Code Generation Software Can Drive a Cognitive Revolution
Artificial intelligence updates
Sign up for myFT Daily Digest to be the first to know about artificial intelligence news.
English writer Percy Bysshe Shelley spoke of poets “The unrecognized lawmakers of the world. “More recently, and more prosaically, Harvard researcher Lawrence Lessig argued that the computer the code is the new law.
But computers, as well as humans, are now starting to write this code. Will this make algorithms the world’s little-known lawmakers?
For the moment, the question is premature. But the rapid progress of OpenAI Codex Code Writing Program, released by the San Francisco-based research firm over the summer, highlights the direction of travel. Derived from OpenAI’s powerful GPT-3 text generation model, Codex can instantly translate natural language into computer code in 12 different computer languages ââ- and can even translate between them.
GitHub, the open source software resource owned by Microsoft, is already using the technology to power Copilot, a software tool that suggests the next line of code, like an autocomplete function in a search engine. When Codex comes out of its current beta testing, OpenAI says it will help democratize access to smart software.
âBy removing barriers to entry, everyone is going to do more,â says Greg Brockman, co-founder and CTO of OpenAI. But he suggests that there is still a long way to go to improve the current Codex limits and ensure that it is used appropriately. Although surprising for its speed and flexibility, Codex currently generates the ârightâ code in 37% of use cases, which means it still requires close monitoring. âThe Codex is a point on the journey, but it’s not the end point,â says Brockman.
One of the areas where Codex is likely to be used first is digital education. Sana Labs, a Swedish start-up led by Joel Hellermark, explores how Codex can help deliver more personalized and dynamic learning courses, teaching users, among other things, the principles of coding. Hellermark’s point of view is that Codex will initially be used by experienced software developers as a tool for writing faster code. But over time, this will allow non-technical users to do all kinds of things that they currently deem impossible.
âYou will still need to have a hunch of how the software works. But it will give more value to computational thinking rather than the exact syntax of a particular computer language, âHellermark explains. âIntelligence is increasingly trivialized. “
In a launch paper published in July, OpenAI researchers trumpeted Codex’s technological capabilities, but also reported many concerns about automated code writing programs. As with GPT-3 itself, the datasets and code resources used to train the Codex model are partial and imperfect and should be examined for built-in errors and biases. “The Codex can generate code with a structure that reflects stereotypes about gender, race, emotions, class,” the newspaper warns.
As with all software, there is also a significant risk of misalignment, when programmers do not specify the desired outputs clearly enough. Think of the Disney movie Fancy in which Mickey Mouse as a sorcerer’s apprentice abuses magic to disastrous effects. By replicating buggy software, Codex can inject security holes into systems. It could also be used to create malware or authorize phishing attacks. Aware of the dangers, OpenAI says it is already focusing on ways to mitigate this potential damage.
The Codex is also part of a much larger software transformation, a cognitive revolution that could be comparable in impact to the Industrial Revolution, according to Kevin Scott, chief technology officer at Microsoft, who invested in OpenAI. This 2.0 software revolution, as Scott calls it, will help programmers move from an era of handcrafted manual work to one of mass production. “It changes the coding process: you very explicitly tell a machine to do a task in terms that are suitable for the machine, you teach a computer how to do a task in terms that are suitable for a human,” Scott says. .
The possibilities for improving human creativity and productivity are attractive. Like Shelley’s Poetry Design, software can awaken and expand the mind and help âlift the veil on the hidden beauty of the worldâ. But the recent history of flawed technology deployments suggests that we should hurry slowly and ensure that humans remain the recognized lawmakers of the world. âHuman oversight and vigilance is necessary for the safe use of code generation systems like Codex,â as the academic launch document rightly concludes.