AI is the interface. Curated information is the foundation.
Language models are powerful at interpreting questions, but generated prose alone is not a substitute for traceable scientific evidence. ChemQuery uses AI to access structured datasets so researchers can examine and verify returned records.
Limitations of transformer based large language models (LLMs)
Tell me what you know, not what you think
General-purpose LLM responses
Useful for synthesis, explanation, and ideation. Specific factual claims may need independent validation and may not expose their underlying structured evidence.
Data-grounded scientific workflows
Translate questions into queries over curated records, preserve identifiers and measurements, and make results easier to inspect. Validation remains important; grounding does not guarantee error-free answers.
Gemini, Claude, and ChatGPT etc. are transformer-based large language models
- Rely on expensive statistical modeling of very large sets of text chunks
- Transform natural language queries into statistically most likely answers
- High confidence for frequent relationships
- Low confidence for rare relationships
- Hallucination possible
Authoritative Intelligence is a vetted-info-based large language model
- Relies on large sets of authoritative facts in validated databases
- Translates natural language queries into complex expert SQL queries
- Explicit statistics for frequent relationships
- High precision - even for rare relationships
- Answers trace back to records you can inspect
We create Authoritative AI applications, such as ChemQuery, integrating chemistry with other disciplines.
Natural language on the front end. Validated data and explicit queries underneath.
Natural-language prompt
Ask a scientific or business question in plain English.
ChemQuery LLM / AI
Generates complex expert SQL against trained data structures.
Authoritative result
Returns traceable data, counts, tables, and structured outputs.