AI-powered coding assistants have raised nearly $1 billion in funding since the beginning of last year, signaling that software engineering is shaping up to be the first “killer app” for generative artificial intelligence.
According to Dealroom, companies including Replit, Anysphere, Magic, Augment, Supermaven and Poolside AI have raised $433 million so far this year alone, bringing the total to $906 million since January 2023.
The rapid push to pour money into AI coding assistants indicates that computer programming will be the first job role transformed by the latest wave of AI technology.
“Today, software engineering and coding are the fields most impacted by AI,” says Hadi Partovi, CEO of the educational nonprofit Code.org and a longtime Silicon Valley investor and adviser to Airbnb, Uber, Dropbox, and Facebook. “At this point, software engineering without AI is like writing without a word processor.”
While Silicon Valley is growing increasingly convinced of the benefits of AI coding, some investors are questioning the economic merits of generative AI and the return on investment of the $1 trillion that big tech companies are expected to spend on computing infrastructure to support the technology over the next few years.
“It's much easier to monetize AI if you can embed your product into existing workflows and show immediate, tangible benefits,” said Hannah Seal, a partner at Index Ventures, which invested in the startup Augment along with Eric Schmidt.
To monetize AI tools, Seal's questions are, “What is the time to value, and how meaningful is that added value?” and, “With Coding Copilot, the answer is very clear,” she added.
The AI craze has startups and tech giants such as Microsoft, Amazon, Meta and Google building AI assistants and agents that can write and edit computer code, vying for an edge in a crowded field.
Code.org's board of directors includes Amazon's head of e-commerce, David Treadwell, and Microsoft's chief technology officer, Kevin Scott, who recently told Partovi that the company will stop hiring people to code without AI by the end of the year.
“The easier it is to program, the more technology you can build, so the demand will increase,” Partovi added.
GitHub, the world's largest software development platform and owned by Microsoft, was one of the first companies to turn a large-scale language model — the software underlying ChatGPT, which can generate text, images and code — into a coding assistant.
“When we worked with OpenAI's first major model, GPT-3, we realized relatively quickly that it was so good at writing code that we could build products on top of it,” said Thomas Dohmke, CEO of GitHub, which was acquired by Microsoft in 2018 for $7.5 billion.
That prototype evolved into GitHub Copilot, an AI coding assistant that was released broadly in 2022 and has nearly 2 million paid subscribers. “Today, the model writes better code than the average developer,” Dohmke said.
As of April, GitHub's revenue was up 45% year over year, and Microsoft CEO Satya Nadella said the company had annual revenue of $2 billion earlier this month.
“Copilot will represent more than 40% of GitHub's revenue growth this year and is already a larger business than the entirety of GitHub at the time of our acquisition,” he said during the company's earnings call on July 30.
According to Nadella, more than 77,000 organizations, from BBVA, FedEx and H&M to Infosys and Paytm, have adopted the two-year-old tool, representing a 180 percent increase over the previous year.
Still, IT departments at large companies have some concerns about the security implications of using automated programming tools to write production-level code.
But Domke said he doesn't expect AI-generated code to be deployed without manual checks and balances.
“Typically, companies that reported their internal statistics saw productivity increases of 20 to 35 percent,” Dohmke said, referring to clients such as Latin American e-commerce giant Mercado Libre and professional services group Accenture.
Recommendation
A McKinsey analysis last year found that AI's direct impact on software engineering productivity could range from 20 to 45 percent of current annual spend on the function, and benefits include generating initial code drafts, code correction, and refactoring.
“By accelerating the coding process, generative AI can boost the skill sets and capabilities needed in software engineering to code and architectural design,” McKinsey said.
Software engineers are already integrating AI assistants into their daily workflows, saying they're helping them not only work faster but also be more creative.
“I personally use GitHub Copilot every day to write code, often in conjunction with ChatGPT,” said Mark Tuscher, deep learning scientist and chief technology officer at German robotics startup Sereact.
He added that GitHub's tools are most useful for “repetitive tasks” like user interfaces and product backends, while ChatGPT helps him solve more abstract problems.
“ChatGPT presents you with some classic ideas or new papers, then you ask, 'How do you do this in Python?' and it generates the code for you,” Tuscher says. “Both tools are very good.”
Tuscher said that every programmer he knows uses these products and that they are “fundamentally changing the way we work,” but that the tools are not a replacement for programmers, only a powerful complement.
“There's no GenAI that knows good software architecture or how to put systems together,” he added. “That's something we have to figure out for ourselves.”
Video: AI power demand could stifle industry growth | FT Energy Source
Source link
