So far this week we’ve seen generative AI come to CRM from toptechtrends.com/2023/03/07/salesforce-plans-to-incorporate-generative-ai-across-the-platform/”>Salesforce and to customer service chatbots from toptechtrends.com/2023/03/08/forethought-aims-to-build-more-accurate-chatbots-with-more-constrained-generative-ai-models/”>Forethought – and that’s just the ones I’ve covered personally. Today, we look at ThoughtSpot’s generative AI entry, which lets you query your data using natural language to get text or a graph back, as appropriate, with the correct response.
This is an approach that Thoughtspot has been working towards for years. In 2019, when I spoke to the company on the occasion of itstoptechtrends.com/2019/08/28/thoughtspot-hauls-in-248m-series-d-on-1-95b-valuation/”> $248 million Series E – at the time the company was valued at $1.95 billion – it was already using AI to turn plain language queries like ‘what is the best selling shoe in Portland’ into SQL behind the scenes and delivering an answer.
That’s not altogether that different from what it’s announcing today, but now it’s relying on GPT-3 to allow users to enter a query and similarly get a result. It just took some time for the technology to catch up with the vision.
“We always wanted to build a pure natural language intent-driven interface. In fact, I can tell you four years ago, we had a project internally to build our own large language model. We paused that because we knew that when the public large language model capabilities came on tap, we would be able to put it on top [of our our products] and deliver the best, most flexible, highly accurate platform – and that’s what we have done,” ThoughtSpot CEO Sudeesh Nair told TechCrunch.
Perhaps the biggest criticism of ChatGPT is that it sometimes gives the wrong answer, yet it is essential for Thoughtspot to deliver an accurate answer when using the technology to query data. In this regard, the company takes advantage of the GPT-3 API to help translate the natural language into SQL, but it also adds its own layer to make sure it delivers the single correct answer because with data there isn’t room for error.
“This is why while large language models make sense, making them trustworthy for business computing, for database queries, is a complete game changer and…we have actually built the stack differently to deliver accuracy and trust at scale at large companies,” Nair said.
The company understands that no matter how hard they try, they won’t always get it right, so they have also built in a feedback loop to let them know when they made a mistake, either through inaccuracy, or the customer presenting the data differently from how the algorithm does.
The user can change the way it measures something by editing the query, or give a thumbs up or thumbs down, based on the response, and the program can use this feedback to fine-tune answers in the future.
Different types of AI come into play, both when the user is asking the question and Thoughtspot is retrieving and generating the answer. What’s more, Thoughtspot AI can help companies data experts build customized data models for their source data.
The company was founded in 2012 and has raised over $660 million, per Crunchbase. A private beta of the new integration with GPT-3 opens today.
toptechtrends.com/2023/03/09/thoughtspot-adds-gpt-3-integration-to-help-customers-query-data/”>ThoughtSpot adds GPT-3 integration to help customers query data by toptechtrends.com/author/ron-miller/”>Ron Miller originally published on toptechtrends.com/”>TechCrunch
About The Author
