Here’s an interesting article in Nature about the use of AI in evaluating embryos with AI — another use of computer vision in the medical field. Could this bring down healthcare costs? What if the algorithm mislabels an embryo? Are there ethical implications?

Deep learning algorithms, in particular convolutional neural networks (CNNs), have recently been used to address a number of medical-imaging problems, such as detection of diabetic retinopathy,18 skin lesions,19 and diagnosing disease.20 They have become the technique of choice in computer vision and they are the most successful type of models for image analysis. Unlike regular neural networks, CNNs contain neurons arranged in three dimensions (i.e., width, height, depth). Recently, deep architectures of CNNs such as Inception21 and ResNet22 have dramatically increased the progress rate of deep learning methods in image classification.23 In this paper, we sought to use deep learning to accurately predict the quality of human blastocysts and help select the best single embryo for transfer (Fig. 1).

Here’s an interesting piece in Forbes on how AI will transform the way we conduct and audit business.

“One of the best-use cases for AI is to look at information and to identify patterns that stand out,” he says, adding that this applies just as readily to looking for unethical behavior as it does to recognizing cat pictures. “As long as there’s data related to it and you can analyze it at scale, AI can detect anomalies that humans simply can’t.”

OpenAI raised some eyebrows last month when it announced it had figured out a way to get an AI to write more naturally. They, however, decided not to release their entire research for fear that it could cause havoc.

From an article in The Register.

Last month, researchers at OpenAI revealed they had built software that could perform a range of natural language tasks, from machine translation to text generation. Some of the technical details were published in a paper, though the majority of materials was withheld for fear that it could be used maliciously to create spam-spewing bots or churn out tons of fake news. Instead, OpenAI released a smaller and less effective version nicknamed GPT-2-117M.

As a machine learning project grows, so should its infrastructure. In this talk, Alejandro Saucedo covers some of the key trends in machine learning operations, as well as libraries to watch in 2019.

The talk is based on the “Awesome Machine Learning Operations” list maintained by The Institute for Ethical AI & Machine Learning, and focuses on the topics of reproducibility, orchestration and explainability.