OpenAI and start-ups race to generate code and transform software industry

Stay notified with free updates

Simply sign up Artificial intellect MEFT Digest – Directly delivered to your inbox.

Artificial intelligence is ready to exclude people in writing codes as a top group, including OpenAI, ethnographic and Google, re -forming the software industry.

Based on San Francisco Open This week a new model suit has revealed that distinct criteria for computer programming are still in the best.

Its new GPT -1.3, and 3, and 4 -minute models are more effective in solving programming problems, using the second, using ‘logic’, giving more time to think through complex questions.

Wednesday OpenAIO Codex announced a freely available system called CLI, it is a so -called AII Models of “Agent” are designed to use users to help users in coding functions.

These steps match the rival ethnographic, Google, Meta and Start-ups a recent attempt to host a host, which has bet on coding as the most clear initial use for big language models.

Emphasis on programming as the next border of AI systems signals to one of the clear examples of how the industry can transform, thousands of software developers already use new models in their work.

“This year … AI is better than people in competitive codes,” said Kevin Wail, Chief Product Officer of the OpenAI. Over -powerful Podcast this week. He compared the progress with the AI ​​in the cheeks several years ago, but argued that it had a more democratic impact “If everyone could create software on the world”.

Statistics of the top industry say that LLM has enhanced the software development process by creating the entire block of the code based on a few text instructions. AI systems can also detect errors and try to correct them.

In the past one year, AI models have become more able to understand complex patterns, logic about programming problems and logically solving.

In 2023, AI systems were able to solve 4.4 percent of the coding problems based on an industrial exam called Need-Bench. This image jumps at 69.5 percent this year.

Meanwhile, Microsoft’s coding platform GitHhab research has shown that 92 percent of US-based developers use AI coding equipment.

“AI coding is saving thousands of dollars for an engineer,” Coding Start-up Reflection AI co-founder and CEO Misha Laskin says. “For a few categories, it is able to do it for you with something that has given you 10,000 for you. We’re entering an unprecedented big market.”

Start-ups like reflection have attracted strong Investors’ interestIt has raised $ 130 million to $ 130 million so far with the funds of Sikoia and LightShid. Coding automation tools have raised $ 2.5bn $ 105 million in January in January.

“We are spending what we mean to do what we mean to do intelligently, we need to revisit some of the role we mean,” said Pulside’s co-founder, including Bain Capital and NVDia, by determining the price of $ 3 billion from investors.

Meta launched a model called Code Lama last year, which used text requests to create and discuss the code. Anthropic has its own coding product, clad code, which was launched in February.

Anthropic Chief Product Officer Mike Criger says the role of software engineer will be increasingly involved “Understanding the requirements” [of users]Working as a team, and determining what you have created is actually the right thing to make it “.

“It’s more about giving advice to your idea or how these things go into effect on how to do these things [and becoming] Almost like a doll master or their orchestra conductor [AI] Agents, ”he added.

“I don’t think the coding will disappear at all,” the Open Source AI platform, the co-founder of the hug face, said Tomas Wolf. “Coders will use this tool to go faster.”

Additional Report of George Hammond in San Francisco