Amazon recently unveiled a new warehouse robot that can “feel,” according to the company.  Called the Vulcan robot, the two-armed beast is already working in Amazon fulfillment centers in Spokane, WA, and Hamburg, Germany, where it’s handled more than half a million orders. The Vulcan’s sense of touch comes from force-sensitive grippers and sensors on its joints, which provide data to its AI about the edges, contours, and resistance of the items it picks up, moves, and places in a new location. With this faculty, Vulcan can grip a soft bag of candy gently, but a heavy coffee table book more firmly without crushing the candy or dropping the book.  One way to look at the benefit of a robot that can detect what it’s gripping is that it provides flexibility, enabling a wider range of products to be handled. According to Amazon, Vulcan can handle about 75% of the million or so products in a typical Amazon warehouse. Aaron Parness and his roughly 250-person robotics team at Amazon created the Vulcan robot. The company revealed it to the world on May 7 at its “Delivering the Future” event in Dortmund, Germany.  Mixed feelings about tactile robots Vulcan is the first Amazon robot with what the company calls a “genuine sense of touch,” thanks to force feedback sensors and AI-powered software that lets it “feel” items, not just see them (cameras are the main sensors on most Amazon robots).  If you’re familiar with my views on anthropomorphicization (i.e., “humanizing”) of robots and AI, you might guess what I’m about to say. Amazon says its Vulcan robot can feel (without quotation marks). This isn’t true.  When a robot like Vulcan “feels” something, it uses sensors that measure force, pressure, and sometimes texture or shape, turning these signals into data that AI can interpret. Vulcan’s sensors are built into its gripper and joints, so when it touches or grasps an object, it detects how much force it’s applying and the contours it’s encountering. Machine learning algorithms then help Vulcan decide how to adjust its grip or movement based on this feedback. By contrast, a person feels with a network of millions of nerve endings in the skin, especially in the fingertips. These nerves send detailed, real-time information to the brain about pressure, temperature, texture, pain, and even the direction of force. The human sense of touch is deeply connected to memory, emotion, judgment, and consciousness. Robots like Vulcan can now match or even exceed humans in detecting pressure or identifying textures, but their “feeling” is purely mechanical and digital. They don’t experience sensation or emotion, and they only know what their sensors can measure and their software can interpret. Humans, on the other hand, feel in a way that’s physical, emotional, and conscious. With that caveat out of the way, it has to be said that Vulcan is pretty amazing.  Giving warehouse workers a hand Parness says Vulcan’s sense of touch is a breakthrough because it brings “physical intelligence” to robots, and that’s the categorical breakthrough here. The main advance robots will undergo over the next decade or two will involve sensors increasingly being able to detect and adapt to real conditions in the real world and being trained in virtual physical AI environments.  Despite the advanced state of Vulcan’s technology, its main job is actually limited:  to pick products from bulk storage and pack them into movable shelves, a task that used to require human dexterity.  Vulcan is neither a humanoid robot nor is its hand modeled after human hands. I’ve weighed in before about how modeling factory robots after human body parts makes no sense.  Vulcan’s “hand” combines a conveyor belt gripper with a spatula-like tool, both of which are fitted with sensors that constantly measure pressure and torque. A ruler-like tool attached between the paddles acts as a spatial guide, nudging existing items in storage bins to create space for the new items it is laying into place.  The robot’s AI, trained on thousands of hours of physical interaction data, calculates the right amount of force for each object in real time. Vulcan can work up to 20 hours a day without ever taking a coffee break or using the bathroom, moving at speeds comparable to a human worker, but always behind a safety fence in case it suddenly goes bananas like that Chinese robot in the TikTok video people have been talking about and misinterpreting. (It just malfunctioned. It didn’t try to “attack” people.)  The Vulcan robot has real limitations. For example, it’s too weak to lift anything heavier than 8 pounds and can only move products from one place to another. If the robot encounters an unfamiliar item or something that exceeds its 8-pound weight limit, it flags a human worker for help. Grasping the importance of tactile robots Vulcan isn’t the only robot with a sense of touch. RoboTact and RoboTouch sensors, developed over decades and now used in everything from humanoid robots to service bots, give machines the ability to sense contact, pressure, and even the shape of objects, allowing for delicate and precise handling.  Sanctuary AI enhanced its Phoenix robot with tactile sensors that let it handle complex, touch-driven jobs. Their technology means Phoenix can detect things like slippage or excessive force, even when it can’t see what it’s doing.  Meta created its Digit 360 sensor, a fingertip-shaped device that can register forces as tiny as one millinewton and pick up details down to seven microns. The sensor is still in the lab, but Meta’s partnership with GelSight and Wonik Robotics is helping to eventually bring these sensors into real-world use.  Eventually, robots with a sense of touch will revolutionize robotics and enable them to perform a wide range of tasks that robots currently cannot do, like picking strawberries, performing surgery, and defusing bombs. But for now, Amazon is using some of the most advanced tactile robots on the planet to help ship bathrobes, books, and batteries to customers.

