Why AI Hasn't Cured Anything...Yet, According to Jennifer Doudna

ELI5/TLDR

Jennifer Doudna won a Nobel for co-inventing CRISPR, a tool that lets us cut and edit DNA the way you’d fix a typo in a document. It has cured a handful of people of genetic diseases, but it’s still slow, brutally expensive, and only works for a few conditions. The big new hype is that AI will speed all of this up and “cure cancer” — and Doudna, very politely, says she sees no sign of that happening yet. Her message: the breakthrough was real, biology is just much harder than the headlines admit, and right now the riskiest thing isn’t failure, it’s the funding cuts pushing science out of the US.

The Full Story

What CRISPR actually is

Start with a problem bacteria have: viruses attack them too. Over millions of years, bacteria evolved a memory-based immune system, and CRISPR is the leftover machinery from it.

Here’s the trick. When a virus infects a bacterium, the bacterium snips out a small piece of the invader’s DNA and files it away — like keeping the mugshot of a burglar. It then copies that mugshot into a short strand of RNA, which acts as a kind of genetic GPS coordinate. Paired with that GPS is a protein called Cas9, which is essentially a pair of molecular scissors. If the same virus ever comes back, the GPS strand guides the scissors straight to the matching DNA and cuts it. Threat neutralised.

What Doudna and her collaborator Emmanuelle Charpentier realised — the insight that won the Nobel — was that the GPS strand could be rewritten. Instead of only pointing at viral DNA, you could feed it any address you wanted, including addresses inside human cells. Suddenly the scissors would go wherever you sent them.

“Suddenly editing DNA was no longer theoretical.”

That’s the whole revolution in one line. A defence system bacteria use against viruses became a programmable text editor for the code of life.

It has cured real people — a few of them

This isn’t a promise; some patients are already living it. Doudna keeps returning to Victoria Gray, the first American with sickle cell disease treated with CRISPR. And in 2025, an infant called “baby KJ” became the first person ever to get a CRISPR therapy designed just for him — a one-off edit for a rare disorder where his body couldn’t break down protein and toxic ammonia was building up in his brain. He’s now walking and climbing on everything.

So why isn’t everyone cured? Two reasons, and they’re the heart of the video.

The first is cost and clumsiness. Baby KJ’s treatment ran about $800,000, stitched together from research grants, university labs, and charity. And the way most CRISPR therapy works today is genuinely unpleasant: doctors take cells out of the patient, edit them in a dish, then transplant them back in.

“It’s quite expensive. It’s unpleasant for patients… it’s not ideal.”

The holy grail Doudna wants is in-vivo editing — delivering the molecular scissors directly into the body, having them travel to exactly the right cells, make the edit, and then stop. Get there, and the cost and misery collapse.

The second reason is humbling: we still don’t understand the machine we’re editing. Doudna offers a number that should make anyone pause. We sequenced the human genome around the year 2000. Twenty-six years later, we still don’t know what roughly 40% of the genes do in a bacterial cell — and bacteria are far simpler than us.

“We’re not going to be able to simulate our way to an understanding of the human body.”

The AI question

This is the part the title is built around. Silicon Valley is making enormous promises — one OpenAI clip says “one day, maybe we can cure all disease with the help of AI,” and Larry Ellison has claimed AI will cure cancer in a 48-hour window.

Doudna’s response is the calm centre of the whole interview. She’s not anti-AI. She thinks it’ll genuinely help — summarising data, writing reports, spotting how drugs interact with the body, making discovery more efficient. But she draws a hard line at the word innovate.

“I’m not seeing chatbots in our own experience innovating… I’m not seeing chatbots coming up with a brand new idea for something that nobody else ever thought of.”

Asked whether AI can’t innovate, she’s careful: “I don’t know if it can’t innovate. I just don’t think it is right now.” Asked about the post-AGI future, she shrugs: “I never say never… but I’m not holding my breath.” And when told an OpenAI executive thinks the company should get a cut of any drug discovered via ChatGPT, her entire answer is two words: “Good luck.”

Her real bottleneck isn’t intelligence, it’s data. Better models need better and more training data about how bodies actually work — and you can’t shortcut your way past physical experiments.

The part that actually worries her

The fear in this video isn’t rogue AI or evil scientists. It’s politics and money.

US federal funding has been cut hard. New NSF grants dropped 24% over three years; about 20% of the federal research workforce — some 25,000 people — has left. Doudna frames it as economic self-sabotage: every NIH research dollar returns about $2.50 in economic benefit, and pulling back simply hands the lead to China, which is pouring state money into biotech and fast-tracking human trials.

