Skip to content
New Salesforce AI chief eyes future with voice-driven coding – TechCrunch


As we start to see AI progress in business, the way we interact with machines is starting to change. Companies like Salesforce are looking for new opportunities for AI to have a more direct impact on customers.

While it’s certainly valuable to use AI to identify the customer most likely to unsubscribe or purchase, it’s only one step in the process, and that’s just the beginning. how AI could change the way we work in the future.

Salesforce’s AI journey began in 2016 with the launch of its AI framework called Einstein. In reality, Einstein was never designed as a product, but rather as a set of intelligence capabilities that could touch every aspect of the Salesforce stack. The original crew that brought this to life has largely moved on, but the work continues.

A year ago, the company tapped former Stanford professor Silvio Savarese to be its chief scientist. One of the reasons he was willing to leave academic life was his ability to pursue advanced research with large datasets, large staff, and the resources of a company like Salesforce.

He said he wanted to continue the research he had been conducting for two decades with the aim of bringing skills to people who lacked specific training. “One of the main directions I’m pushing here is really getting AI to empower people in business in new ways, and I’m really excited to deliver that power with experiences that are so simple that anyone who can use them,” Savarese explained.

To achieve this overall goal, one of the main initiatives he and his 100-person research team have pursued is the voice programming approach the company has dubbed CodeGen. The idea is to let people simply describe in plain language what they want to do, and the AI ​​will produce code based on the natural language instructions.

But it’s not just about telling AI technology what you want; Savarese said it was more of a conversation. “CodeGen truly offers a new way to develop software. Rather than writing code directly, users would simply describe the problem they are trying to solve in plain language in a conversation. So the conversation part is very, very important,” he explained.

What he means is that you can ask something and the AI ​​will ask for clarification and there will be a back and forth like the example provided in a Salesforce blog post explaining CodeGen:

Example of conversational coding from Salesforce using the CodeGen tool. Picture credits: Selling power

Although still in the experimental stage of development, they are making progress in building models that would suit two distinct audiences. “The goal is to address a couple of users. One is for more experienced developers, who in this case CodeGen will help them write the code and take care of the manual parts of the processing, those parts that aren’t so interesting from a coding perspective. The second user is people with no coding experience, so almost no coding expertise, but CodeGen can still give them a way to build software to solve real problems,” he said.

Salesforce is trying to achieve something with conversational coding that has never been done before. While Microsoft is working on something similar with the GPT3 framework, it’s what Savarese calls large-scale deep learning and it involves extremely complex models.

“This is a fundamental model for coding, so CodeGen is built on a massive autoregressive model with 16 billion parameters, which is trained with a very large amount of data,” he said. Here it bifurcates use cases with examples for the model depending on whether the user is experienced or non-coder.

While the project is still in the proof-of-concept stage, the next step is to release it to the internal Salesforce developer community, which will happen when Savarese presents his findings at an internal conference later this month. -this.

If the project goes beyond the experimental stage, the idea would be to give data scientists and business analysts the means to use Tableau, the company acquired by Salesforce in 2019 for nearly $16 billion, to create programs based on data and make it more commercially accessible. .

Voice-driven coding might just be the first step here, as other capabilities like content creation, website layout, and other tasks might be similarly reduced to a simple description. “The inspiration comes from the need to have an easy way to communicate with AI systems, and the ability to use language to create better communication to inform certain processes.”

techcrunch

Not all news on the site expresses the point of view of the site, but we transmit this news automatically and translate it through programmatic technology on the site and not from a human editor.