Macworld At a glanceExpert's Rating Pros Robust Construction Effective fan Good data transfer rates Informative display Cons Not many extra ports Fan not silent No carrying case Our Verdict Anyone who opts for the Smart USB-C Hub 7-in-1 will do so mainly because of the option to integrate an NVMe SSD. Price When Reviewed This value will show the geolocated pricing text for product undefined Best Pricing Today Price When Reviewed129 Dollar Best Prices Today: Dockcase Smart USB-C Hub 7-in-1 with M.2 SSD Enclosure Retailer Price Dockcase 129,00 € View Deal Price comparison from over 24,000 stores worldwide Product Price Price comparison from Backmarket At first glance, what stands out most about the Smart USB-C Hub 7-in-1 with M.2 SSD Enclosure hub is its relatively large case with an equally large display. The 4.3 by 2.3 centimeter display shows, among other things, the connected devices and the current data transfer rate. It also serves as a touchscreen for settings such as data protection and fan speed. The large display looks playful, but is very informative. An integrated SSD health monitor issues early warnings. Speaking of that integrated SSD, inside the hub is a connector for an M.2 SSD module that you provide. It supports the compact 2230 and 2242 NVMe SSDs with a maximum capacity of 2TB. An optional read-only mode prevents unwanted changes or deletions and thus protects sensitive data, if required. An integrated, temperature-controlled fan and ventilation openings ensure efficient cooling of the NVMe SSD. Installtion an SSD module is effortless and Dockcase provides a small screwdriver. The ports on the Smart USB-C Hub are one USB-C and USB-A (both USB 3.2 Gen 2 at 10GBps), a 100W Power Delivery USB-C power, an HDMI 2.1 port with support for 4K at 120Hz, an SD card slot, and a TF port. If you want a hub that gives you more connectors, other hubs and docks are better suited in our Best USB-C hubs and adapters for Mac 2025 roundup. Not everyone needs a lot of extra ports; primarily, this is an external SSD enclosure with a few useful ports. The data transfer rate of the integrated NVMe SSD impresses across the board in the AJA System Test Lite with 875MBs for writing and 955MBps for reading. The fan occasionally starts up in the read/write endurance test; it is relatively quiet, but not silent. In our endurance test, however, the small NVMe SSD can get very warm after more than 30 minutes of intensive use. The write rate then decreases significantly. However, this is a respectable performance for a USB-C hub. With the fan deactivated for the test (the temperature was read via the internal sensor), the temperature rose during continuous operation of the SSD. After a short time, a warning was displayed at 69 degrees Celsius, and the data transfer rate dropped significantly in some cases. At 72 degrees Celsius, we activated the fan to protect the SSD and within a short time, the temperature dropped to below 60 degrees Celsius. The full metal housing is robust and also ensures good temperature equalization thanks to the many ventilation slots. However, the fan can be noisy and annoying to sensitive ears. Dockcase provides a fan control with four sppeds or the ability to turn it off completely–but doing so puts the SSD at risk. With an OWC Ultra SDXC memory card, we measured 253MBps for writing and 284MBps for reading–an excellent performance. with a Lexar Armor 700 USB-C SSD, the 890MBps for writing and 939MBps for reading are only aabout 100MBps below the values directly on the MacBook Air M3. Should you buy the Dockcase Smart USB-C Hub 7-in-1 with M.2 SSD Enclosure? Anyone who opts for the Smart USB-C Hub 7-in-1 will do so mainly because of the option to integrate an NVMe SSD. If you need an external storage device and could use card reader slots and an extra charging port, the Smart USB-C hub may work. But there are plenty of other options out there if you want to add more ports.

A Northeastern University student demanded her tuition money back after discovering her business professor was secretly using AI to create course materials. Ella Stapleton, who graduated this year, grew suspicious when she noticed telltale signs of AI generation in her professor's lecture notes, including a stray ChatGPT citation in the bibliography, recurring typos matching machine outputs, and images showing figures with extra limbs. "He's telling us not to use it, and then he's using it himself," Stapleton told the New York Times. After filing a formal complaint with Northeastern's business school, Stapleton requested a tuition refund of about $8,000 for the course. The university ultimately rejected her claim. Professor Rick Arrowood acknowledged using ChatGPT, Perplexity AI, and presentation generator Gamma. "In hindsight, I wish I would have looked at it more closely," he said. Read more of this story at Slashdot.

Credit: Leon Neal/Getty Images TL;DR: Unblock Pornhub from North Carolina with a VPN. The best service for unblocking porn sites is ExpressVPN.More than a third of U.S. states have introduced restrictions for online adult content, including North Carolina. After the introduction of age verification laws, millions of users in North Carolina are now unable to access popular sites like Pornhub. There are complicated reasons for this restriction, but for those left in the dark, there is a simple workaround.If you want to unblock porn sites like Pornhub for free from North Carolina, we have all the information you need. You May Also Like How to unblock Pornhub for free in North CarolinaVPNs are useful tools that can hide your real IP address (digital location) and connect you to secure servers in other locations. This straightforward process bypasses geo-restrictions so you can access sites like Pornhub from anywhere in the world. Unblock Pornhub by following these simple steps: Mashable Trend Report: Coming Soon! Decode what’s viral, what’s next, and what it all means. Sign up for Mashable’s weekly Trend Report newsletter. By clicking Sign Me Up, you confirm you are 16+ and agree to our Terms of Use and Privacy Policy. Thanks for signing up! Sign up for a VPN (like ExpressVPN)Download the app to your device of choice (the best VPNs have apps for Windows, Mac, iOS, Android, Linux, and more)Open up the app and connect to a server in a location that supports access to PornhubVisit Pornhub Opens in a new window Credit: ExpressVPN ExpressVPN (2-Year Subscription + 4 Months Free) $139 only at ExpressVPN (with money-back guarantee) The best VPNs for unblocking porn sites are not free, but most do offer free-trial peiods or money-back guarantees. By leveraging these offers, you can unblock porn sites like Pornhub without actually spending anything. This is obviously not a long-term solution, but it does give you the opportunity to temporarily retain access to Pornhub before recovering your investment.If you want to retain permanent access to sites like Pornhub, you'll need a subscription. Fortunately, the best VPN for bypassing content restrictions is on sale for a limited time.What is the best VPN for Pornhub?ExpressVPN is the top choice for unblocking porn sites like Pornhub, for a number of reasons: Servers in 105 countries Easy-to-use app available on all major devices including iPhone, Android, Windows, Mac, and moreStrict no-logging policy so your data is always secureFast streaming speeds free from throttlingUp to eight simultaneous connections30-day money-back guaranteeA two-year subscription to ExpressVPN is on sale for $139 and includes an extra four months for free — 61% off for a limited time. This plan also includes a year of free unlimited cloud backup and a generous 30-day money-back guarantee.Unblock Pornhub for free in North Carolina with ExpressVPN. Joseph Green Global Shopping Editor Joseph Green is the Global Shopping Editor for Mashable. He covers VPNs, headphones, fitness gear, dating sites, streaming, and shopping events like Black Friday and Prime Day.Joseph is also Executive Editor of Mashable's sister site, AskMen.

