It can take years to learn to write computer code properly. Source, the first in Paris, thinks that programming should not be such a big deal.
The company is well-prepared for the tool it uses artificial intelligence write the number based on a brief description of what the number should do. Tell the company’s tool to “multiply the two numbers provided by the user,” for example, and it strikes twelve lines in Python to do the same.
SourceAI goals are an indication of major changes in software design. Advances in learning technology have led to a number of writing tasks, ranging from finalizing code-making and algorithming in such a way that you can track a starting number and detect dangerous bugs.
Installing a flexible machine can change the size of the software, but the limitations and blind spots of modern AI can bring new challenges. Machine learning methods can perform unexpectedly, and machine-generated machines can contain harmful bugs only if they are carefully monitored.
SourceAI, and other similar programs, are looking to take advantage of GPT-3, a powerful AI-speaking program announced in May 2020 and OpenAI, a San Francisco-based company focused on advancing AI. The founders of SourceAI were among the first hundred people to gain access to GPT-3. OpenAI did not release the GPT-3 code, but allows other users to access the format via the API.
GPT-3 is a large neural network that is trained in the main components extracted from the Internet. It does not understand the meaning of the text, but it is able to draw out the language well enough to make notes on a given topic, to summarize the text, or to answer questions about the text.
“By testing this tool, we realized that it can generate code,” says Furkan Bektes, founder of SourceAI and CEO. “That’s when we had the idea to create a SourceAI.”
He was not the first to realize the potential. Once GPT-3 is released, one developer showed that it can make traditional programs, including buttons, text entries, and colors, by changing the pieces of code that were fed. The company has, Make, Plans to sell technology.
SourceAI seeks to allow users to create multiple programs in multiple languages, thus helping to create more programs. “Developers should keep track of writing time, while people who don’t have the experience to write can also create programs,” says Bektes.
The company has, TabNine, used the old version of the OpenAI language, GPT-2, released by OpenAI, to create a tool that allows you to complete a line or task once the developer has started typing.
Some software giants seem to be having fun with it. Microsoft invested $ 1 billion in OpenAI in 2019 and has agreed to grant the GPT-3 license. On giant programs Build a meeting in May, Sam Altman, founder of OpenAI, showed how GPT-3 can fulfill your number to the developer. Microsoft has refused to comment on how to use AI in its software development tools.
Brendan Dolan-Gavitt, an assistant professor in the Computer Science and Engineering department at NYU, says language models such as GPT-3 can be used to help software readers. Some marketers will use these colors to “recognize the flaws that can occur when you write them, by looking at things that are ‘amazing’ in the language,” he says.
Using AI to create and analyze code can be challenging, however. In an online paper in March, researchers at MIT demonstrated that AI-trained authentication program the code can run smoothly can be tricked and make small changes, such as changing other colors, to make a bad program. Shashank Srikant, a PhD student involved in the project, says the types of AI should not be relied upon too much. “Once these species start to form, things can quickly get worse,” he says.
Dolan-Gavitt, a professor at NYU, says that the variety of languages used in the development of registration tools also poses problems. “I think the use of language forms can create a trailer and instability,” he says. Apart from that, they are taught about human literature, which is often small and insecure. ”
