Let us know about free updates
Artificial intelligence is poised to surpass humans in writing as a major group, including the major group of Openai, Humanity, Google and other groups, as well as competition to release systems that rebuild the software industry.
San Francisco-based Openai released a series of new models this week, suggesting by its independent benchmark.
The new GPT-4.1, O3, and O4-MINI models are more effective at solving programming problems, the latter using “inference” to give time to think through complex queries.
Openai on Wednesday also announced a free-to-use system called Codex CLI, a so-called AI “agent” designed to use models to assist with coding tasks.
These moves coincide with recent efforts from rivals from Google, Google, Meta and numerous startups, betting on coding as one of the clearest initial uses for large-scale language models.
With an emphasis on programming as the next frontier of AI systems, it presents one of the most concrete examples of how technology can transform industry.
“This is the year when AI is forever superior to humans in competitive code,” said Kevin Weil, Openai’s chief product officer, in this week’s overwhelming podcast. He compared AI advances in chess a few years ago that surpassed AI, claiming that it had a more democratic impact “if anyone could create software, the world.”
According to major industry figures, LLM expanded the software development process by generating an entire block of code based on several textual instructions. AI systems can also try to identify errors and fix them.
Over the past year, AI models have become much more aware of the inference and complex patterns of logical resolution for problems presented in programming.
In 2023, AI systems only solved 4.4% of coding problems based on industry tests called SWE Benches. This figure jumped to 69.1% this year.
Meanwhile, a survey by Microsoft’s coding platform Github found that 92% of US-based developers use AI coding tools.
“AI coding saves thousands of dollars for engineers,” said Misha Laskin, co-founder and CEO of Coding Startup Reflection AI. “For some of these categories, you can do that on demand with something that might have paid $10,000. We’re entering a massive, unprecedented market.”
Startups such as Reflection have attracted strong investors’ interest, raising $130 million so far to date to fund Sequoia and Lightspeed. Anysphere raised $155 million in January at a valuation of $2.5 billion, after Coding Automation Tool Cursor.
“We’re reducing the cost of doing intelligent work that means we have to rethink some of our roles,” says Isocanto, a poolside co-founder who received a $500 million valuation in October with a $3 billion valuation from investors including Bain Capital and nvidia.
Recommended
Last year, Meta launched a model called Code Llama. This uses text prompts to generate and discuss code. Anthropic has its own coding product, Claude Code, which was launched in February.
Mike Krieger, chief product officer in anthropology, said the role of software engineers increasingly involves “understanding (the user’s) requirements, working as a team, understanding that what you build is actually the right thing to build.”
“It’s about defending your ideas, or seeing how those things unfold (and become) like the puppet master or the orchestra conductor of these (AI) agents,” he added.
“I don’t think coding will disappear at all,” said Thomas Wolf, co-founder of Hugging Face, an open-source AI platform. “The coder simply uses this tool to go faster.”
Additional Reports from George Hammond of San Francisco
