Probably raises $9M to build more reliable AI with error-reducing tech

As large language models (LLMs) have grown more powerful, errors have proven difficult to avoid. Probably, which just raised $9 million in seed funding from Andreessen Horowitz, is trying to build a more rigorous way to catch those errors. Probably's goal is to prevent hallucinations and simple factual errors from ever reaching the user, and achieve 99.99% accuracy.

Founder Peter Elias says bringing LLMs to that level of accuracy requires rethinking many of the basic assumptions of AI engineering. Probably's first product is a data science tool that produces quick answers from complex datasets, each with a citation and audit trail. The tool uses an elaborate harness system, described as a "data science mech suit," to keep errors from creeping into summaries.

The LLM's first-pass answers are checked against a deterministic validator system, which bounces back any results that don't match the dataset. "What we learned building this was that the better your harness engineering is, the weaker the model can be," Elias says. "If you can refine the context enough, the model does not have to work very hard to do the right thing.

Basically, it's an exercise in reducing ambiguity." This allows Probably's data science tool to run on significantly smaller AI models, which can be run on local hardware, reducing token costs associated with AI use. The company's approach could have far-reaching implications for industries that require high accuracy, such as accounting and medical services. Elias notes that the same engine can be extended to cover use cases like these, which he describes as "any precision-sensitive use case." "I think it's really interesting that the big AI labs have not even attempted to do this," Elias says.

"They're incentivized not to, because they make money the more times you have to correct the model." Why this matters: The AI industry's struggle with hallucinations and factual errors has significant implications for businesses and consumers. Probably's approach, which prioritizes accuracy and transparency, could set a new standard for AI development. By reducing token costs and enabling smaller AI models, Probably's technology could make high-accuracy AI more accessible to a wider range of users.

However, it's unclear whether the company's approach can be scaled up to meet the demands of more complex use cases. As AI continues to play a larger role in industries like healthcare and finance, the need for reliable and accurate AI systems will only continue to grow. With $9 million in seed funding, Probably is poised to make a meaningful impact on the AI landscape, but the company's true test will come as it works to implement its technology in real-world applications.