AI-powered coding assistants have amassed nearly $1 billion in funding since early last yr, an indication that software development is emerging as the primary “killer app” for generative artificial intelligence.
Companies like Replit, Anysphere, Magic, Augment, Supermaven and Poolside AI have raised $433 million to date this yr alone, in line with Dealroom, bringing the whole raised since January 2023 to $906 million.
The rush to pour money into AI coding assistants is a sign that computer programming shall be the primary job function to be transformed by the most recent wave of AI technology.
“Software development and programming are the areas most affected by AI today,” says Hadi Partovi, executive director of the nonprofit educational organization Code.org and a longtime investor and advisor to Airbnb, Uber, Dropbox and Facebook in Silicon Valley. “Software development without AI today is a bit like writing and not using a word processor.”
The growing conviction in Silicon Valley concerning the advantages of AI programming contrasts with the doubts of some investors concerning the economic advantages of generative AI and the likely returns from the trillion-dollar investments planned by major technology corporations within the computing infrastructure to support this technology in the approaching years.
Hannah Seal, a partner at Index Ventures, which invested within the startup Augment together with Eric Schmidt and others, said it’s “much easier to generate profits with AI when you may embed your product into an existing workflow and make the advantages immediately visible.”
For AI tools to generate profits, Seal must ask herself: “How long does it take so as to add value, and the way significant is that value?” She adds: “When programming as a co-pilot, the reply could be very clear.”
The craze for artificial intelligence has driven startups and tech giants like Microsoft, Amazon, Meta and Google to vie for dominance in a crowded sector by developing AI assistants and agents that may write and edit computer code.
A senior executive on Code.org's board of directors, which incorporates David Treadwell, Amazon's head of e-commerce, and Kevin Scott, Microsoft's chief technology officer, recently told Partovi that his company will stop hiring individuals who program without artificial intelligence by the top of the yr, he said.
“The easier (programming) becomes, the more demand increases because so way more technology could be built,” Partovi added.
Microsoft-owned GitHub, the world's largest software development platform, was certainly one of the primary to rework a big language model – the software that underlies ChatGPT and may generate text, images or code – right into a coding assistant.
“When we used GPT-3, OpenAI's first major model, we realized relatively quickly that it was so good for writing code that we could construct a product on it,” said Thomas Dohmke, CEO of GitHub, which was acquired by Microsoft in 2018 for $7.5 billion.
The prototype became GitHub Copilot, an AI programming assistant that launched broadly in 2022 and has nearly 2 million paying subscribers. “Now the model writes higher code than the common developer,” Dohmke said.
GitHub's revenue increased 45 percent year-over-year in April, and annual revenue was $2 billion at the start of this month, in line with Microsoft CEO Satya Nadella.
“Copilot has driven over 40 percent of GitHub's revenue growth this yr and is already a bigger company than GitHub overall after we acquired it,” he said on a July 30 quarterly earnings call.
More than 77,000 organizations – from BBVA, FedEx and H&M to Infosys and Paytm – have adopted the two-year-old tool, Nadella said, a number that represents a 180 percent increase from last yr.
However, IT departments in large corporations have some reservations concerning the security risks related to using automated programming tools to create production-ready code.
However, Dohmke said he wouldn’t expect AI-generated code for use without manual checks and balances.
“In general, corporations which have presented internal statistics have seen productivity gains of between 20 and 35 percent,” said Dohmke, citing clients comparable to Latin American e-commerce giant Mercado Libre and consulting firm Accenture.
A McKinsey evaluation last yr found that the direct impact of AI on software development productivity might be between 20 and 45 percent of current annual spending on the function, with advantages comparable to generating first drafts of code, code fixes and refactoring.
“By accelerating the coding process, generative AI could transfer the talents and competencies required in software development to code and architecture design,” said McKinsey.
Software developers say they’ve already integrated AI assistants into their each day workflow and that this not only helps them be faster but additionally more creative.
“I personally code with GitHub Copilot each day, often along with ChatGPT,” said Marc Tuscher, deep learning scientist and chief technology officer of Sereact, a German robotics start-up.
GitHub's tool is most useful for “repetitive tasks,” comparable to user interfaces and the backend of products, he added, while he uses ChatGPT to assist solve more abstract problems.
“ChatGPT will present some classic ideas and a few latest papers and you then can ask, 'How would this be done in Python?' and it produces code,” Tuscher said. “Both tools are very, very cool.”
Although all programmers he knows use these products and “this fundamentally changes the way in which we work,” Tuscher believes that the tools are nothing greater than powerful helpers for programmers and never a substitute.
“No GenAI knows good software architecture or methods to put systems together,” he added. “That's still something we have now to take into consideration ourselves.”
