Winning the War on Error: Solving Halting Problem, Curing Cancer
summary
Claims a practical workaround for the halting problem using abstracting abstract machines and abstract interpretation, applied in DARPA PAC app-audits where accuracy rises from 0% to 100% by later engagements; details include address abstraction, store-allocated continuations and SAT solvers. Proposes a computable pipeline for precision medicine illustrated by NGLY1/prevacid discovery: trio sequencing diagnosis, worm and fly models (survival up to ~80–90%), in silico docking that flagged prevacid, medicinal chemistry to reach ~60× potency, and a UAB clinic matching drugs for ~71% of cancer patients.
AI-written summary · about
chapters (28)
- 0:00Intro
- 1:06The War on Error
- 2:08Unifying theme
- 2:25Main ideas
- 4:09Fixing Software
- 5:13Need to engineer software.
- 5:30an engineer uses prediction
- 6:10example: bridges
- 8:17DARPA: CRASH
- 8:37Solve the halting problem.
- 8:50Play the static analysis game.
- 10:04Abstract Interpretation
- 11:06Interpreter
- 13:19Step 1: Diagnose the problem
- 14:12Step 2: Abstract
- 23:52optimizing abstract abstract machines
- 25:22Complexity attacks
- 25:49What about people?
- 26:09What is precision medicine?
- 26:49Cancer is not a disease.
- 27:00Curing rare diseases.
- 29:27INSERT alanine
- 30:26Rare diseases...
- 31:54Tumors evolve!
- 47:5070 compounds!
- 54:05Precision oncology
- 54:24Handling rare
- 58:32What about cancer?
related talks
53mThe Algorithm for Precision Medicine
50mThe algorithm for precision medicine
07mCollaboration in AI benefits humanity
04mThe Intersection of Neuroscience, AI, and Hardware: A Conversation
09mCan Tech Eliminate death?
20mTranslating a Trillion Pointo Diseases of Data into Therapies
13mWhen AI Gets a White Coat
21mTransforming Healthcare Through Data-Driven Intelligence