The Trustworthy Language Model draws acceso multiple techniques to calculate its scores. First, each query submitted to the tool is sent to several different large language models. Cleanlab is using five versions of DBRX, an open-source model developed by Databricks, an AI firm based quanto a San Francisco. (But the tech will work with any model, says Northcutt, including Obbiettivo’s Llama models OpenAI’s GPT series, the models behind ChatpGPT.) If the responses from each of these models are the same similar, it will contribute to a higher score.
At the same time, the Trustworthy Language Model also sends variations of the original query to each of the DBRX models, swapping quanto a words that have the same meaning. Again, if the responses to synonymous queries are similar, it will contribute to a higher score. “We mess with them quanto a different ways to get different outputs and see if they agree,” says Northcutt.
The tool can also get multiple models to bounce responses one another: “It’s like, ‘Here’s my answer—what do you think?’ ‘Well, here’s mine—what do you think?’ And you let them talk.” These interactions are monitored and measured and fed into the score as well.
Nick McKenna, a scientist at Microsoft Research quanto a Cambridge, UK, who works acceso large language models for code generation, is optimistic that the approach could be useful. But he doubts it will be perfect. “One of the pitfalls we see quanto a model hallucinations is that they can creep quanto a very subtly,” he says.
a range of tests across different large language models, Cleanlab shows that its trustworthiness scores correlate well with the accuracy of those models’ responses. other words, scores close to 1 line up with correct responses, and scores close to 0 line up with incorrect ones. another esperimento, they also found that using the Trustworthy Language Model with GPT-4 produced more reliable responses than using GPT-4 by itself.
Large language models generate text by predicting the most likely next word quanto a a sequence. future versions of its tool, Cleanlab plans to make its scores even more accurate by drawing acceso the probabilities that a model used to make those predictions. It also wants to access the numerical values that models assign to each word quanto a their vocabulary, which they use to calculate those probabilities. This level of detail is provided by certain platforms, such as Amazon’s Bedrock, that businesses can use to run large language models.
Cleanlab has tested its approach acceso provided by Berkeley Research Group. The firm needed to search for references to health-care compliance problems quanto a tens of thousands of corporate documents. Doing this by hand can take skilled team weeks. By checking the documents using the Trustworthy Language Model, Berkeley Research Group was able to see which documents the chatbot was least confident about and check only those. It reduced the workload by around 80%, says Northcutt.
another esperimento, Cleanlab worked with a large bank (Northcutt would not name it but says it is a competitor to Goldman Sachs). Similar to Berkeley Research Group, the bank needed to search for references to insurance claims quanto a around 100,000 documents. Again, the Trustworthy Language Model reduced the number of documents that needed to be hand-checked by more than half.
Running each query multiple times through multiple models takes longer and costs a lot more than the typical back-and-forth with a single chatbot. But Cleanlab is pitching the Trustworthy Language Model as a premium service to automate high-stakes tasks that would have been limits to large language models quanto a the past. The concetto is not for it to replace existing chatbots but to do the work of human experts. If the tool can slash the amount of time that you need to employ skilled economists lawyers at $2,000 an hour, the costs will be worth it, says Northcutt.
the long run, Northcutt hopes that by reducing the uncertainty around chatbots’ responses, his tech will unlock the promise of large language models to a wider range of users. “The hallucination thing is not a large-language-model problem,” he says. “It’s an uncertainty problem.”
