When all the greatest things get fused into one: AI using neural networks to learn about healthcare through global incentivised cryptoeconomic ecosystem. The future is here!
Hey there! Please write a bit more about why you are interested in this project and what research you’ve done We’ve introduced some rules when it comes to ICO discussions in order to avoid spam https://thecrypto.pub/t/about-the-upcoming-icos-category
I’ve looked into this system and will try to break down the idea as I see it. I’m also not invested in this project in any way but starting to think about it.
The current healthcare system can be considered a top down approach. Patient has a problem -> practitioner learns about their symptoms -> practitioner may refer the patient a number of times -> a practitioner tries to establish a root cause and prescribes a treatment based on their understanding of the scientific literature.
• As science progresses the knowledge base for general practitioners increases
• As science progresses the complexity of specialisations increases.
• Good rigorous science tends to be slow.
• Large observational studies are expensive for the researchers.
• Self-diagnosing is very hard
Potential solutions offered by doc.ai:
• Use AI to take pressure off general practitioners by providing consultation and prescriptions directly to the patient.
• Use AI to support general practitioners by providing relevant patient information and suggestions.
• Use AI to support specialists by drawing on subtle and easily overlooked possibilities.
• Provide a system that relies on fundamental data only, avoiding the need for lengthy analysis and peer review.
• Provide a database that can be the foundation for almost any observational.
• Minimise admin and logistics associated with large observational studies by incentivising patients to enter their own data as they accumulate it.
Doc.ai uses a bottom up approach powered by machine learning. It takes fundamental facts about your health (from blood tests, geneitics, diet, microbiome and more) -> it runs them through models and relationships observed from the population -> provides general recommendations on health and specific recommendations based on illnesses and treatments as they arise.
I see this as being complimentary to the current healthcare system, and if it gets traction could improve patient outcomes directly through accurate prescriptions and indirectly through accelerating research. Studies have been done using machine learning to predict illnesses with good results before, so not all the tech here is new.
I’m not sure this needs to have its own token, but they are trying to take advantage of cryptoeconomics, as are many other projects. What are the risks here?
So what is your opinion on this project? Will it be massive as suggested by the size of the healthcare industry or will it have to settle for some niche application? Will it explode on the crypto market because it sounds awesome? Is the tech and market opportunity promising enough to invest?
- the problem that they want to solve is very complex. Predict your illness based on medical data? I would expect a serious number of medical experts + AI experts working together.
- why do they need a token? I did not really get it yet. If I will be able to only buy some medical service inside their system, that means that this token will gain value only if the service is very very popular and very high quality. In this case this ICO is more like IPO. The token has significant potential to grow only if it solves problems of storing/sending value or selling/buying a valuable or popular goods/services. I don’t really see what value am I going to get with their token now
Yes, I think predictions are good when there is lots of data collected on them but not good for lesser known illness - this is simply the nature of AI. I think a huge part of the benefit in medical AI is around these common illness, it blows me away to see the number of people going to hospital because they have a cough. They are that uninformed about the common cold and don’t realise that they take the time of busy professionals. I wonder if a whole government would get behind doc.ai?
Here’s an example of wasted resources: 20% of Emergency room visits are dental emergencies. Something the ER has no resources to treat. They give a few prescriptions and tell them to see their dentist.
I think AI will greatly aid in the treatment of illnesses. As it stands now, physicians are all trained using a decision tree model to diagnose and treat. If symptom X, then possible cause Y and or Z. Multiple symptoms narrow down the possible causes. All this can be computerized and learned by AI. Proactive patients can benefit by visiting the physician with most of the diagnoses already accomplished. Reactive patients could be instructed to fill out their signs and symptoms on a tablet in the waiting room and have the AI save the physician a lot of wasted time. Add to this the cross referencing the variety of patient medications for adverse reactions and the AI will make this easy to prevent over medication.
Wow, that’a big savings for hospitals if they can filter those out early.
I like this cross referencing idea! I think this concept is of huge value because it’s one of the slowest things for humans to do, while being one of the fastest things for computers to do, and AI will be able to cross reference things we don’t even know to check for.