Elissa Coulon / « Le Monde » « S’il te plaît, ChatGPT » : pourquoi est-on poli avec les IA ? Publié aujourd’hui à 07h00 Article réservé aux abonnés DécryptageUne grande partie des utilisateurs font preuve de politesse en s’adressant à ChatGPT, Siri, Alexa et autres agents conversationnels. A tort ou à raison ? « Je ne saurais pas l’expliquer, mais je me sens mal si je ne dis pas “s’il te plaît” à ChatGPT… Ça me paraît impoli compte tenu du service que l’IA me rend. » Pauline Blanquet a 25 ans. Comme elle, de nombreux utilisateurs font preuve de courtoisie avec ChatGPT, Siri, Alexa et autres agents conversationnels. Ils seraient même majoritaires : en 2019, le Pew Research Center estimait que 54 % des propriétaires américains d’enceintes connectées leur disaient « s’il te plaît » ; plus récemment, un sondage du groupe de presse tech Future évaluait à 67 % la proportion d’Américains polis avec des chatbots. « Je sais pourtant qu’il ne s’agit que d’un robot, reconnaît en soupirant la jeune Boulonnaise, mais je ne peux pas m’en empêcher. » Comment l’expliquer, alors qu’il ne lui viendrait pas à l’idée de remercier d’autres outils, comme son tournevis ou son grille-pain ? La différence réside dans les signaux « humains » envoyés par la machine. C’est ce que démontraient dès 1996 deux chercheurs de Stanford, Clifford Nass et Byron Reeves : si un appareil imite suffisamment certains comportements sociaux, « il sera traité comme un humain, même si les gens savent que c’est idiot ». Ainsi, face à un chatbot poli, nous avons tendance à respecter des normes sociales, développe la sociologue du numérique Valérie Beaudouin, directrice d’études à l’Ecole des hautes études en sciences sociales. « Quand on analyse les conversations humaines, si je vous dis “bonjour”, vous me dites “bonjour”. Or, les chatbots sont formatés pour commencer par une phrase comme “bonjour, que puis-je faire pour vous ?”. Nous sommes donc incités à être polis, car ne pas le faire serait enfreindre les règles de la conversation. » « Je deviens un horrible personnage » Il vous reste 74.97% de cet article à lire. La suite est réservée aux abonnés.

Decision trees are a popular supervised learning algorithm with benefits that include being able to be used for both regression and classification as well as being easy to interpret. However, decision trees aren’t the most performant algorithm and are prone to overfitting due to small variations in the training data. This can result in a completely different tree. This is why people often turn to ensemble models like Bagged Trees and Random Forests. These consist of multiple decision trees trained on bootstrapped data and aggregated to achieve better predictive performance than any single tree could offer. This tutorial includes the following:  What is Bagging What Makes Random Forests Different Training and Tuning a Random Forest using Scikit-Learn Calculating and Interpreting Feature Importance Visualizing Individual Decision Trees in a Random Forest As always, the code used in this tutorial is available on my GitHub. A video version of this tutorial is also available on my YouTube channel for those who prefer to follow along visually. With that, let’s get started! What is Bagging (Bootstrap Aggregating) Bootstrap + aggregating = Bagging. Image by Michael Galarnyk. Random forests can be categorized as bagging algorithms (bootstrap aggregating). Bagging consists of two steps: 1.) Bootstrap sampling: Create multiple training sets by randomly drawing samples with replacement from the original dataset. These new training sets, called bootstrapped datasets, typically contain the same number of rows as the original dataset, but individual rows may appear multiple times or not at all. On average, each bootstrapped dataset contains about 63.2% of the unique rows from the original data. The remaining ~36.8% of rows are left out and can be used for out-of-bag (OOB) evaluation. For more on this concept, see my sampling with and without replacement blog post. 2.) Aggregating predictions: Each bootstrapped dataset is used to train a different decision tree model. The final prediction is made by combining the outputs of all individual trees. For classification, this is typically done through majority voting. For regression, predictions are averaged. Training each tree on a different bootstrapped sample introduces variation across trees. While this doesn’t fully eliminate correlation—especially when certain features dominate—it helps reduce overfitting when combined with aggregation. Averaging the predictions of many such trees reduces the overall variance of the ensemble, improving generalization. What Makes Random Forests Different In contrast to some other bagged trees algorithms, for each decision tree in random forests, only a subset of features is randomly selected at each decision node and the best split feature from the subset is used. Image by Michael Galarnyk. Suppose there’s a single strong feature in your dataset. In bagged trees, each tree may repeatedly split on that feature, leading to correlated trees and less benefit from aggregation. Random Forests reduce this issue by introducing further randomness. Specifically, they change how splits are selected during training: 1). Create N bootstrapped datasets. Note that while bootstrapping is commonly used in Random Forests, it is not strictly necessary because step 2 (random feature selection) introduces sufficient diversity among the trees. 