OpenAI’s GPT-3 based pair programming model – Codex – is now open to private beta testers via an API • The Register

0

In short OpenAI has released its new and improved version of Codex, its AI code completion model, to beta testers via an API.

Co-founders Greg Brockman and Ilya Sutskever demonstrated the capabilities of the Codex during a livestream broadcast on Twitch earlier this week. They showed that the system was able to generate Python code for simple tasks such as printing and formatting text, from simple English instructions.

If you want to verify that the Codex is correct then, yes, you will need to know how to write Python. Codex is the backbone of GitHub Copilot, a paired programming software that GitHub originally launched a few months ago.

The AI ​​model has been described as “a descendant of GPT-3” by OpenAI. GPT-3 was trained on text taken from the Internet and excels at natural language tasks. In order to specialize it in code, however, engineers had to tweak it on billions of lines of code copied from GitHub.

“OpenAI Codex is the best performer in Python, but it also masters over a dozen languages, including JavaScript, Go, Perl, PHP, Ruby, Swift and TypeScript, and even Shell. It has 14KB of memory for it. Python code, compared to GPT-3 which is only 4KB – so it can take in over 3 times the contextual information when performing any task, ”said startup AI .

You can register to access the API here.

Samsung is designing its own AI chips for its future smartphones

South Korean tech giant Samsung has confirmed that new local hardware will allow machine learning algorithms to run on its Exynos chipset for smartphones.

The kit was designed using software from Synopsys AI, Wired first reported. Not much is known about the chip’s specs or whether they have entered production yet, but there is more than enough competition in the AI ​​smartphone circuit.

“What you see here is the first of a true commercial processor design with AI,” exclaimed Aart de Geus, President and Co-CEO of Synopsys.

The news follows Google’s announcement that it has also built its own AI chips for its upcoming Pixel 6 and Pixel 6 Pro handsets coming later this year. Building accelerators optimized for a set of custom AI models seems to be becoming the norm as smartphones increasingly use machine learning techniques to process images etc.

Southwest Institute Engineers Get $ 34 Million To Build Self-Contained Humvees For The U.S. Army

The Southwest Institute nonprofit research organization has won a $ 34 million contract to help build heavy-duty autonomous vehicles for the U.S. military.

“SwRI is proud to continue to develop the latest autonomous and robotic systems for the US military,” Joseph Hernandez, a senior engineer working at the institute, who manages the program, noted in a report.

“This contract offers a more direct link with our end customers and allows us to provide more innovative solutions to support our fighters.

Engineers will develop the software needed to autonomously operate high mobility multipurpose wheeled vehicles, more commonly known as Humvees. SwRI has contracted with the U.S. military to help them understand how to integrate sensors and machine learning algorithms into cars and drones for over a decade.

Funding of $ 34 million will be distributed over five years. ®


Source link

Leave A Reply

Your email address will not be published.