Artificial Intelligence (AI) is one of the most discussed topics in the world of technology. It is also one of the most promising tools that people hope will improve our lives. For many, AI is something unbelievable and difficult to understand. However, we utilize AI in our day-to-day life in many ways without even knowing about it. According to statistics, only 33% of people know they use AI in their daily life.

By now, it will have been difficult to avoid the screenshots of the flirty messages Adam Levine sent to his fans - and the resulting memes. The Maroon 5 frontman is married to model Behati Prinsloo, with whom he has two children, and she is currently pregnant with their third. But, over the last week, five women have come forward accusing 43-year-old Levine of sending them explicit messages in recent years. The first was Instagram model Sumner Stroh, 23, who uploaded a bombshell TikTok video in which she claimed the two had had an affair last year and revealed a slew of flirtatious DMs. Then comedian Maryka and a woman named Alyson Rose shared screenshots of their alleged Instagram conversations with the singer. Levine's former yoga teacher, Alanna Zabel, next accused him of sending a dirty text to her, before fitness influencer Ashley Russell claimed he started messaging her on social media this year.

Machine learning (ML) is helping companies remain competitive. In fact, many companies' core business today is based on machine learning and image/speech recognition. Google, for example, uses machine learning in image recognition for Google Photos and speech recognition for Google Home and Google Assistant. Millions of people talk to Siri, Apple's virtual assistant. The company extended the application of its virtual assistant through HomePod, a smart home device.

These AI systems create images based on instructions or directions. They will, if told to, create an image of a kiwi bird eating a kiwi fruit while sitting on a big padlock key. They can be used to create ads, fashion designs or movie production storyboards. DALL-E, Midjourney and Wombo Dream are examples of AI image generators. AI can also create three-dimensional spaces and objects, both real and digital. It can design buildings, rooms and even whole city plans, as well as virtual spaces for gameplay or metaverse-style collaboration.

We present a new method based on episodic Knowledge Graphs (eKGs) for evaluating (multimodal) conversational agents in open domains. This graph is generated by interpreting raw signals during conversation and is able to capture the accumulation of knowledge over time. We apply structural and semantic analysis of the resulting graphs and translate the properties into qualitative measures. We compare these measures with existing automatic and manual evaluation metrics commonly used for conversational agents. Our results show that our Knowledge-Graph-based evaluation provides more qualitative insights into interaction and the agent's behavior.

AI is taking the world by storm. While it was once a thing of sci-fi movies, it's no longer fiction. In fact, AI--tech that can think, learn, and make autonomous decisions--is now seeping its way into our lives. Think of self-driving cars (Tesla), navigation (Google Maps), or even virtual assistants (Siri): all of them rely heavily on AI. You might have also heard about the AI-powered robot from Boston Dynamics that can do a summersault, a handstand, or even a split leap that's now making a lot of noise online.

The need for improvement has been changing the world for thousands of years. Numerous inventions have filled gaps with intended and unintended uses in response to this desire for advancement. Take the wheel, for example. It was developed to assist potters with their clay back in Mesopotamia around 3500 B.C. (Gambino, 2009). Now, we get thousands of different uses out of the wheel.

Graph Convolutional Networks (GCNs) and their variants have achieved significant performances on various recommendation tasks. However, many existing GCN models tend to perform recursive aggregations among all related nodes, which can arise severe computational burden to hinder their application to large-scale recommendation tasks. To this end, this paper proposes the flattened GCN~(FlatGCN) model, which is able to achieve superior performance with remarkably less complexity compared with existing models. Our main contribution is three-fold. First, we propose a simplified but powerful GCN architecture which aggregates the neighborhood information using one flattened GCN layer, instead of recursively. The aggregation step in FlatGCN is parameter-free such that it can be pre-computed with parallel computation to save memory and computational cost. Second, we propose an informative neighbor-infomax sampling method to select the most valuable neighbors by measuring the correlation among neighboring nodes based on a principled metric. Third, we propose a layer ensemble technique which improves the expressiveness of the learned representations by assembling the layer-wise neighborhood representations at the final layer. Extensive experiments on three datasets verify that our proposed model outperforms existing GCN models considerably and yields up to a few orders of magnitude speedup in training efficiency.

Sparsity of the User-POI matrix is a well established problem for next POI recommendation, which hinders effective learning of user preferences. Focusing on a more granular extension of the problem, we propose a Joint Triplet Loss Learning (JTLL) module for the Next New ($N^2$) POI recommendation task, which is more challenging. Our JTLL module first computes additional training samples from the users' historical POI visit sequence, then, a designed triplet loss function is proposed to decrease and increase distances of POI and user embeddings based on their respective relations. Next, the JTLL module is jointly trained with recent approaches to additionally learn unvisited relations for the recommendation task. Experiments conducted on two known real-world LBSN datasets show that our joint training module was able to improve the performances of recent existing works.

Cold start is an essential and persistent problem in recommender systems. State-of-the-art solutions rely on training hybrid models for both cold-start and existing users/items, based on the auxiliary information. Such a hybrid model would compromise the performance of existing users/items, which might make these solutions not applicable in real-worlds recommender systems where the experience of existing users/items must be guaranteed. Meanwhile, graph neural networks (GNNs) have been demonstrated to perform effectively warm (non-cold-start) recommendations. However, they have never been applied to handle the cold-start problem in a user-item bipartite graph. This is a challenging but rewarding task since cold-start users/items do not have links. Besides, it is nontrivial to design an appropriate GNN to conduct cold-start recommendations while maintaining the performance for existing users/items. To bridge the gap, we propose a tailored GNN-based framework (GPatch) that contains two separate but correlated components. First, an efficient GNN architecture -- GWarmer, is designed to model the warm users/items. Second, we construct correlated Patching Networks to simulate and patch GWarmer by conducting cold-start recommendations. Experiments on benchmark and large-scale commercial datasets demonstrate that GPatch is significantly superior in providing recommendations for both existing and cold-start users/items.