2). For each tree, at each node, a random subset of features is selected as candidates, and the best split is chosen from that subset. In scikit-learn, this is controlled by the max_features parameter, which defaults to 'sqrt' for classifiers and 1 for regressors (equivalent to bagged trees). 3). Aggregating predictions: vote for classification and average for regression. Note: Random Forests use sampling with replacement for bootstrapped datasets and sampling without replacement for selecting a subset of features.  Sampling with replacement procedure. Image by Michael Galarnyk Out-of-Bag (OOB) Score Because ~36.8% of training data is excluded from any given tree, you can use this holdout portion to evaluate that tree’s predictions. Scikit-learn allows this via the oob_score=True parameter, providing an efficient way to estimate generalization error. You’ll see this parameter used in the training example later in the tutorial. Training and Tuning a Random Forest in Scikit-Learn Random Forests remain a strong baseline for tabular data thanks to their simplicity, interpretability, and ability to parallelize since each tree is trained independently. This section demonstrates how to load data, perform a train test split, train a baseline model, tune hyperparameters using grid search, and evaluate the final model on the test set. Step 1: Train a Baseline Model Before tuning, it’s good practice to train a baseline model using reasonable defaults. This gives you an initial sense of performance and lets you validate generalization using the out-of-bag (OOB) score, which is built into bagging-based models like Random Forests. This example uses the House Sales in King County dataset (CCO 1.0 Universal License), which contains property sales from the Seattle area between May 2014 and May 2015. This approach allows us to reserve the test set for final evaluation after tuning. Python"># Import libraries # Some imports are only used later in the tutorial import matplotlib.pyplot as plt import numpy as np import pandas as pd # Dataset: Breast Cancer Wisconsin (Diagnostic) # Source: UCI Machine Learning Repository # License: CC BY 4.0 from sklearn.datasets import load_breast_cancer from sklearn.ensemble import RandomForestClassifier from sklearn.ensemble import RandomForestRegressor from sklearn.inspection import permutation_importance from sklearn.model_selection import GridSearchCV, train_test_split from sklearn import tree # Load dataset # Dataset: House Sales in King County (May 2014–May 2015) # License CC0 1.0 Universal url = 'https://raw.githubusercontent.com/mGalarnyk/Tutorial_Data/master/King_County/kingCountyHouseData.csv' df = pd.read_csv(url) columns = ['bedrooms',             'bathrooms',             'sqft_living',             'sqft_lot',              'floors',              'waterfront',              'view',              'condition',              'grade',              'sqft_above',              'sqft_basement',              'yr_built',              'yr_renovated',              'lat',              'long',              'sqft_living15',              'sqft_lot15',              'price'] df = df[columns] # Define features and target X = df.drop(columns='price') y = df['price'] # Train/test split X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0) # Train baseline Random Forest reg = RandomForestRegressor(     n_estimators=100,        # number of trees     max_features=1/3,        # fraction of features considered at each split     oob_score=True,          # enables out-of-bag evaluation     random_state=0 ) reg.fit(X_train, y_train) # Evaluate baseline performance using OOB score print(f"Baseline OOB score: {reg.oob_score_:.3f}") Step 2: Tune Hyperparameters with Grid Search While the baseline model gives a strong starting point, performance can often be improved by tuning key hyperparameters. Grid search cross-validation, as implemented by GridSearchCV, systematically explores combinations of hyperparameters and uses cross-validation to evaluate each one, selecting the configuration with the highest validation performance.The most commonly tuned hyperparameters include: n_estimators: The number of decision trees in the forest. More trees can improve accuracy but increase training time. max_features: The number of features to consider when looking for the best split. Lower values reduce correlation between trees. max_depth: The maximum depth of each tree. Shallower trees are faster but may underfit. min_samples_split: The minimum number of samples required to split an internal node. Higher values can reduce overfitting. min_samples_leaf: The minimum number of samples required to be at a leaf node. Helps control tree size. bootstrap: Whether bootstrap samples are used when building trees. If False, the whole dataset is used. param_grid = {     'n_estimators': [100],     'max_features': ['sqrt', 'log2', None],     'max_depth': [None, 5, 10, 20],     'min_samples_split': [2, 5],     'min_samples_leaf': [1, 2] } # Initialize model rf = RandomForestRegressor(random_state=0, oob_score=True) grid_search = GridSearchCV(     estimator=rf,     param_grid=param_grid,     cv=5,             # 5-fold cross-validation     scoring='r2',     # evaluation metric     n_jobs=-1         # use all available CPU cores ) grid_search.fit(X_train, y_train) print(f"Best parameters: {grid_search.best_params_}") print(f"Best R^2 score: {grid_search.best_score_:.3f}") Step 3: Evaluate Final Model on Test Set Now that we’ve selected the best-performing model based on cross-validation, we can evaluate it on the held-out test set to estimate its generalization performance. # Evaluate final model on test set best_model = grid_search.best_estimator_ print(f"Test R^2 score (final model): {best_model.score(X_test, y_test):.3f}") Calculating Random Forest Feature Importance One of the key advantages of Random Forests is their interpretability — something that large language models (LLMs) often lack. While LLMs are powerful, they typically function as black boxes and can exhibit biases that are difficult to identify. In contrast, scikit-learn supports two main methods for measuring feature importance in Random Forests: Mean Decrease in Impurity and Permutation Importance. 1). Mean Decrease in Impurity (MDI): Also known as Gini importance, this method calculates the total reduction in impurity brought by each feature across all trees. This is fast and built into the model via reg.feature_importances_. However, impurity-based feature importances can be misleading, especially for features with high cardinality (many unique values), as these features are more likely to be chosen simply because they provide more potential split points. importances = reg.feature_importances_ feature_names = X.columns sorted_idx = np.argsort(importances)[::-1] for i in sorted_idx:     print(f"{feature_names[i]}: {importances[i]:.3f}") 2). Permutation Importance: This method assesses the decrease in model performance when a single feature’s values are randomly shuffled. Unlike MDI, it accounts for feature interactions and correlation. It is more reliable but also more computationally expensive. # Perform permutation importance on the test set perm_importance = permutation_importance(reg, X_test, y_test, n_repeats=10, random_state=0) sorted_idx = perm_importance.importances_mean.argsort()[::-1] for i in sorted_idx:     print(f"{X.columns[i]}: {perm_importance.importances_mean[i]:.3f}") It is important to note that our geographic features lat and long are also useful for visualization as the plot below shows. It’s likely that companies like Zillow leverage location information extensively in their valuation models. Housing Price percentile for King County. Image by Michael Galarnyk. Visualizing Individual Decision Trees in a Random Forest A Random Forest consists of multiple decision trees—one for each estimator specified via the n_estimators parameter. After training the model, you can access these individual trees through the .estimators_ attribute. Visualizing a few of these trees can help illustrate how differently each one splits the data due to bootstrapped training samples and random feature selection at each split. While the earlier example used a RandomForestRegressor, here we demonstrate this visualization using a RandomForestClassifier trained on the Breast Cancer Wisconsin dataset (CC BY 4.0 license) to highlight Random Forests’ versatility for both regression and classification tasks. This short video demonstrates what 100 trained estimators from this dataset look like. Fit a Random Forest Model using Scikit-Learn # Load the Breast Cancer (Diagnostic) Dataset data = load_breast_cancer() df = pd.DataFrame(data.data, columns=data.feature_names) df['target'] = data.target # Arrange Data into Features Matrix and Target Vector X = df.loc[:, df.columns != 'target'] y = df.loc[:, 'target'].values # Split the data into training and testing sets X_train, X_test, Y_train, Y_test = train_test_split(X, y, random_state=0) # Random Forests in `scikit-learn` (with N = 100) rf = RandomForestClassifier(n_estimators=100,                             random_state=0) rf.fit(X_train, Y_train) Plotting Individual Estimators (decision trees) from a Random Forest using Matplotlib You can now view all the individual trees from the fitted model.  rf.estimators_ You can now visualize individual trees. The code below visualizes the first decision tree. fn=data.feature_names cn=data.target_names fig, axes = plt.subplots(nrows = 1,ncols = 1,figsize = (4,4), dpi=800) tree.plot_tree(rf.estimators_[0],                feature_names = fn,                 class_names=cn,                filled = True); fig.savefig('rf_individualtree.png') Although plotting many trees can be difficult to interpret, you may wish to explore the variety across estimators. The following example shows how to visualize the first five decision trees in the forest: # This may not the best way to view each estimator as it is small fig, axes = plt.subplots(nrows=1, ncols=5, figsize=(10, 2), dpi=3000) for index in range(5):     tree.plot_tree(rf.estimators_[index],                    feature_names=fn,                    class_names=cn,                    filled=True,                    ax=axes[index])     axes[index].set_title(f'Estimator: {index}', fontsize=11) fig.savefig('rf_5trees.png') Conclusion Random forests consist of multiple decision trees trained on bootstrapped data in order to achieve better predictive performance than could be obtained from any of the individual decision trees. If you have questions or thoughts on the tutorial, feel free to reach out through YouTube or X. The post Understanding Random Forest using Python (scikit-learn) appeared first on Towards Data Science.

