Explainable NLP (ExNLP) has increasingly focused on collecting human-annotated explanations. These explanations are used downstream in three ways: as data augmentation to improve performance on a predictive task, as a loss signal to train models to produce explanations for their predictions, and as a means to evaluate the quality of model-generated explanations. In this review, we identify three predominant classes of explanations (highlights, free-text, and structured), organize the literature on annotating each type, point to what has been learned to date, and give recommendations for collecting ExNLP datasets in the future.

GCHQ published a paper called Ethics of AI: Pioneering a New National Security which explains why the technology – enabling problem-solving at scale and speed – will be at the heart of our mission to keep the country safe in an increasingly complex world. The paper also details how GCHQ will ensure they use AI fairly and transparently, applying existing tests of necessity and proportionality. This includes establishing an AI ethical code of practice to recruiting more diverse talent to help develop and govern our use of AI, protecting privacy and striving for systematic fairness. Source: GCHQ to use AI to tackle child sex abuse, disinformation… – GCHQ.GOV.UK

I have been an ML engineer for over 2 years in a US based company in India. Working in this company(service based) I saw my whole life playout: Senior MLE in a year, Solution Architect in a couple more, and finally leading a project in a couple more years with proportional increase in pay of course. Seeing my next 5-7 years pan out this way I suddenly realized that this wont be sufficient to satisfy the intellectual in me and also made me realize how much I am actually interested in research (I have a paper published in IEEE related to Deep Learning and Instrumentation) and how much I enjoy making something new. So in short, as my first priority I am looking for something research driven. Since I am research/innovation driven, post MS I will be looking for a job in a research lab (or a research wing in a company).

The possibility always exists, even if we neither accept it nor appreciate it! There is a reason to develop and that reason takes us ahead towards the impossible mission with speed and higher abilities. The idea of deep learning can be a reality since science has started walking towards it with the goal that it will be achieved. Nothing stands as impossible forever if we as professionals aim still higher during each passing semester. The fruit will be sweet of deep learning!

The deepfake technology this TikToker is using to impersonate Tom Cruise, apparently (via The Daily Beast). The deeptomcruise account appears to feature someone who's pretending to be the famous actor, and the impression isn't just jaw-dropping because of how well the person replicates Cruise's famous laugh. How can you tell it's not actually Tom Cruise using some sort of strange Twitter name? First, his voice is close, but it's not quite there, especially since the lip-syncing is a tiny bit off at some points. You can hear what the real Cruise sounds like in this video (unless, of course, James Corden also deepfaked him in).

Wattpad uses a machine-learning program called StoryDNA to scan all the stories on its platform and surface the ones that seem like candidates for …

Reinforcement learning (RL) is the area of machine learning that is concerned with how software is able to take the right decision.

Intellectual property is the lifeblood of entertainment, and Hollywood badly needs a transfusion. Stuck recycling the same ideas and producing the same franchises, Hollywood is now looking to podcasts, news articles, and even songs and paintings to dredge up stories that could be hits with global audiences. Finding the right stories--and developing them into things that actually exist on screen--can be an arduous, money-draining process. One company thinks it has found an efficient new way to sift through millions of user-generated stories to find the next Hollywood hits: artificial intelligence. And, on occasion, it actually brings fans of those stories into the arcane process of developing them for the screen.

In collaboration with Sorbonne University, Facebook AI introduced a new benchmark for continual learning (CL). This presents excellent means to improve traditional machine learning (ML) methods by training AI models to mimic the way humans learn new tasks. The team addresses one of the most vital bottlenecks associated with modern AI systems: their reliance on massive human-labeled datasets. Humans can learn from past experiences but applying the same to AI is not easy. Generally, ML systems are trained in isolation, thus require specific models for each new task.

Three days ago, a TikTok account going by @deeptomcruise began posting video clips of the Hollywood actor Tom Cruise doing everything from golfing, to tripping and telling a joke in what appears to be a men's clothing store in Italy, to performing a magic trick with a coin. In each of the three videos, Cruise delivers his signature maniacal laugh--you know, the one he repeatedly unleashed in that batty Scientology recruitment video years back--before launching into some sort of bit, and in all of them, it looks just like Cruise. There are a few giveaways, of course. Also, his voice is hollow and scratchy, a la that scene in Face/Off where John Travolta-as-Nicolas Cage is trying to adjust his vocals to that Cage-ian timbre. Still, the Cruise TikToks managed to bewilder and horrify a number of people.