Brockman’s AI medical forecast faces a clinical evidence test | Magica
Brockman forecasts AI medical gains in 12–18 months, with longevity still unproven
Editorial Team
••📖6 min read
OpenAI president Greg Brockman predicts significant medical advances within 12 to 18 months and raises the possibility of death becoming a choice. OpenAI’s cell research and a separate early drug trial show narrower progress, with financial ties, safety questions and clinical testing still central to judging the promise.
OpenAI’s president predicts significant AI-driven medical advances within 12 to 18 months; his suggestion about death becoming optional has no comparable deadline.
OpenAI reports cell-reprogramming gains with longevity startup Retro Biosciences, which has financial ties to OpenAI’s chief executive. The model is not broadly available.
A separate AI-discovered drug reached a controlled human trial, but its early lung-function signal came with safety concerns and substantial missing measurements.
Greg Brockman, president and co-founder of AI developer OpenAI and its founding chief technology officer, predicts significant advances in biology, health and disease treatment over the next 12 to 18 months. At WSJ Tech, he also raised the possibility of death becoming a choice, according to a November 5 account of his remarks.
Brockman helped build OpenAI from its start: its December 2015 founding announcement named him CTO, previously CTO of Stripe. His forecast comes from a technology executive whose company is already working on longevity research. That gives the prediction a concrete research context, while making OpenAI an interested party in assessing AI’s medical potential.
The crucial distinction is between accelerating research, helping patients and overcoming mortality. Existing experiments and clinical results offer ways to judge the first two. They do not establish the third.
Cells treated with RetroSOX, RetroKLF, OCT4 and MYC at day 10 in company-reported laboratory research. Credit: OpenAI/Retro Biosciences. Source: OpenAI life sciences research.
A near-term forecast without a defined clinical milestone
Brockman acknowledged that promises of disease cures have become familiar in AI, then predicted substantial progress within a year to a year and a half. A second account of the appearance quoted him discussing death becoming a choice:
The Cambridge startup has secured backing described as eight figures in capital and compute, including UK government support. Its wager is that cheaper, accurate quantum calculations can supply better AI training data; full performance results remain forthcoming.
Editorial Team
“All of this is possible.”
That expresses possibility, not a demonstrated intervention. The reported passages identify no particular disease, patient population, treatment or clinical endpoint for the 12–18-month forecast. Nor do they put a deadline on making death optional.
A new biological insight, a promising drug candidate and a treatment that improves patients’ lives would each be consequential. They represent different stages of progress, so the forecast needs a defined milestone before it can be assessed meaningfully.
OpenAI’s longevity work has results—and financial ties
In an August 2025 research announcement, OpenAI described working with longevity biotech startup Retro Biosciences on GPT-4b micro, a small experimental model derived from GPT-4o and specialized for protein engineering. Retro supplied cell-reprogramming expertise and laboratory screening; OpenAI supplied the model work.
The target was an existing biological technique. Proteins known as Yamanaka factors can turn adult cells into pluripotent stem cells, which can develop into different cell types. OpenAI said conventional reprogramming typically converts fewer than 0.1% of cells and can take three weeks or more, with still poorer efficiency in cells from aged or diseased donors.
OpenAI reported that redesigned proteins produced more than 50 times the expression of stem-cell reprogramming markers compared with unmodified controls in laboratory experiments. It said the findings were replicated across donors, cell types and delivery methods, and that resulting stem-cell lines showed pluripotency and genomic stability. It also reported reduced DNA-damage markers in a separate cell assay.
These are measures of cellular behavior. The 50-fold figure does not measure lifespan, and the experiments do not demonstrate rejuvenation of a person. They also depended on scientists screening and validating model-generated proteins in a laboratory. OpenAI itself cautioned that computational protein benchmarks often have limited value because better scores may not translate into practical utility.
The collaboration carries a financial connection. OpenAI disclosed that Sam Altman, its chief executive, invests in Retro and said its research team formed the partnership. Altman supplied Retro’s entire $180 million seed round, according to a January 2025 report.
OpenAI also said GPT-4b micro was developed for research and was not broadly available. Its announcement establishes neither a generally deployable product nor the cost of delivering a therapy based on the work.
Company-reported alkaline phosphatase staining at day 10 compares standard OSKM controls with engineered RetroSOX/KLF variants. Purple colonies indicate activity associated with stem-cell reprogramming. Credit: OpenAI/Retro Biosciences. Source: OpenAI life sciences research.
A human drug trial offers a different test
A separate example comes from Insilico Medicine, the drug developer that sponsored a randomized phase 2a study published in June 2025. Its experimental drug rentosertib inhibits TNIK, a protein implicated in fibrosis; the researchers used generative AI to identify the target and design the compound. The team reported reaching preclinical candidate nomination in 18 months and completing phase 0/1 testing in under 30 months of starting target discovery—milestones short of proving a treatment works in patients.
The subsequent phase 2a trial enrolled 71 patients in China with idiopathic pulmonary fibrosis, a progressive lung disease, and compared three dosing regimens with placebo over 12 weeks. Its primary endpoint was safety, measured by the share of patients experiencing at least one treatment-emergent adverse event. Lung function was a secondary endpoint.
The strongest lung-function signal came from the 60-milligram once-daily group. Its mean change in forced vital capacity—the volume of air forcibly exhaled after a full breath—was an increase of 98.4 milliliters, with a 95% confidence interval of 10.9 to 185.9 milliliters. The placebo group’s mean change was a decline of 20.3 milliliters, with an interval from a 116.1-milliliter decline to a 75.6-milliliter increase. Those groups contained only 18 and 17 patients, respectively; the authors described the result as a trend warranting further study.
