AMIS African Master's Degree in Machine Intelligence provides brilliant young Africans with state-of-the-art training in machine learning.

While OpenAI's GPT-3 system has proved to be remarkably effective at many tasks, including question-answering (QA), it is still out of reach for many organizations, being only available to approved users for a fee. While there are a few other pretrained QA systems available, none has quite matched GPT-3's few-shot QA performance -- until now. AI2 has just released Macaw (multi-angle question-answering), a versatile, generative question-answering (QA) system that exhibits strong zero-shot performance on a wide range of question types. On a suite of 300 challenge questions, Macaw outperformed GPT-3 by over 10%, even though Macaw is an order of magnitude smaller (11 billion vs. 175 billion parameters). Even better, Macaw is publicly available for free.

In recent years, artificial intelligence has become mainstream. Today, more enterprises in more industries are looking to technology like AI for a competitive edge and to help them manage the overwhelming volumes of data they both generate and collect. A recent study by IBM put the "global AI adoption rate" at 35% and found that 44% of global organizations are currently working on embedding AI into their operations. But despite this continued uptick in AI adoption, one element sure to put the brakes on many initiatives is the global skills shortage. Successful AI requires competent data scientists.

A mathematician by trade, Molei Tao, typically uses mathematics to design algorithms and solve physical science problems like how planets move. Recently, he became attracted to machine learning, an area that according to him, contains numerous interesting problems that are mathematically exciting and can benefit from modern mathematical tools. This year, Tao published his first machine learning conference paper, and this work was awarded the best paper award at the 23rd International Conference on Artificial Intelligence and Statistics (AISTATS). In flat spaces, the approach of adding momentum for accelerating the training of machine learning models has already been tremendously successful, and this new progress expands the applicability of the popular and powerful idea. Tao felt fortunate to win this recognition.

Science and technology are developing at a very fast pace and many new techniques are incorporated to make several tasks easier. There are several newer techniques used for educating the masses and one of them is Artificial Intelligence (AI). Artificial Intelligence enables machines and computers to mimic the capabilities of the human brain in decision-making and problem-solving. There are other similar terms used that include machine learning and deep learning. These terms are frequently used interchangeably, but they are not the same.

What is another way to re-cast this? In the case of Image segmentation, we can think of it as an optimization problem. For any image, you can choose and iterate between a variety of segment configurations, and choose the one where the difference between pixel intensities is the highest. Learning problems are trained by minimizing the error between predictions and actual values across a large data set. In other words, they are trained by optimization methods that are stochastic. Learning and optimization problems have many things in common structurally.

Camp 2 Data Science Academy (DSA) was held online from August 8, 2022 to August 13, 2022 via the Zoom platform. This camp was mentored by Rendi Chevi, an AI Research Scientist from kata.ai. Curious about the excitement of Camp 2 of DSA? Let's see what it's all about! The first day began with remarks from the MCs, followed by a presentation by Rendi Chevi about "Introduction to Machine Learning". First of all, Rendi gave an overview of Machine Learning along with a Roadmap of what will be discussed during the second camp of DSA, then Rendi gave a Venn Diagram of the relationship between Machine Learning and Artificial Intelligence which gave rise to a sub-section, namely Deep Learning.

Google has announced the expansion of content advisory for searches where its AI (Artificial Intelligence) system does not have high confidence in the overall quality of the results available for search. Google Search Vice President Pandu Nayak said that this does not mean that there is no useful information available or that a particular result is of low quality. "These notices provide context about the full set of results on the page, and you can always see the results of your query, even if the advice is present," he said in a blog post. Nayak also said in a blog post that Google has invested deeply in both information quality and information literacy for Search and News, and today we have some new developments regarding this important task. Google has also introduced the latest AI model called Multitask Unified Model (MUM) to improve search result quality in'snippets' that are shown at the top of the page for search.

We propose and demonstrate a new approach for fast and accurate surrogate modelling of urban drainage system hydraulics based on physics-guided machine learning. The surrogates are trained against a limited set of simulation results from a hydrodynamic (HiFi) model. Our approach reduces simulation times by one to two orders of magnitude compared to a HiFi model. It is thus slower than e.g. conceptual hydrological models, but it enables simulations of water levels, flows and surcharges in all nodes and links of a drainage network and thus largely preserves the level of detail provided by HiFi models. Comparing time series simulated by the surrogate and the HiFi model, R2 values in the order of 0.9 are achieved. Surrogate training times are currently in the order of one hour. However, they can likely be reduced through the application of transfer learning and graph neural networks. Our surrogate approach will be useful for interactive workshops in initial design phases of urban drainage systems, as well as for real time applications. In addition, our model formulation is generic and future research should investigate its application for simulating other water systems.

Practising and honing skills forms a fundamental component of how humans learn, yet artificial agents are rarely specifically trained to perform them. Instead, they are usually trained end-to-end, with the hope being that useful skills will be implicitly learned in order to maximise discounted return of some extrinsic reward function. In this paper, we investigate how skills can be incorporated into the training of reinforcement learning (RL) agents in complex environments with large state-action spaces and sparse rewards. To this end, we created SkillHack, a benchmark of tasks and associated skills based on the game of NetHack. We evaluate a number of baselines on this benchmark, as well as our own novel skill-based method Hierarchical Kickstarting (HKS), which is shown to outperform all other evaluated methods. Our experiments show that learning with a prior knowledge of useful skills can significantly improve the performance of agents on complex problems. We ultimately argue that utilising predefined skills provides a useful inductive bias for RL problems, especially those with large state-action spaces and sparse rewards.