How AI Is Changing Medicine — Drugs, Diagnosis & the Risk - video Dailymotion

DailyAIWizard

In 2025, a drug with both its target and molecule designed entirely by AI completed human clinical trials. It worked. That's just one of the things AI is quietly doing inside medicine right now. In Episode 4 of AI at Work, we go inside healthcare AI — how it reads scans, discovers drugs, writes clinical notes, and personalises treatment to your biology — plus the serious risks of bias, hallucination and a wrong diagnosis.  Hosted by Oliver & Olga, with deep-dives from Anastasia & Sophia.  🔍 What you'll learn: • How AI reads scans and outperforms specialists on specific tasks • The first fully AI-designed drug that worked in human trials • How "self-driving labs" are compressing drug timelines from 10+ years to 3–6 • Why ambient AI scribes cut physician burnout from 51.9% to 38.8% in 30 days • The bias problem: why medical AI can fail minorities and women • Why AI scribes can hallucinate drug dosages — and what's being done about it • What medical foundation models and continuous monitoring mean for your future health   🎙️ About AI at Work A series breaking down how AI is transforming real industries.  Previously: Banking, Insurance, Government. Next: Energy.  🔔 Subscribe:    / @dailyaiwizard    💬 What do you want us to cover next? Tell us in the comments.  #AIinHealthcare #MedicalAI #ArtificialIntelligence #DrugDiscovery #HealthTech #DailyAIWizard  SOURCES 1,247 AI medical devices FDA-authorised Aug 2025; 75%+ radiology — FDA / NCBI / IntuitionLabs, Aug 2025 295 AI/ML medical device 510(k) clearances in 2025 alone — Innolitics, Dec 2025 AI improves radiologist sensitivity 6–8%; pooled sensitivity/specificity  mor than 90% in fracture detection — NCBI fracture review 2025 IDx-DR: first FDA-authorised autonomous AI diagnostic — IntuitionLabs / FDA First fully AI-designed drug (target + molecule) Phase IIa completion, dose-dependent improvement — Drug Target Review Feb 2026; Nature Medicine June 2025 AI drug timelines: 3–6 years potential vs 10+; 80–90% Phase I success vs 40–65% traditional — Lifebit.ai, 2025 Dozens of AI-designed candidates in clinical trials by mid-2025 — ScienceDirect Nov 2025 AlphaFold protein-structure predictions: undruggable targets unlocked — DeepMind / industry Physicians: 5+ hours EHR per 8 hours scheduled; documentation nearly triples burnout odds — Veradigm Mar 2026 / JAMA studies Prior authorisation: 13 hours/week, 39 requests/physician/week, 89% say it increases burnout — AMA survey / Definitive HC Mar 2026 Ambient AI scribe (Abridge): burnout 51.9% → 38.8% in 30 days — JAMA Network Open, Oct 2025 (Olson et al.) 8.5% less EHR time, 15%+ less note-composition time — Advisory.com / UChicago Medicine JAMA study, Feb 2026 Aidoc CARE1: first FDA-cleared foundation model, Feb 2025 — IntuitionLabs, Mar 2026 Carnegie Mellon/UPMC $10M CPACE center targeting underserved cancer screening — IntuitionLabs, Nov 2025 EU AI Act: medical AI classified high-risk, mandatory human oversight — EU AI Act 2024 Congress:

Transcript

00:00Here's everything AI is doing inside medicine in two minutes.

00:05Diagnostics first. Deep learning models trained on millions of medical images

00:10can now flag breast cancer on a mammogram, detect diabetic retinopathy from a retinal

00:16photo, and spot stroke on a CT scan, often before a human has opened the file.

00:22The FDA authorized over 1,200 AI medical devices as of August 2025, with radiology accounting

00:31for more than three quarters. These models improve radiologists' diagnostic sensitivity

00:37by 6-8% when they work together. AI making the human better, not replacing them.

00:44Next, drug discovery. In 2025, a drug with both its target and molecule designed entirely by AI

00:53completed Phase 2A clinical trials and showed clinical improvement, published in Nature Medicine.

01:00Traditional drug development takes over a decade and $2 billion per drug, with a 90% failure rate.

01:08AI-assisted pipelines are compressing that to three to six years.

01:13Self-driving robotic labs run experiments 24-7. AlphaFold has unlocked thousands of previously

01:22undruggable disease targets. Then-admin, the invisible crisis. Physicians spend more than five

01:29hours on electronic health records for every eight hours with patients. Ambient AI scribes listen to

01:36consultations and write the clinical note. A 2025 JAMA study tracked six health systems over

01:4330 days. Physician burnout dropped from 51 to 38%. Doctors felt present with patients again and

01:52personalized care. Risk prediction before symptoms appear. Genomic matching of your tumor to the right

01:59therapy. Treatment optimized for your specific biology. But the risks here are the highest of any

02:06industry we've covered. Biased training data performs worse on minorities. Scribes can hallucinate drug dosages.

02:15And a wrong diagnosis can cost a life. The guardrail. The clinician stays in the loop. Always.

02:21That's AI in healthcare. Full episode. Link in the bio.

02:26That's AI in healthcare. Full episode. Link in the bio.

How AI Is Changing Medicine — Drugs, Diagnosis & the Risk

Category

Transcript

Comments

Recommended