👾 ML that Improves Code Writing
Weekly news digest curated by the industry insiders
📝 Editorial
The idea of machine learning (ML) models that can create production-ready code has been one of the iconic use cases in the history of AI. A few months ago, the AI world was astonished by the release of models like OpenAI’s Codex and DeepMind’s AlphaCode that can generate sophisticated text from natural language. These types of models are an important step for code generation, and they have opened the door to all sorts of research about using ML models to improve the process of writing software.
The lifecycle of code in software applications goes beyond authoring. Typically, developers in peripheral tasks such as testing or bug fixing. This is an area that ML can help too. Just this week, there were several research contributions to the ML-first code space. Microsoft Research unveiled Jigsaw, a model that can fix bugs in the code generated by large language-to-code models. Similarly, Stanford University published details about a model used to evaluate coding tests. Salesforce is also jumping into the ML code generation space with CodeGen, a mode that can generate code from language instructions. From code generation to testing to bug fixing, ML is rapidly penetrating all areas of software development.
🔺🔻 TheSequence Scope is our Sunday free digest. To receive high-quality educational content about the most relevant concepts, research papers, and developments in the ML world every Tuesday and Thursday, please subscribe to TheSequence Edge 🔺🔻
🗓 Next week in TheSequence Edge:
Edge#179: we do a recap of GANs series;
Edge#180: a deep dive into SuperAnnotate, an annotation platform for real-world scenarios.
Now, let’s review the most important developments in the AI industry this week
🔎 ML Research
Generating Code from Natural Language
Salesforce Research published a paper introducing CodeGen, an autoregressive language model that turns language instructions into code →read more on Salesforce blog
Grading Coding Assignments
Stanford University published a paper proposing a reinforcement learning model that can grade coding assignments particularly related to coding games →read more on Stanford AI Lab blog
Fixing Bugs in AI-Generated Code
Microsoft Research published a paper discussing Jigsaw, a tool that can fix bugs in programs generated by large language models →read more on Microsoft Research blog
Improving Wikipedia Content
Meta (Facebook) AI Research (FAIR) published a paper proposing a method to provide more inclusive content in Wikipedia →read more on FAIR blog
🤖 Cool AI Tech Releases
DataRobot Cloud 8.0
DataRobot announced the release of AI Cloud 8.0, a new release with a strong focus on predictive models and continuous ML →read more in DataRobot’s press release
Arize AI vNext
Arize AI released a new version of its ML observability platform →read more in Arize AI’s press release
🗯 Interesting Tweet
🛠 Real World ML
GNNs at Amazon
Amazon Research published a blog post detailing their use of gran neural networks across different scenarios →read more on Amazon Research blog
TensorFlow.js at LinkedIn
LinkedIn published a detailed blog post about their use of TensorFlow.js to customize the performance of web and mobile apps →read more on LinkedIn Engineering blog
💸 Money in AI
Data/AI/ML
Search infrastructure company Pinecone Systems Inc. raised a $28 million Series A funding round led by Menlo Ventures. Hiring in Tel Aviv/Israel and New York, NY/US.
Тo-code machine learning infrastructure company Black Crow AI raised a $25 million Series A round led by Imaginary Ventures. Hiring remote.
AI-powered
Online furniture marketplace Kaiyo raised $36 million Series B funding round comprised of equity and debt, led by Edison Partners. Hiring in New York, NY/US.
Revenue intelligence startup BoostUp.ai raised $28.5 million in a Series B funding round led by NGP Capital. Hiring in California/US and remote.
Building design platform BeamUP raised $15 million in a seed funding round led by StageOne Ventures. Hiring in Tel Aviv/Israel.