From what I have seen, I am sceptical about this project. I work in healthcare and my organisation is funding ground-breaking machine-learning projects. For example, people with cystic fibrosis suffer exacerbations (severe infections that damage the lungs) resulting in hospital admissions, intravenous antibiotics and reduced life-expectancy. Through machine-learning it is possible to predict, and even prevent, these exacerbations. Days before an exacerbation occurs vital signs change and activity levels drop. We have been monitoring hundreds of patients over many years.
Undertaking research in this small area of health has involved - partnerships, the testing of and supply of medical / monitoring devices, clinical trials, peer-review and debate amongst specialist clinicians on when and how patient interventions should be made based on the data. Models of healthcare will not change without clinical consensus. If you get the parameters wrong - there could be too many clinical interventions placing a greater burden on medical teams or too few inventions making patient outcomes worse. Also, what are the unintended consequences of patients spending less time in hospital - it’s not just about the treatment, but psychological support etc. The consequences of getting these things wrong could be life-threatening.
These important issues are not reflected in your proposal.
You only talk about collecting data and analysing it in very broad terms.
There is massive potential for healthcare to benefit from blockchain technology but, for reasons I have mentioned, the industry is very slow to adopt. To put this into perspective, the UK National Health Service is the biggest purchaser of fax machines in the world and uses Windows 95.
Healthcare projects on the blockchain need to focus on very specific problems that the blockchain can solve. Not a general approach promising to change the world. Team also need a roadmap that highlights partnerships and how clinical support is likely to be achieved - this does not mean having a doctor on the team, it means engaging with patient organisations, medical associations and trade bodies, trials, publishing emperical evidence and debate.
That is very good perspective. I know that doc.ai is really hot in the valley and thats causing quite a lot of hype, its easy to get caught up in all the articles and chatter about doc.ai. But given what you said, it looks like doc.ai is more risky than meets the eye.
Yea, the hype is real! It is well presented and had a well sourced team. But I couldn’t find a realistic assessment of the challenges they face, it was all quite glossy. I’m optimistic that further developments in AI will make something out of this line of thought, but I admit as excited as I am about the concept I wasn’t confident in doc.ai enough to throw money at it.
@jrbarrow, I’d suggest the reason why the issues you raised are not reflected in my proposal and why I cover it in broad terms is partly due to the nature of AI. For a neural network to develop most accurately we give it the raw data. No controls, no models, no statistics no causation. There are no parameters, only the raw data and a neural network which runs on it’s own steam. The neural framework will be developed by computer scientists and the black box evolves from there. So it is the opposite to the good science you describe by definition. It’s quite a paradigm shift to throw everything into a black box compared to connecting each and every dot with good science.
Does this paint a clearer picture of the advantages of this approach? Looking forward to your criticism once again! =)
@genebeveridge thanks for your response. It makes sense… in terms of collecting raw data, I’m sure the project will be a success. I was just making the point that from there, even after analysing the data, there is a lot of work to do before there is any impact on the provision of healthcare. I hope to be on the constructive side of critical
So their success rides largely on the incentives perceived by the user of the system. This is a bold project for sure and I wonder if tackling a smaller one first would give them some more credibility.
And yea, it’s good that we can get stuck into this, not only are these ideas interesting, there is money to be made by getting the facts right and picking that winner early!
It seems to me that AI and a cryptocurrency are two very different things, and that the latter is only a funding mechanism for the former.
Did anyone read that an AI machine beat the world champion Go masters recently. The programmers took a very unique approach to teaching their machine how to win. They just taught it the rules and let it play.
It seems to me that taking a similar approach can benefit Doc AI. Instead of a list of illnesses and symptoms, teach it how the body works, then let it run contemporaneously with existing doctor intake procedures. Of course, it will be able to “look over the shoulder” of many doctors across the country at the same time.
The decision tree concept described by Doc has some significant limitations, including the diagnosis being limited to things the diagnosing doctor is familiar with. The benefit of AI in this style is just a larger database of known illnesses to pick from. This hardly seems like progress.
My perspective is that entrenched interests in healthcare want to make changes on the margins to protect their turf. Something that saves money is great, like a tablet that allows a patient to input their symptoms instead of having a human being do that. However, radical progress is not made by small steps.