[ad_1]
Tom Smith gave him a job interview as soon as he got his hands on Codex, a new artificial intelligence technology that writes its own computer programs.
When interviewing for big money jobs at Silicon Valley companies like Google and Facebook, he asked if he could overcome the “coding challenges” that programmers often face. Can one write a program that replaces all spaces in a sentence with a hyphen? Better yet, can one write one that identifies invalid zip codes?
He did both instantly before completing several other tasks. “These are problems that are difficult for many people, myself included, to solve, and they type the answers in two seconds,” said Mr. Smith, a veteran programmer who runs an artificial intelligence startup called Gado Images. “It was scary to watch.”
Codex looked like a technology that would soon replace human workers. As Mr. Smith continued to test the system, he realized that his skills went beyond his ability to answer ready-made interview questions. It can even translate from one programming language to another.
But after working with this new technology for several weeks, Mr. Smith believes it poses no threat to professional coders. In fact, like many other experts, he sees it as a tool to increase human productivity. It can even help a whole new generation learn the art of computers by showing them how to write simple bits of code, almost like a personal tutor.
“This is a tool that can make a coder’s life so much easier,” said Mr. Smith.
About four years ago, researchers at labs like OpenAI started designing neural networks. analyzed an enormous amount of prose, including thousands of digital books, Wikipedia articles, and any other text posted online.
By detecting patterns in all these texts, the networks learned to predict the next word in a string. When someone writes a few words to them “universal language models”, they can complete the thought with whole paragraphs. In this way, a system – an OpenAI product called GPT-3 – could write its own Twitter posts, speeches, poems and news articles.
To the surprise of even the researchers who built the system, he could even write his own computer programs, although they were short and simple. Apparently, he had learned from countless programs posted on the internet. So OpenAI went one step further and trained a new system, Codex, on an enormous array of both prose and code.
The result is a system that understands both prose and code, up to a point. You can ask for it to snow on a black background in plain English and this will give you a code that creates a virtual snowstorm. If you want a blue bouncing ball, he will give it to you too.
“You can tell him to do something and he will do it,” said Ania Kubow, another programmer using the technology.
Codex can create programs in 12 computer languages and even translate between them. However, he often makes mistakes and, although his skills are impressive, he cannot reason like a human. He can recognize or mimic what he has seen in the past, but is not agile enough to think for himself.
Sometimes programs created by Codex do not work. Or they contain security flaws. Or they can’t even come close to what you want them to do. OpenAI estimates that Codex generates the correct code 37 percent of the time.
When Mr. Smith used the system as part of a “beta” testing program this summer, the code he produced was impressive. Sometimes, however, it only worked if he made a small change, such as tweaking a command to suit his software installation, or adding a digital code needed to access the internet service he was trying to query.
In other words, Codex was only really useful to an experienced programmer.
But it can help programmers do their daily work much faster. It can help them find the basic building blocks they need or guide them to new ideas. GitHub, a popular online service for programmers, now offers Co-pilot, a tool that uses technology to suggest your next line of code.
“It’s a way to write code without having to write a lot of code,” said Jeremy Howard, who founded the Fast.ai artificial intelligence lab and helped create the language technology OpenAI’s work is based on. “Not always true, but close enough.”
Mr. Howard and others believe that Codex can also help novices learn to code. It is particularly good at creating simple programs from short English descriptions. And it works in the other direction as well, explaining complex code in plain English. Some, including Joel Hellermark, an entrepreneur in Sweden, are already trying to turn the system into a teaching tool.
The rest of the AI landscape looks similar. robots increasingly stronger. So are chatbots. designed for online conversation. DeepMind, an artificial intelligence lab in London, recently built a system. Instantly identifies the shape of proteins in the human bodyIt is an important part of designing new drugs and vaccines. This task once took scientists days or even years. But these systems replace only a small part of what human experts can do.
In the few areas where new machines can replace workers instantly, these are usually jobs that the market fills slowly. Robots, for example, are becoming increasingly useful in expanding shipping hubs that are struggling to find the workers they need to keep up.
With his newly founded Gado Images, Mr. Smith began building a system that could automatically sort the photo archives of newspapers and libraries, resurrect forgotten images, automatically write captions and tags, and share photos with other publications and businesses. But technology could only do part of the job.
It can browse a large library of photos, identify the types of images that might be useful, and browse captions faster than humans. But finding the best and most important photos and tagging them appropriately still required an experienced archivist.
“We thought these tools would completely eliminate the need for humans, but we learned years later that this was not really possible – you still needed a skilled human to review the output,” said Mr. Smith. “Technology gets things wrong. And it can be biased. You still need someone to review what they’ve done and decide what’s good and what’s bad.”
Codex expands on what a machine can do, but is another indication that technology works best with humans at the controls.
“The AI is not working as expected,” said Greg Brockman, chief technology officer at OpenAI. “It felt like it was going to do this job and this job, and everyone was trying to figure out which one would go first. Instead, it doesn’t replace any work. But it takes the heavy lifting out of them all at the same time.”
[ad_2]
Source link