Machines Writing Code Are Making Huge Advancements
A major accomplishment for DeepMind AI announced claims that the software has enough smarts to write simple code and solve some basic problems that were not programmed into it. It solved them on its own.
A first glance it may not appear to be that big of a deal, but looking down the road it could be a real game-changer. DeepMind Technologies is a British AI subsidiary of Alphabet Inc., and a research laboratory founded in September 2010. DeepMind was acquired by Google in 2014.
DeepMind has created an AI capable of writing code to solve arbitrary problems posed to it, as proven by participating in a coding challenge and placing — well, somewhere in the middle. It won’t be taking any software engineer jobs just yet, but it’s promising and may help automate basic tasks.
The team at DeepMind, a subsidiary of Alphabet, is aiming to create intelligence in as many forms as it can, and of course these days the task to which many of our great minds are bent is coding. Code is a fusion of language, logic, and problem-solving that is both a natural fit for a computer’s capabilities and a tough one to crack.
This isn’t the first or only attempt at something like this. OpenAI has its own Codex natural-language coding project, and it powers both GitHub Copilot and a test from Microsoft to let GPT-3 finish your lines.
Deepmind Gives Other AI Advice
Deepmind’s recent paper points out some flaws it found in another algorithm.
“Recent large-scale language models have demonstrated an impressive ability to generate code, and are now able to complete simple programming tasks. However, these models still perform poorly when evaluated on more complex, unseen problems that require problem-solving skills beyond simply translating instructions into code.”
The article in techcrunch.com, written by Devin Coldewey, explains that Deepmind’s status was only in the middle of the pack when it came to solving the code problems compared to humans doing the same work. But code writers may want to up their game when it comes to competing with a writing algorithm.
To take on the domain, DeepMind trained a new model using selected GitHub libraries and a collection of coding problems and their solutions. Simply said, but not a trivial build. When it was complete, they put it to work on 10 recent (and needless to say, unseen by the AI) contests from Codeforces, which hosts this kind of competition.
However, the test Deepmind completed was its first crack at it. If a human had reached the same level it would be considered an average outcome as it would have taken many more tries.
“I can safely say the results of AlphaCode exceeded my expectations,” said Mike Mirzayanov. “I was skeptical because even in simple competitive problems it is often required not only to implement the algorithm, but also (and this is the most difficult part) to invent it. AlphaCode managed to perform at the level of a promising new competitor.”
You can dive deeper into the way AlphaCode was built, and its solutions to various problems, at this demo site.
The author includes an example of Deepmind’s work in his article. This really could be a turning point in writing code, especially if it is writing at the speed of AI.
read more at techcrunch.com