Wētā Digital – now part of Unity – developed many of the tools and solutions used to bring the world of Avatar: The Way of Water to life. Here, we take a look at the CGI technology behind the water. If you’re interested in being among the first to access some of the tools used in the film, you can register for the Unity Wētā Tools beta through our website.James Cameron is no stranger to working with water. Titanic aside, in 2012 he made a record-breaking solo dive, piloting a submarine to the bottom of the Mariana Trench in the Pacific Ocean: Earth’s lowest point at nearly 11 kilometers deep. As he said in the resulting 2014 documentary, Deepsea Challenge, “Down here you feel the power of nature’s imagination, which is so much greater than our own.”It must have been truly remarkable, then, seeing as the world of Pandora and its stunning visuals ultimately came from Cameron’s own imagination.Translating Cameron’s vision, which for the sequel included the new reef village of the aquatic Metkayina clan, required extensive use of visual effects – especially for the dominant water setting.The tools and solutions used to create the film’s VFX – including the award-winning water effects – were developed by Wētā Digital, now a division of Unity.To ensure that the interactions between the characters and water elements were as realistic as possible, a team of experts, including Unity and Wētā’s water simulation VFX specialists Alexey Stomakhin, Steve Lesser, Joel Wretborn, and Sean Flynn, were brought together to form the “Water Taskforce”. This team’s water toolset was recently recognized with a win at the Visual Effects Society (VES) Awards, with the Emerging Technology Award.Extreme attention to detail saw the taskforce conduct extensive research and experimentation in collaboration with New Zealand’s National Institute of Water and Atmospheric Research (NIWA) to find the best approach to creating CGI water. This included taking into account the effects of tides, wind, and the sea floor on aquatic environments.Avatar: The Way of Water required water effects for 2,225 shots, some taking up to eight days of simulation to achieve the high resolution needed.There were also numerous scenes where water interacted with over 50 creatures in a single shot. This presented the challenge of needing simulations to be accurate at scale, from large domains for bigger creatures, to submillimeter resolution for thin film on skin.As it was not computationally feasible to create a single-representation water system, the toolset was developed with a number of distinct solvers to keep compute times to a minimum.“The Loki water state machine was crucial for delivering the sheer volume of large-scale water shots in this movie. In a typical VFX-heavy movie, water shots of this complexity are few and far between and require many iterations and passes from very experienced artists. In contrast, our state machine approach was able to deliver great results after just a single pass, even by artists who had just entered the industry.” – Sean Flynn, simulation lead, Unity x Wētā DigitalA majority of the water tools developed by the team sit within Wētā’s proprietary simulation framework, Loki. This piece of tech includes solvers for multiple water states, including procedural water waves, bulk water, spray, mist, hero bubbles, diffuse bubbles, foam, capillary surface waves, thin film, and residual wetness.State machineMany of these solvers sit within the Loki state machine – an airborne spray system. The water states are coupled with the surrounding air, with transitions between states handled in a mass- and momentum-conserving way.Rather than a one-size-fits-all approach, the Loki state machine allows multiple solvers to run in tandem. Each solver is optimized for the level of detail required by its respective state, such as bulk water, spray, and mist. This helps keep large-scale water simulations efficient while still capturing the very fine droplet interactions required by spray and mist.All of the states including the surrounding air are completed in a single simulation pass. As all solvers are computed with proper physical interactions between them, this is what helped to create such natural and realistic water interaction throughout the film.During SIGGRAPH 2019, a practical approach for modeling close-up water interaction with characters was presented, with a focus on high-fidelity surface tension and adhesion effects as water moves over and drips from skin. Using a scene from Alita: Battle Angel (a screenplay also written by Cameron), the team showed how this method allowed for a resolution of effects that was performant enough – on the scale of a fraction of a millimeter – to cover a whole character with a layer of water.The approach was to adapt an existing particle-in-cell (FLIP/APIC) solver to capture small-scale water-solid interaction dynamics. This technique was then advanced during the production of Avatar: The Way of Water to handle any sequence that involved characters emerging from water.“This was not a cheap solution, as we had to simulate water dynamics at sub-millimeter scales. The results would often take days to compute. We had to ensure our solver was scalable, robust and reliable enough to produce physically plausible visuals out of the box, with minimal tuning required from artists.” – Alexey Stomakhin, principal research engineer, Unity x Wētā DigitalTo achieve believable dynamics in underwater scenarios – for example, when characters breathe underwater in Avatar: The Way of Water – the approach to underwater bubbles was to simulate them together with a narrow band of water around the region of interest. The bubbles themselves would be represented in two parts: a hero and diffuse counterpart.The hero counterpart captures bigger bubbles with more explosive and turbulent behaviors. It utilizes an incompressible two-phase Navier-Stokes solve on a Eulerian grid, with the air phase represented by FLIP/APIC particles to facilitate volume conservation and accurate interface tracking.The diffuse counterpart captures the motion of smaller bubbles below the resolution of the Eulerian grid. The team has developed a novel scheme for coupling diffuse bubble particles with bulk fluid that could also be applied to other submerged, porous objects such as sand, hair, and cloth.To enhance the visual detail of a water surface simulation, the team from Wētā Digital and IST Austria developed a method of post-processing that took a simulation as an input, and increased its apparent resolution by simulating detailed Lagrangian water waves on top of it.Linear water wave theory was extended to work in non-planar domains with Lagrangian wave packets attached to spline curves that would evolve over the bulk fluid surface. This method produces high-frequency ripples with dispersive wave-live behaviors, customized to the underlying fluid simulation.A technique was developed for the realistic movement of underwater bubbles – created by movement in the water – reaching the water surface and converting into foam. This was important for nearly all of the water scenes in Avatar: The Way of Water.Grid-based Navier-Stokes simulators – usually reserved for capturing large-scale motion such as bulk fluid – are inherently limited by their grid resolution, making this method impractical for small-scale phenomena like spray and mist from breaking waves. These whitewater effects are usually simulated as independent Lagrangian particles.“One key aspect of our whitewater method is the interaction of two solvers: a grid-based fluid solver coupled with bubbles, and a SPH solver for foam constrained to the water surface. The declarative solver framework in Loki is what makes building and supporting these complex systems possible in production without having to develop new solvers from scratch.” – Joel Wretborn, senior research engineer, Unity x Wētā DigitalThe key aspect most of the existing solvers neglect are the collective effects: groups of bubbles rise faster than single bubbles due to their combined buoyancy, and the collection of many bubbles can have a significant impact on the motion of the water.The new technique addresses this limitation by simulating bubbles two-way coupled with the surrounding fluid. This effectively captures collective bubble effects, and creates a more connected look between bubbles and the motion of the fluid. As bubbles reach the surface they transition into "wet" foam particles constrained to the water surface, discretized with smoothed particle hydrodynamics (SPH). In the end this created believable whitewater dynamics in both close-up and large ocean shots.The simulation technology used by the Water Taskforce was created by present and former colleagues at Wētā Digital, as well as friends from Wētā FX and academic institutions, including: Alexey Stomakhin, Joel Wretborn, Kevin Blom, Gilles Daviet, Steve Lesser, John Edholm, Noh-Hoon Lee, Eston Schweickart, Xiao Zhai, Sean Flynn, Andrew Moffat, Gary Boyle, Tomas Skrivan, Andreas Soderstron, John Johansson, Christoph Sprenger, Ken Museth, and Chris Wojtan. Learn more about Unity Wētā Tools beta.

html PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN" "http://www.w3.org/TR/REC-html40/loose.dtd" The box office still belongs to Marvel thanks to the success of Thunderbolts*. The MCU offering has grossed $278 million worldwide heading into its third weekend in theaters. However, Final Destination Bloodlines is poised for a franchise-best opening weekend that would top the box office. Theaters will be offering new summer blockbusters every week for the next three months. When you can’t make it to the theater, try streaming a free movie on a FAST service like Tubi and Pluto TV. These services continue to increase in popularity due to their no-cost options. Our recommendations for this weekend include a surprise horror hit, a Mission: Impossible entry, and a puzzling thriller. Recommended Videos We also have guides to the best new movies to stream, the best movies on Netflix, the best movies on Hulu, the best movies on Amazon Prime Video, the best movies on Max, and the best movies on Disney+. Barbarian (2022) 20th Century Studios Cash your ticket if you had a member of The Whitest Kids U’ Know crafting one of the best horror movies of the last five years. Zach Cregger surprised audiences in 2022 with Barbarian, a horror movie set at a rental home in Detroit. Tess Marshall (Georgina Campbell) is surprised to find Keith (Bill Skarsgård) staying at her Airbnb. Despite her instincts telling her to leave, Tess spends the night with Keith. The next morning, things go off the rails. For fear of spoilers, I’ll stop there. Barbarian is a movie that you must see to believe. The twist is less effective the second time around, but if this is your first time, enjoy the madness. Stream Barbarian on Tubi. Mission: Impossible — Rogue Nation (2015) Paramount Pictures It’s Mission: Impossible month in Hollywood, as audiences everywhere prepare for Tom Cruise to entertain them as Ethan Hunt one last time in Mission: Impossible — The Final Reckoning. As part of your rewatch,Rogue Nation should be on your list. After the death of a station operative, the Senate decommissions the IMF and forces its employees to join the CIA. Ethan refuses and goes on the run as he hopes to track down The Syndicate, a nefarious organization that threatens the global powers. Now a fugitive, Ethan recruits his old team and Ilsa Faust (Rebecca Ferguson) to join him in his hunt to destroy this rogue nation. To trigger your memory, Rogue Nation is when Ethan hangs off the side of a plane. Stream Mission: Impossible — Rogue Nation for free on Pluto TV. Escape Room (2019) Sony Pictures Final Destination Bloodlines is a franchise built on insane death scenes. These unfathomable sequences are like Rube Goldberg machines for murder. While Escape Room does not feature the violence of Final Destination, the elaborate life-or-death situations are something the two movies have in common. Six strangers are invited to take part in an escape room challenge. If the players complete the tasks, they will win $10,000. Once the first challenge begins, the group quickly realizes failure to complete a mission will result in physical harm, with death being one of the possibilities. Starring Taylor Russell and Logan Miller, Escape Room is a brisk and entertaining 100-minute thriller specifically made for people who love to solve puzzles. Stream Escape Room for free on Tubi.

