Gates Foundation
Institution - official statements with no named speaker · USA
Track record: No forecast checked yet (5 not due, 0 being checked).
5forecasts & commitments on the record
60%can be checked as stated
44average ambiguity (0-100)
0goalpost moves recorded
0confirmed U-turns
Timeline
came true partly did not not due / being checked / too vague
Calibration score: shown after 10 checked forecasts with a stated likelihood (now 0).
How they speak about the future: based on data 4, stated on authority 1.
Kinds of claim: forecast 5. · Download this report card as a picture
In the nine words
How to read this
Each box is one sentence a public figure or institution said about the future, word for word, with a link to the official text. Can be checked means it names what will happen, by when, and how likely - so it can be compared with official figures later. Too vague to check means it does not. Ambiguity (0-100) is how many different ways the sentence can be read. When the date passes, the forecast is compared with official data and marked came true, partly or did not come true. Labels judge the sentence, never the person. Full guide and glossary.On the record (5)
The program expects to identify an initial package of innovations by early 2027 and announce its first participating country partner in the coming months.
The multi-year roadmap will scale AI applications from an initial reach of 50 million farmers to 200 million smallholders across critical climate-adaptation and food-security challenges.
MOUNTAIN VIEW & SEATTLE (Friday, 18 September) — The Gates Foundation and Google today unveiled an expanded initiative to bring advanced AI tools to 200 million smallholder farmers across Sub-Saharan Africa and South Asia, helping them anticipate climate risks, improve farm productivity, and build resilience in a changing climate.
Today’s commitment will help scale up the work of those partners over the next two years, with the funding distributed roughly as follows across the foundation’s priority areas:
More than 90% of the data used to train early large language models came from English-language sources, leaving many communities that could benefit from AI poorly represented in the data on which these tools were built.