3. What is going on now?
Major IT companies created medical
AI departments:
• Google – Calico, Deepmind Health
• Facebook – Chan-Zuckerberg
initiative
• Intel – Intel health and science
• IBM – IBM Watson
• Microsoft – Microsoft Health, Allen
Brain Atlas, malaria research in
Gates foundation
4. What is going on now?
Major IT companies created medical
AI departments:
• Global AI startup financing hit $5
bn in 2016 (Nikkei)
• Medicine – 15% of total spendings
5. What is going on now?
From Virtual Nurses To Drug Discovery:
106 Artificial Intelligence Startups In
Healthcare
Deals to healthcare-focused AI startups went up from
less than 20 in 2012 to nearly 70 in 2016.
9. Medical AI tasks overview
1. Medical research and drugs development
2. Disease prevention
3. Diagnostics
4. Therapy
5. Rehabilitation and nursing
10. Non-medical AI tasks for life
extension
1. Brain interfaces (BCI) – This sphere directly related to life extension
as it may eventually lead to “uploading” of human consciousness to a
computer;
2. Human security – AI can do a lot to reduce the non-medical risks;
3. Digital immortality and cryonics
4. Partnership with robots
11. ❶ Narrow AI
❷ AGI = human level AI
❸ Superintelligence
AI epochs:
12. Stages of AI development
Narrow AI. now
Human level AI, or AGI. somewhere in middle а 21 century
Superintelligence, as Bostrom-Yudkowsky - end of 21 century
When?
13. Narrow AI
Aging – cause of death
Combinations of geroprotectors
We need AI to calculate
is most effective in
scientific research
14. Geroprotectors
• Geroprocteotrs research is one of several promising
approaches to fight aging, where AI may help
• Other are genetics manipulations, SENS, regenerative
medicine, microbiota
• Geroprotectors are drugs which prevent aging and
extend life
• Many drugs, supplements and food could extend life
of mice 10-40 per cent individually
15.
16. Will geroprotectors
work on humans?
• We don’t know
• Very long and expensive experiments are needed
• Clinical trial approved only for methformin (but no
funding)
• Aging is not a disease according to WHO
17. Biomarkers of aging based
on Big data will help
• Shorter experiments
• Wearable devices
• Patient organizations – less regulation
• Powerful computer and algorithms are needed to
analyze the data.
21. Data is needed to
train narrow AI
IBM bought Truven for $ 2.6 billion, which has hundred
millions of patients medical records.
http://fortune.com/2016/02/18/ibm-truven-health-acquisition/
22. Startups
Deepmind Healthcare:
Stream app. “Streaming
the right information to
the right clinician at the
right time”
AiCure:
confirms medication
ingestion
Babylon Health:
AI expert system +
videochat
23. Similar applications of AI
to search for new things
Project Atomnet – chemical materials properties
predicting on the basis of convolutional neural networks.
http://www.atomwise.com/introducing-atomnet/
Predicting the electrical properties of new molecules
using a multilayer neural network http://
www.worldcommunitygrid.org/about_us/
viewNewsArticle.do?articleId=450
Ligand properties predictions: http://
www.ncbi.nlm.nih.gov/pmc/articles/PMC4510302/
24. Other application of AI in
medical scientific research
New drug candidates could be found
• Generative neural networks may create plausible candidates
• They could be tested on digital models of cells, molecular pathways or receptors
Model of aging as instability of genetic
networks
• AI could help in calculating complex theoretical models
• “Gero” company use system dynamic to analyse instabilities in genetic networks
Robots in the lab: automation
26. Upgrade Network
Elon Mask: neural
lace and Open AI: net
of self improving
individuals
We: Best way for such
merge: medical AI
• BCI + “neuronet”
• Self-improvement
• Cyborgization
• Human-like AI assistants
• Humoral interface
• Social network of patients
27. Value alignment problem:
Humans are self-improving nodes of the net = no need
to upload values.
It will be best solution, as it doesn’t require one AI taking
over the world
Upgrade Network
28. Human level AI
as personal physician
• Inside your body
• Provide advise
• Language understanding
• Will know source of pain
• Could work as autopilot for human: prevent dangerous
behavior. Ethical?
29. Human level AI as
a research catalyst
Human level AI will accelerate research in life extension
• AI-scientists
• Automated labs
• AI as a research institute
ARUP Automation Systems
30. Human level AI as
a research catalyst
Problem:
1. 80-90 per cent of medical studies are not replicable.
2. Human life extension is especially vulnerable to fraud as
results are remote. Snake oil wins.
Task of verification of truth by AI:
• replicability,
• total recording of experiments,
• blockchain,
• fraud and plagiarism
• detection,
• altruistic selfless agents
31. Human level AI as
a research catalyst
Problem:
1. FDA require tons of legal documents for clinical rials
2. They cost billions
AI lawyer
• prepare documents, read laws
• Etherium-style contracts
38. Surviving until creation
of Strong AI
• The older the person, the less chances
• Longevity escape velocity [de Grey]
• Patient organisations and clinical trials
• Competence
• Accelerate medical AI creation (if it safe)
39. Narrow AI
AGI
Superintelligence
People are dying
Immortality
2 Wearables
3 Big data
1 Upgrade net
2 Humoral interface
3 Robo-scientists
1 Nanotech
2 Uploading
3 Resurrection
1 Geroprotectors