Coming Soon to a Workplace Near You: An AI-based Engineer

Adept, a new startup, is developing an AI-powered digital assistant that can interface with all known software and APIs.

“Hey, Siri? Can you do my job for me?”

For most of us, AI-based “assistants” are becoming inescapable in our everyday lives. From Siri to Alexa, many of us now rely on these hands-free digital interfaces to answer simple questions and perform straightforward tasks, like setting a timer.

But many companies are already dreaming of the next generation of AI-powered assistants. The goal? General intelligence that can perform advanced tasks and meaningfully contributing to productivity. In addition to the technology giants like Google, many startups are making waves when it comes to natural language processing (NLP) and the development of advanced algorithms that can create these types of advances in AI technology.

Adept’s current team includes engineers with experience at Google, DeepMind and Meta. (Image courtesy of Adept.)

Adept’s current team includes engineers with experience at Google, DeepMind and Meta. (Image courtesy of Adept.)

In mid-August 2022, it was announced that the startup company, Adept, is using Oracle Cloud Infrastructure (OCI) along with NVIDIA solutions to develop an AI-powered “teammate” that can perform engineering tasks. Adept has developed a solution that uses a natural language computing interface to perform a range of tasks that engineers and technicians would typically perform on a computer or using the Internet.

According to Adept and Oracle, the newly developed “AI engineer” can be operated via voice command. Engineers can then use their AI-based “teammate” to streamline daily tasks that are difficult or tedious. In a way that is similar to Siri calling a friend for you, Adept’s new AI assistant can help engineers perform a range of technical tasks with only a few words.

“AI continues to rapidly grow in scope but until now, AI models could only read and write text and images; they couldn’t actually execute actions such as designing 3D parts or fetching and analyzing data,” said David Luan, chief executive officer at Adept. “With the scalability and computing power of OCI and NVIDIA technology, we are training a neural network to use every software application, website, and API in existence—building on the capabilities that software makers have already created. The universal AI teammate gives employees an ‘extra set of hands’ to create as fast as they think and reduce time spent on manual tasks. This in turn will help their organizations become more productive, and nimble in their decision-making.”

The Road to an AI-Powered Engineer

The team at Adept is no stranger when it comes to AI. The founding team brings backgrounds from Google, DeepMind and Meta and has joined forces to develop the next generation of AI-powered assistants. In imaging the future of AI, the cofounders want to build technology that can perform useful actions for people, not just engage in conversation and execute menial tasks.

Two of Adept’s cofounders, Ashish Vaswani and Niki Parmar, invented the Transformer in 2017, long considered a game changer in AI. The Transformer algorithm was the first neural network-based machine learning model that actually worked in the majority of use cases. As described on Adept’s website, following its creation, the Transformer has been adopted as the foundational architecture for most large-scale AI models, including the current engine that powers the Google search tool.

However, the team behind Adept saw one major limitation with the AI tools available on the market. Although most of the models were advanced enough to read, write and dictate, they could not perform actions that were truly “useful” to others. So, the Adept technology is being developed as a new type of neural network that can utilize any software tool or API in the world and perform technical tasks.

Oracle and NVIDIA Are Powering the Development of Adept’s AI

To build an assistant capable of performing advanced engineering, the team needed a fast and relatively cheap way to develop and train their AI. They chose to run thousands of NVIDIA GPUs on clusters of the OCI bare metal compute instances. Oracle and NVIDIA customized a solution for Adept so the company could use thousands of NVIDIA’s A100 Tensor Core GPUs.

Oracle describes its custom solution as ideal for startups as it brings access to high performance computing (HPC) capabilities with the flexibility of the cloud. So, Adept could massively parallelize its HPC jobs using ultra-low latency remote direct memory access (RDMA) cluster networks. However, the company remains elusive on exactly how it is training the AI and the full range of software with which it can currently interface.

Adept is not alone in its investment in the future of AI-based assistants. The company just closed a $65 million Series A funding round.

“With brilliant minds from DeepMind, OpenAI, and other AI and ML pioneers, Adept is building the next generation of user interfaces for software applications,” said Kari Briski, vice president, AI and high-performance computing (HPC) software development kits, NVIDIA. “By working with Oracle to provide Adept with an industry-leading GPU engine and a wide range of AI and ML software tools, we’re making innovative AI systems possible.”

So, Do Engineers Need to Worry About Their Jobs?

Adept’s website describes the power of its technology as a “universal collaborator.” We all have parts of our job that we find more or less interesting, and the Adept tool will allow engineers and other professionals to focus on the parts of their jobs that they enjoy the most. So, as an example, you could ask your AI-based assistant to make your expense report for the month or modify elements of a blueprint based on feedback.

The main concern that is likely on every engineer’s mind is what this might mean for their future employment prospects. For now, it seems that engineers can sleep easily, knowing they are not without a job. Even with Adept’s flashy marketing, it appears to be billing the AI “teammate” as a valuable tool for engineers, not a solution capable of replacing the entire profession. The tool seems to be described as a method for automating and simplifying the elements of technical jobs that engineers find less enjoyable. When it comes to making creative or advanced decisions, human intelligence will likely still be required to direct the tool through its use of “every software in existence.”

But this might not bode well for more entry-level positions, where the relatively straightforward tasks of some junior engineers or technicians might be more readily automated. Although Adept has not commercially released its solution, and it remains vague about the technology’s complete capabilities, some companies might still view the solution as a way to reduce the need for entry-level personnel.

Others might be concerned about what this means for the development of general intelligence and the realization that computers are truly competent at tasks previously believed to require human oversight. Interestingly, Adept seems particularly excited about its software precisely because it does not achieve that level of general intelligence. It prefers to focus on the technology’s ability to interface with all existing software tools, as opposed to developing an unsupervised approach to general intelligence. It seems that the final product will hopefully improve the specificity and utility of digital assistants, making it possible for technical jobs to be better supported by these types of tools.

What Will the Future Hold?

Of course, Adept is not alone in this space, and DeepMind is working on its own AI tool that can perform hundreds of tasks, like booking a flight, with what the company describes as “human-level” accuracy. However, the research paper describing this AI relied on people performing millions of demonstrations of computer-based tasks to train the technology. So, although the Adept technology is not close to being on the market, it seems like its competitors are also firmly rooted in the R&D stage.

Meanwhile, we will have to wait and see what these tools are truly capable of when they are released in their beta stage. One thing is for sure, based on their Series A funding, Adept is more than capable of selling the potential of its advanced AI assistant.