Google’s AI wants to remove every disease: what’s actually happening?
A Reddit post making the rounds claims that Google’s Isomorphic Labs (the DeepMind-adjacent drug discovery outfit) is moving AI-designed drugs into human clinical trials, with an ambition to “wipe out all diseases”. The source cited is Fortune plus the Wikipedia page for Isomorphic Labs.
“They’re about to start clinical trials with drugs made by an AI.”
Big claim, big implications. If even partly true, it could reshape how medicines are discovered and who benefits. Below is a plain-English look at what this means, where the hype ends, and why UK readers should care.
Sources from the Reddit post: Fortune, Wikipedia: Isomorphic Labs, and the discussion thread on Reddit.
What “AI-designed drugs” actually means
Drug discovery is a multi-step process that starts with finding a biological target (often a protein) and ends, years later, with a safe, effective medicine. AI can accelerate early steps by running large-scale “in silico” (computer-based) experiments. In practice, this includes:
- Target discovery – spotting proteins or pathways linked to disease.
- De novo design – proposing new molecules with desired properties.
- Property prediction – estimating safety, absorption, and toxicity before any lab work.
- Prioritisation – ranking candidates to test in wet labs, saving time and cost.
AI doesn’t remove the need for lab experiments or clinical trials. It narrows the search space and reduces dead ends. That’s valuable, but it’s not a magic wand.
Clinical trials: the slow, expensive part remains
The Reddit post suggests human trials are imminent. Key details are not disclosed: the indication (which disease), the phase (I/II/III), which regulators are involved, or timelines. These matter because:
- Phase I tests safety in healthy volunteers or patients (small numbers).
- Phase II tests effectiveness and dosing in patients (larger sample).
- Phase III confirms efficacy and safety at scale before approval.
Even with AI, trials take years, cost serious money, and face high failure rates. Regulators like the MHRA (UK) or FDA (US) evaluate evidence from these trials. No-one gets a pass because a model suggested the molecule.
Will this “cook” pharma—or transform it?
“If this even half works, pharma as we know it is kinda cooked.”
Expect transformation more than destruction. Pharma’s core advantages—clinical development, regulatory strategy, manufacturing, distribution, and pharmacovigilance—don’t disappear. Instead, AI shifts where value is created:
- Discovery becomes faster and cheaper, with more shots on goal.
- Smaller, AI-native firms may partner earlier with big pharma for trials and commercialisation.
- Competition could intensify for certain targets and therapeutic areas.
Where AI helps—and where it doesn’t
| Stage | Traditional bottleneck | AI’s potential contribution | Still hard/regulatory |
|---|---|---|---|
| Target discovery | Finding causative biology | Pattern-finding across omics and literature | Biological validation required |
| Hit/lead design | Slow iterations in chemistry | De novo design, property prediction | Synthesis feasibility, novel chemistries |
| Preclinical | Animal and toxicity studies | Toxicity prediction, prioritisation | Must run real studies |
| Clinical trials | Recruitment, cost, attrition | Better patient stratification, trial design | Years of evidence and oversight |
“Wipe out all diseases”: ambition vs biology
Ambition is healthy. Total eradication is another matter. Some realities:
- Complex diseases (e.g., Alzheimer’s, many cancers) are multi-factorial and evolve over time.
- Pathogens (e.g., bacteria, viruses) adapt—antimicrobial resistance is a moving target.
- Safety margins are unforgiving; off-target effects can kill promising programmes.
AI can help us find and test more ideas. It cannot abolish uncertainty, diversity in human biology, or the need for robust evidence.
Why this matters for the UK
The UK is well-placed if AI-driven discovery takes off:
- Regulation: The MHRA has been open to innovative trial designs and digital tools. Expect scrutiny on model validation, data provenance, and safety monitoring.
- NHS and access: For UK patients, the win is faster access to effective treatments at fair prices. NICE’s cost-effectiveness bar still applies.
- Data and privacy: Any use of patient data must comply with GDPR and UK data protection law. Transparency around datasets and consent is critical.
- Ecosystem: The Oxford-Cambridge-London triangle has deep biotech and AI expertise. Partnerships, jobs, and spinouts could grow if these platforms deliver.
Reading the headlines critically
When you see claims like “AI drug enters human trials” or “cure all diseases”, ask:
- Which disease? Which patient population? Not disclosed.
- Which trial phase and geography? Not disclosed.
- What evidence supports the candidate—preclinical data, safety signals, biomarkers? Not disclosed.
- How is AI used—target ID, molecule design, trial optimisation? Not clearly stated.
- What is the company’s plan for manufacturing, regulatory approval, and pricing? Not disclosed.
Lack of detail doesn’t mean it isn’t real—it means the proof will come from trial registries, peer-reviewed papers, and regulatory filings rather than press lines.
Balanced outlook: excitement with guardrails
Reasons to be optimistic:
- AI accelerates hypothesis generation and reduces wasted lab work.
- Better patient stratification could make trials smaller, faster, and more likely to succeed.
- Rare diseases might benefit from lower discovery costs and targeted designs.
Risks and trade-offs:
- Data quality and bias can mislead models and patients.
- Models can “hallucinate” plausible but wrong structures or relationships if not rigorously validated.
- Overpromising erodes public trust and can distort funding away from essential, non-glamorous science.
For developers and professionals: what to do now
You don’t need a wet lab to gain value from AI. Focus on robust data practices, clear validation, and auditable workflows—principles that apply from life sciences to finance and ops. If you’re integrating AI into day-to-day work, start with high-leverage automations that respect compliance boundaries.
Related read: if you’re experimenting with light-touch automation, here’s a practical guide to connecting LLMs with spreadsheets for structured workflows: How to connect ChatGPT and Google Sheets.
Bottom line
Isomorphic Labs moving into human trials—if confirmed—would be a meaningful milestone for AI in drug discovery. But “remove every disease from Earth” is marketing shorthand, not a plan. For UK readers, the prize is faster, safer medicines reaching the NHS, stewarded by sensible regulation and fair pricing. Watch for specifics—indication, phase, evidence—and judge the science, not the slogan.
