Shortly AI is unlikely to replace software engineers, but will dramatically change the way they work in the future, especially if it can instruct natural language machines to generate code.
Several organizations – from OpenAI and Microsoft to Amazon and research labs like DeepMind – have trained neural networks to learn to code. A recent survey of more than 2,000 developers by GitHub found that the vast majority of respondents felt that GitHub’s Copilot helped increase their productivity because the AI tool can act like a super autocomplete and help developers , Boilerplate code for programs to write faster.
But will programming jobs be taken over by machines in the future? “I don’t think AI will come close to replacing human developers,” Vasi Philomin, Amazon’s vice president of AI services, told IEEE Spectrum.
It’s possible that developers don’t need to learn the syntax and vocabulary of programming languages and instead focus on understanding concepts and systems to design programs while letting the AI do all the boring, basic programming work, he said. In other words, you describe how an application works and a machine learning model outputs the appropriate code to compile or run.
Peter Schrammel, co-founder of Diffblue, a company focused on automating Java code, agreed that programming jobs will change and engineers can focus more on difficult, creative problems.
“Software developers are not going to lose their jobs because an automation tool replaces them,” he said. “There will always be more software that needs to be written.”
Private medical images in public AI training data set
Photos of people taken in medical settings were scraped into a public dataset to train text-to-image models, all without consent for that particular use case.
An artist named Lapine was horrified to see that two private images, taken almost a decade ago for surgical purposes, are in the LAION-5B dataset used to train popular models like Stable Diffusion and Google’s Imagen. Lapine told Ars Technica she has dyskeratosis congenita, a rare genetic condition that affects bone marrow function and affects skin tissue.
“It affects everything from my skin to my bones and teeth,” she said. “In 2013 I underwent a small series of facial contour restoration surgeries after undergoing so many mouth and jaw surgeries. These images are from my most recent procedures with this surgeon.” Lapine said the surgeon who kept the medical photos died in 2018 and somehow the data was preserved, shared online and downloaded.
Lapine now wants her photos removed from the record to prevent more models from being trained with sensitive, private information. “I would like a way for anyone to request that their picture be removed from the record without sacrificing personal information. Just because they removed it from the internet doesn’t mean it should be public information or even the web at all,” she said.
OpenAI releases free, open speech recognition model
OpenAI has released an open source neural network called Whisper capable of speech recognition across different languages and accents.
Whisper was trained on a whopping 680,000 hours of audio data scraped from the internet. The model splits input data into 30-second chunks to feed to an encoder. A decoder is trained to generate subtitles for the audio snippet; It is able to identify languages and transcribe speech to English text automatically.
Examples posted by OpenAI show that Whisper can accurately transcribe fast and muddled speech with a heavy Scottish accent and translate clips of Korean pop songs.
“We provide open-source models and inference code to serve as a basis for building useful applications and further research into robust language processing,” OpenAI announced. “We hope that Whisper’s high level of accuracy and ease of use will enable developers to add voice interfaces to a much broader range of applications.”
You can read more about the model here [PDF] and access the code here.
How can we prevent AI from ripping off our work?
Artists are thinking about how best to protect their works from being ripped off and copied by internet users with AI models. Especially when people type descriptions like “a summer afternoon in Times Square, New York City in the style of Rembrandt” into ML software and save the output.
Established artist Greg Rutkowski’s name has been entered as a text prompt into art-generating models more than 93,000 times, more than some of the world’s most celebrated artists like Pablo Picasso or Leonardo da Vinci, who have each appeared in about 2,000 prompts or fewer, MIT Technology Review reports . In other words, people are getting AI models to produce artworks that specifically rip off Rutkowski’s style, let alone other artists.
In fact, people messing around with tools like Midjourney or Stable Diffusion can produce multiple images that look like Rutkowski’s epic imaginative digital paintings in a matter of seconds. No skills are required other than a text description. Artists like Rutkowski are trying to figure out how these text-to-image systems will affect his work and livelihood in the future.
Some want their work removed from training datasets so models can’t reproduce their style, and others believe AI companies should seek to build working relationships with museums and artists to better support their work, according to illustrator Karla Ortiz .
“It’s not just artists. It’s photographers, models, actors and actresses, directors, cinematographers,” she said. “Any type of imaging professional has to grapple with that particular question right now.”
Cohere For AI Scholars program
The non-profit research arm of language model startup Cohere has launched a program to hire engineers who want to start a career in machine learning research but haven’t published any articles yet.
Candidates are not required to have a specific degree or experience in academic work. Those accepted into the program will be paired with experts and work remotely to investigate a specific problem in natural language processing from January to August 2023 and receive financial support.
“We designed this program to provide more entry points into machine learning and expand access to world-class research and engineering knowledge,” said Sara Hooker, director of Cohere for AI The registry.
“The best and brightest minds in machine learning push boundaries and often take different paths to research. That is why we are working on fundamentally changing where, how and by whom research is carried out. This program is a step in that direction.”
“Supporting the next generation of emerging NLP researchers is critical to groundbreaking new advances in machine learning. Unfortunately, today there are very few facilities to conduct research on innovative NLP problems and limited access to large-scale ML experiments. By expanding access to participation in basic research – particularly among people from alternative backgrounds – the Scholars Program aims to change that,” she said.
The application deadline for the program is November 7th. ®