem of decentralizing the data. At the same time we are setting up community to experiment with parameter servers and new nodes to distribute the optimization methods, that is a lot harder.
Doc.ai’s ICO is a very tempting one for sure! Their Neuron project sounds awesome and even a bit scary when we think about all the possible outcomes of such a concept. It is also refreshing to have a project that aims to be used by everyone rather than just a niche. The team is fine and ambitious but it will clearly need to be well complemented when the ICO is done. The hard cap is smooth, maybe the 37% reserved for the investors is a bit too greedy.
There are two big downsides though. The first one is the severe lack of technical input. There is no working product and all the applications advertised on their website seem to be rather ideas than applied concepts. The technical aspect is also absent from the documentation and the team background, which means that everything is going to be built from scratch only after the end of the crowdsale. The second one is the token distribution, which is expected to happen during spring 2018, but may also happen whenever during the second half of 2018 or the whole 2019 year. To us, it is a big turn off for any flipping intention, especially with that much uncertainty on the market.
For these reasons, we decided to not give Doc.ai any FLIPPING mark, because we think their ICO is just not a flipping opportunity [UPDATE: We set it to 50% to reflect the uncertainty. since flipping is always possible not necessarily lucrative in this case]. It seems to be a good opportunity for a long-term investment though. 7 is the mark we chose for HODLING since it is still a slightly risky bet and potentially a really long-term one.