Why the Next 10 Years May Add 50 to Your Lifespan | Dr. Derya Unutmaz

He explains the difference between instant models and reasoning/pro models. The pro model thinks for minutes, iterating on scenarios. He pushed GPT-5.5 Pro to think for 2 hours on a dataset, which is like years of human thinking. This allows it to not just analyze data but generate deep insights and predict experimental outcomes with high accuracy.

He explains that even a few hundred data points (like the UK BioBank) have high predictive power. By providing daily data (glucose, steps, sleep) and keeping it in the same context window, the AI can correlate changes (e.g., starting a supplement) with outcomes. He suggests using a personal database or 'Viki' (like Karpathy described) to maintain long-term context.

He discusses the UK BioBank study where AI predicted ~1,000 diseases from data collected years prior. He emphasizes that cancer takes years to develop and leaves traces in metabolism, proteins, etc. An AI health coach could analyze this data and say, 'If you continue this lifestyle, your chance of getting this disease is 85%.'

He explains that AI is expanding exponentially, not linearly. The next 10 years will see more advancement than the last century. This acceleration will lead to a point where every year you live adds more than a year to your life. He cites GLP-1 drugs and muscle generators as early examples of lifespan-extending technologies. He believes we will reach a point in 15-20 years where we can completely reverse the aging process, taking an 80-year-old back to 30 or 40.

He describes testing GPT-5.5 Pro by giving it a complex experiment he had already completed. The model predicted the outcome with 98% accuracy, matching his own intuition. He pushed the model to 'think' for two hours on a massive dataset, producing a 40-page report with insights and next questions. He finds this 'mind-boggling' because the model acquired experience he spent 30 years building.

He explains the revolution of immunotherapy and CAR-T. He highlights mRNA vaccines as a way to train the immune system against a patient's specific mutations. AI will model every possible mutation and screen millions of compounds, leading to hundreds of new drugs per month. He cites the story of a man using AI to create an mRNA vaccine for his dog's melanoma.

He references the UK BioBank study where AI predicted ~1,000 diseases from data collected years prior. He argues that cancer doesn't develop in days; there are always signs in metabolism, proteins, etc. He envisions everyone having an AI health coach that continuously analyzes data (glucose, proteins, hormones) to predict and prevent disease.