She’s equally blunt about the anti-science mood — the “MAHA” movement, anti-vaccine sentiment, a measles outbreak. Coming from someone whose RNA work underpinned the COVID vaccine, the verdict lands hard:

“It’s a dangerous moment. It’s very dangerous.”

Designer babies, carefully

The interview circles the scary edge: CRISPR-edited babies. A Chinese scientist already did this in 2018, to global condemnation. Doudna predicts edited babies within 25 years, but pours cold water on the Silicon Valley dream of selecting embryos for intelligence or height. Those traits are governed by thousands of interacting genes, and tweaking one can quietly break others.

“If we tweak one gene or even a few, we can’t always be sure what the outcome is gonna be.”

Her proposed compass is to sort uses into buckets: editing out a devastating disease with a single, well-understood genetic cause is defensible; reaching for eye colour or “musculature” is a different moral category entirely. Asked if the future is Gattaca or paradise, she lands on “somewhere in the middle… hopefully closer to Nirvana.”

Key Takeaways

CRISPR is repurposed bacterial immunity. Bacteria store fragments of past viral attackers as “guide RNA” and use a protein (Cas9) as scissors to cut returning viruses. Doudna’s insight was that the guide RNA can be reprogrammed to target any DNA sequence, including in human cells.
A handful of real cures exist. Victoria Gray (sickle cell, first US CRISPR patient) and “baby KJ” (2025, first fully personalized CRISPR therapy, for a urea cycle disorder) are living proof — not promises.
The current method is crude: cells are removed from the patient, edited in a lab, and transplanted back. The goal is in-vivo editing — delivering CRISPR directly into the body to the right cells.
Cost is the wall. Baby KJ’s bespoke therapy cost ~$800,000, funded by a patchwork of grants, academia, and philanthropy. Scaling means driving cost down and standardising the pipeline.
We barely understand the genome. ~26 years after sequencing the human genome, we still don’t know the function of ~40% of genes in a bacterial cell — a measure of how far we are from “simulating” human biology.
Doudna’s AI line: helpful, not creative. AI can summarise data, write reports, and model drug interactions, but in her lab’s experience it does not yet innovate — it doesn’t generate genuinely new ideas. The real bottleneck is good training data, not intelligence.
IGI scale: Doudna’s Innovative Genomics Institute has spun out 31 companies, ~$9B combined valuation, ~2,500 jobs.
Funding is the bigger threat than the tech failing. US: ~24% drop in new NSF grants over 3 years, ~20% of the federal research workforce (~25,000 people) gone. Every NIH research dollar returns ~$2.50 economically. China is investing heavily and streamlining human trials — the lead could shift.
Designer babies are overhyped near-term. Traits like intelligence and height involve thousands of interacting genes; editing for them risks unpredictable side effects. Single-cause disease edits are far more tractable and defensible.
The Larry Ellison “cure cancer in 48 hours” and OpenAI “cut of drug sales” claims: Doudna’s reaction is polite skepticism (“I’d be overjoyed… but I don’t see it right now”) and, on the revenue-share idea, just “Good luck.”

Claude’s Take

This is a well-produced Bloomberg profile, and its real value is using a genuine authority to deflate a specific balloon: the Silicon Valley story that AI is about to brute-force biology into curing everything. Doudna is the ideal person to do it — she has a Nobel, real cured patients, and zero incentive to badmouth a technology her own institute is trying to harness. When she says “I’m not seeing chatbots innovating,” that carries weight a pundit’s hot take wouldn’t.

The honest framing — breakthrough was real, biology is just genuinely hard, the gap between a lab discovery and a delivered cure is enormous — is the most useful thing here. The 40%-of-bacterial-genes-still-mysterious stat is the kind of fact that recalibrates your priors instantly.

Where I’d dock points: it’s an interview, not an investigation. The hard questions get gestured at but not pressed — the $800k price tag, who actually pays, why in-vivo delivery is so stubborn, all stay at altitude. The science-funding and anti-vax segments are real but lean editorial, and Doudna is, after all, a stakeholder in “fund more science.” Worth noting but not damning. A 7: substantive, honest, genuinely clarifying on the AI-hype question, but it’s a polished profile rather than a deep dive, and a curious viewer will finish wanting the follow-up questions the format didn’t allow.

Why AI Hasn't Cured Anything...Yet, According to Jennifer Doudna | The Circuit