AlphaCode is an AI coding software that outperforms most humans.
The software was entered into a coding competition where it scored better than 46.7% of human coders.
The company is currently working on improving its capabilities
AlphaCode, a brand-new AI system, may bring us a step closer to a world where various jobs are outsourced to AI. Researchers from DeepMind, which is a subsidiary of Alphabet, believe that this system will someday assist coders, but might not replace them.
“It’s very impressive, the performance they’re able to achieve on some pretty challenging problems,” says the Head of the computer-assisted programming group at MIT, Armando Solar-Lezama.
What is AlphaCode capable of?
Previously, OpenAI had released the Codex system, which was taken as a benchmark for AI code writing. OpenAI had developed GPT-3, which is a language learning model (also used by OpenAI’s ChatGPT), that allows it to imitate and interpret the human text to carry out a host of functions. Codex can be given a description of what one wants the software to do, and it will write the corresponding code. This software did face challenges when given tricky problems to solve, however, we’ve already seen ChatGPT display a penchant for writing complex code.
The team at AlphaCode was focused on solving complex problems. They also developed a language learning model that contained GBs of code from GitHub. The researchers used the model to turn problem descriptions into code. After its basic training was complete, AlphaCode was able to solve 34% of the complex problems it was presented with. Moreover, to understand
how well the AI software performs within the industry, DeepMind entered into various coding competitions and learned that it outperformed at least 45.7% of other participants. On checking the code it generated, researchers found that it did not plagiarise code from its learnings.
“It continues to be impressive how well machine-learning methods do when you scale them up,” he says. The results are “stunning,” said Wojciech Zaremba, a co-founder of the Codex paper, and OpenAI.
Speaking about the ability of the software to start working for humans, Solar-Lezama said “Even if this kind of technology becomes supersuccessful, you would want to treat it the same way you treat a programmer within an organization. You never want an organization where a single programmer could bring the whole organization down.”
Follow Us
Kajoli Anand Puri
Kajoli is a tech-enthusiast with a soft-spot for smart kitchen and home appliances. She loves exploring gadgets and gizmos that are designed to make life simpler, but also secretly fears a world run by AI. Oh wait, we’re already there. View Full Profile