The safety and missing-data findings matter alongside that signal. Sixteen patients discontinued treatment across the trial. Seven stopped because of liver injury or dysfunction; four of those seven were also taking nintedanib, an existing antifibrotic treatment. Three patients in the 60-milligram once-daily group experienced acute disease exacerbations requiring hospitalization, compared with one exacerbation without hospitalization in the placebo group. These small counts do not establish causation, but the authors called for longer studies to assess the risk.
Separately, 22 patients lacked week-12 lung-function measurements. The analysis filled missing values statistically under an assumption that they were missing at random. The short follow-up and geographically homogeneous population further limited conclusions about durable benefit and safety. Several authors were Insilico employees, and sponsor-employed authors participated in study design, analysis and interpretation.
This was a test against placebo, with continued standard medication permitted—not a head-to-head demonstration of superiority over existing treatments such as nintedanib or pirfenidone. It provides evidence that AI-assisted discovery can reach controlled human testing, while leaving the treatment’s longer-term value unresolved.
Week-12 spirometry was missing for 22 trial participants. Missing values were imputed assuming they were missing at random. These counts describe missing measurements rather than treatment discontinuations. Source: Nature Medicine.
The next test is sustained patient benefit
Brockman’s forecast needs disease-specific commitments: which interventions, which patients and which outcomes he expects within 12 to 18 months. For cell reprogramming, the unresolved question is whether laboratory gains can become safe, effective therapies. For rentosertib, the study authors called for larger, longer and more diverse trials, including assessment of interactions with existing drugs.
Those studies can test whether AI helps produce durable medical benefits. Even success against a particular disease would leave the broader claim about optional death unanswered: neither research program reported here demonstrates an extension of human lifespan.
Japan’s Nikkei fell 7.09% in Friday morning trading, with AI, chip and data center-related shares hit hard after Wall Street losses. Reports linked the rout to fears of U.S. regulation following Democratic election victories, without identifying an enacted restriction or a cut in AI spending.
Potential investors are pressing Anthropic for a lower valuation as a technology selloff threatens its November IPO timetable. Rapid revenue growth and substantial cash complicate a financing debate driven by losses and long-term infrastructure commitments.
Visa's new benchmark finds that AI shopping models often miss customer requirements and struggle to repeat successful decisions. The tests cover product discovery, leaving checkout and payment reliability for future rounds.
Agility Robotics began trading after its Churchill Capital Corp XI merger, bringing a commercially deployed humanoid robot maker to U.S. public markets. Its next test is turning conditional Digit 5 orders into deliveries and sustainable economics.
Goldman Sachs sees temporary halts potentially affecting about 45% of planned US data center expansion after Democratic election victories. In Texas, where approvals were already paused, the next fight is over how long restrictions last and what developers must pay for.
Core Scientific’s third-quarter colocation revenue rose 57.6% to $215.4 million. Its expansion beyond CoreWeave could diversify sales from 2027, but the next projects require additional financing rather than the same customer funding that supported its existing conversions.
Meta is rolling out voice conversations and animated avatars for Muse in the US. The update adds personality to a cloud-based task agent whose costs, permissions and reliability matter more than its expressions.
Anthropic is rolling out Personal Finance to paid Claude users in the US, using Plaid to read bank data. Its pitch rests on ad-free answers and privacy controls, while its own documentation warns that balances and budgets can be incomplete.
Citrini Research is recasting its AI investments around relative value as it warns that financing and construction constraints threaten returns. Its reported Oracle–Nebius trade faces a crucial counterargument: Nebius has substantial customer commitments, although delivery remains conditional.
Shield AI has agreed to acquire Echodyne for up to $5.3 billion, backed by a $2.5 billion financing package. The radar maker brings a profitable, defense-heavy business and commercial customers, with a promise to keep supplying outside system builders.
Gemini Omni 1.2 Flash lowers Google's quoted 720p generation price to $0.07 per second and extends native clips from 10 to 15 seconds. Editing limits and early commercial tests show why cheaper output does not guarantee the same savings on finished videos.
Mercedes-Benz plans to add Liquid AI’s on-device models to its North American voice assistant. The deal combines local processing with existing cloud AI, while customer availability, measured performance and contract economics remain unconfirmed.
Mistral is expanding Studio with managed agents, apps and workflow tools, extending its existing enterprise platform. Its private-deployment pitch sits alongside cloud-hosted app services and vendor-stored agent sessions, making the chosen configuration central to customer control.
Meta president Dina Powell McCormick expects tougher conditions for new US data centers after Democratic election victories, but says states still want investment. The test is whether electricity commitments, local jobs and tax benefits can secure community support.
OpenAI is releasing GPT-6 Cyber to selected verified defenders with stronger monitoring. Its Linux research shows the scale and cost of automated exploitation, while mixed benchmark results and unfinished patching limit what the launch proves about safer software.
Black Hills plans $1.8 billion of gas generation under agreements to serve Google’s proposed $17 billion Cheyenne campus through 2048. Google has advanced equipment funding, but delivery, fuel-cell conditions and local impacts remain unresolved.
Dropbox agreed to buy Tigris for about $350 million in cash and unveiled DBX Agent Cloud, starting with standalone object storage. The expansion accompanies AI actions for existing file customers, while pricing and evidence of competitive performance and productivity gains remain absent from the announcement.
ElevenMusic creators can now publish directly to SoundCloud, with a three-month Pro offer for Artist Pro subscribers. The integration shortens distribution while preserving disclosure and rights requirements.
A researcher demonstrated how a browser extension could use Claude Dispatch to execute code through a connected desktop agent. Anthropic reportedly acknowledged the flaw in May and proposed retiring Dispatch, but existing users still have access, leaving questions about authorization across devices.
