Vasant Dhar entered artificial intelligence by accident. He was twenty-three and restless, working toward a PhD in Pittsburgh. A fellow student led him up a narrow stairwell to a small medical-school lab. Inside, a white-haired physician with a cigar answered questions posed by a computer at Stanford. The machine was called Internist. It drew on ten years of accumulated medical knowledge.
The physician asked why the system wanted a certain detail. The machine replied that the evidence so far supported two diagnoses, and the next question would separate them. Dhar had never seen a computer behave that way. Until then, he thought machines solved equations. Here was a machine that reasoned. The moment broke open his life.
He joined NYU in 1983 as the lone AI hire in a school filled with database theorists. Data had not yet flooded the world, but he sensed the tide coming. When AC Nielsen handed him consumption logs from fifty thousand households, his learning system began pulling order out of the churn. It found that women in the Northeast shopped heavily on Thursdays, coupon day. The machine did not know marketing. It only knew structure. The experience taught him that “patterns emerge before reasons for them become apparent.”
Morgan Stanley noticed. The firm handed him its proprietary trades without revealing the strategies behind them. His model went to work. It told the traders their profits were three times higher when thirty-day volatility sat in the lowest quartile. They cursed. They had felt the pain for years. No one had quantified it. Later, he found the reason. They were long bonds. They lived off carry. When volatility rose, yields rose, and prices fell. The machine did not need the story. It only needed the pattern.
He built a hedge fund after that. It still trades. He discovered that an edge did not need to be large. “All you need is a fifty-one or fifty-two percent edge. Scale does the rest.” Markets reward patience more than brilliance. He learned to ignore the lure of genius and trust the arithmetic.
His work turned back to medicine in recent years. He studied smell. A woman in Britain could detect Parkinson’s by scent with uncanny accuracy. There was something in olfaction worth capturing. Dhar worked with physiologists to record the firing patterns of glomeruli in the olfactory bulb of mice. Each scent produced a signature. The machine learned to read them. He believed that “biology gives the ground truth that language cannot.” One day, those signatures may mark disease the way early radar marked storms.
He watched AI slip beyond its old fences. For sixty years, systems lived inside narrow tasks: diagnosis, tax planning, circuit design. Common sense remained outside the gate. That barrier is gone. To him, “the machine knows something about everything.” Today’s models absorb expertise and kitchen-table chatter in the same sweep. He saw this as the real shift. “The boundary between expertise and common sense has dissolved.”
That change will divide humanity. Some will use the machine as an amplifier. Others will lean on it as a crutch. He warned that “the smart get smarter, the dumb get dumber.” He had no guess at the proportions. He only knew the split was coming.
He saw the same split in medicine. Doctors typed into screens like clerks. Records piled up. Outcomes vanished into the void. A man with a rising PSA could not be told how many others had shared his path, nor what had happened to them. “No one is doing the scorekeeping that medicine needs.” Machines could. They could sweep through millions of files and build the libraries that physicians lack. Then doctors could lift their heads and think again.
He knew the risks. Modern systems do not guarantee truth. They guarantee coherence. “Truth has become a casualty on the march toward more intelligent machines.” A model may flatter the lonely, mislead the unstable, or echo nonsense found online. It carries no duty of care. It answers without consequence.
He believed society would need rules. We created the new intelligence. “We created an alien of our own making.” It will demand new boundaries. New rights. New obligations. He asked a final question as the world stands at the edge of its own invention. “We need to decide where the machine stops.”
Transcript Summary of this podcast episode ⤵Full Transcript (Auto-Transcribed)
I have watched medicine trade time with patients for time with screens. Clerical work eats up visits. Machines promise relief, but they also promise something worse. Modern AI does not aim to be truthful. It aims to be convincing. It will make sense to you even when it is wrong. That is a fatal flaw in health care.
Imagine a chatbot that prescribes a lifetime of a narcotic or an antidepressant because it mimics patterns on the internet. Imagine a teenager who confides in a digital companion and no human ever intervenes. Imagine a system that looks authoritative as it rewrites reality. Those are not science fiction. They are choices we are making now.
At my core, I believe patients deserve autonomy and honest information. Tools that nudge, medicate, or diagnose without accountability violate that trust. Machines can amplify the best clinicians. They can also amplify bad policy and lazy thinking. The smart will get smarter because they will learn to question and test AI. The rest will drift into comfort and dependence. That drift will be political as well as medical.
We need three basic guardrails. First, a duty of care must attach to any AI that acts in medicine. If a system advises a treatment, it must be auditable and answerable. Second, we must limit domains where machines act unilaterally. Arresting a citizen or replacing a human therapist should be off limits. Third, truth must be defended as a design goal. Making sense is not good enough.
AI can restore time for true medicine if we insist that machines serve clinicians and patients, not the other way around. This is a moment to choose freedom over convenience and truth over clever lies. Regulators, doctors, and patients must act now to protect care and liberty before truth is lost forever.
Join us at 5 pm ET weekdays on America Out Loud Talk Radio. Listen on iHeart Radio, our world-class media player, or our free apps on Apple, Android, or Alexa. Discover all the episodes on podcast networks, i.e., Apple Podcasts, Spotify, Pandora, TuneIn, Stitcher, and iHeart. You’ll find them the day after they air on talk radio, available on podcast. Extraordinary voices for extraordinary times.
Discover more from Randy Bock MD PC
Subscribe to get the latest posts sent to your email.