OpinionMay 14, 20255 min readFederal Budget Cuts Would Sabotage NASA’s Plans to Find Alien LifeNASA’s astrobiology ambitions are at risk of collapsing under the White House’s proposed budget. But your voice can make a differenceBy Michael L. Wong An artist’s illustration of a potentially habitable exoplanet orbiting a red dwarf star. NASA's Goddard Space Flight CenterWe’ve never been so close to discovering life beyond Earth. Our generation could be the one that finds it—provided two essential ingredients exist. First, that there’s life out there. Second, that we’re willing to look.Alien life’s existence is outside our control, but the universe seems to encourage our attention. Many people rest their optimism about alien life on the remarkable fact that our cosmos is brimming with planetary possibilities. To date, we’ve discovered nearly 6,000 exoplanets, most of them around only the nearest of the Milky Way’s hundreds of billions of stars. That means all our astonishingly successful planet-hunting surveys have studied just a mere teardrop of a vast cosmic sea—and implies there are at least as many planets as stars in our galaxy alone, plus some 1025 worlds in the rest of the observable universe. Chances are we’re not alone—so long as the probability that planets spring forth life is not astronomically miniscule.Discovering alien life, on the other hand, rests squarely on us. For the first time in human history, we can meaningfully answer once-timeless questions. Countless generations before us could only ask “Are we alone?” as passive stargazers. Today our rockets reliably reach otherworldly destinations, our robotic emissaries yield transformative knowledge about our planetary neighbors, and our telescopes gaze ever farther into the heavens, revealing the subtle beauty of the cosmos.On supporting science journalismIf you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.NASA has led the way on this work, but it now faces an existential threat in the form of short-sighted budget cuts proposed by the White House. If passed into law by Congress, these cuts would axe critical space missions, gut NASA’s workforce, and abandon one of the most captivating quests in all of science. Additional sweeping cuts planned for the National Science Foundation would be similarly ruinous for ground-based astronomy and a host of other endeavors that support NASA’s work at the high frontier.Led by NASA, for more than a half-century the U.S. has been building toward a golden age of astrobiology, a field of research the space agency helped invent. The groundwork was laid on Mars, beginning with the Viking missions of the 1970s and continuing into today, where the agency has “followed the water” to dried-up lakebeds. In 2014 NASA’s Curiosity rover uncovered clues pointing to an ancient, life-friendly Mars, and more recently NASA’s Perseverance rover has been caching promising rock samples for return to Earth. Researchers eagerly await their arrival, because if Mars ever did harbor life, then some of Perseverance’s specimens may well contain some sort of Martian fossils.Besides our own familiar Earth, Mars isn’t the only promising incubator of life around the sun. In the outer solar system, NASA’s Galileo probe and Cassini orbiter lifted the icy veils of Jupiter’s moon Europa and Saturn’s moon Enceladus, respectively. Beneath thick shells of ice, both moons harbor global subsurface oceans, which could be teeming with bacterial or even macroscopic denizens. NASA’s Clipper spacecraft launched in 2024 and is hurtling toward Europa, where it will make close flybys of the moon to assess its habitability. The agency has developed concepts for follow-on missions to land on both worlds and taste the chilly chemistry there for the telltale signs of life.Every organism on Earth requires liquid water, but perhaps that’s not a strict requirement elsewhere. Astrobiologists speculate about “weird life” in Venus’s sulfuric acid clouds and in the liquid hydrocarbon seas of Saturn’s frigid moon Titan. NASA plans to visit each of these worlds with state-of-the-art spacecraft—two to Venus and one to Titan—in the 2030s.And then there’s the great expanse of exoplanets. NASA’s James Webb Space Telescope has furnished unprecedented data about exoplanet atmospheres, most of them hot and puffy—the easiest to observe. But the most alluring exoplanets for astrobiologists—those the size and temperature of Earth—are just beyond our sight. Currently, teams of scientists are conceptualizing NASA’s next great eye in space, the Habitable Worlds Observatory, whose mission is as its name suggests: to image and examine dozens of notionally Earth-like planets for the global exhalations of alien biospheres.Taken together, these recent developments mean we could be at the doorstep of the next Copernican revolution, the next paradigm shift, the next epoch of human discovery.But the president’s recently proposed budget for the 2026 federal fiscal year strikes NASA’s Science Mission Directorate with a devastating 47 percent cut. Many of the boldest, most transformative space missions will be on the chopping block if the proposal passes. It specifically defunds the Mars Sample Return project, a cancellation that would squander billions of dollars and decades of investment. It also cancels the upcoming missions to Venus, which would investigate how the only other Earth-sized planet in our solar system turned out so drastically different from our own. And it scraps the launch of the already-built Nancy Grace Roman Space Telescope, a project which among other things is a proving ground for imaging technologies essential to future exoplanet investigations; the loss of Roman would render prospects for the Habitable Worlds Observatory perilously dim. Also at grave risk are funding sources for brilliant early-career scientists working to make astrobiology’s future as bright as can be.NASA simply cannot continue its trend of breakthrough discoveries on only half its present budget. As talent departs the U.S. and organizational memory fades, brain drain will doom its global leadership in space science in what experts have called an “extinction-level event.”And because there is no profit-driven incentive for discovering life on a distant world, corporate entities cannot and will not fill NASA’s void. SpaceX is great at building rockets, not robotic geologists on wheels. Commercial rocket companies hone their success by reliably building the same product over and over again, but most every NASA exploration mission must do something new.A mentor of mine once described astrobiology as “a gateway drug to science.” Astrobiology invites anyone—regardless of age or background—to cultivate curiosity, creativity, humility and patience. It motivates collaboration across fields and across borders. Even if we never discover life beyond Earth, astrobiology would still offer humanity a profound gift, allowing us to marvel as never before at our existence on this lonely and precious blue-green dot.So, we must choose to do astrobiology. That means you, dear reader, have the power to influence this field’s fate. Whether through contacting elected officials, informing your friends and family about NASA’s precarious position, or simply sharing your love for space exploration, your actions can make a difference in humanity’s search for life in the universe.In the best of times, we have only a few opportunities per generation to launch revolutionary space missions, let alone ones that could forever change our sense of place in the cosmos—and perhaps even our destiny. Now we have just a fleeting moment to prevent a multigenerational disaster. If we fail, we’ll lose the future of astrobiology and all the insight it could bring.Worst of all, we wouldn’t even know what we’d be missing.This is an opinion and analysis article, and the views expressed by the author or authors are solely their own and not those of any organization they are affiliated with or necessarily those of Scientific American.

Apple is launching a brand new Home product in the coming year. I’ve been calling it ‘HomePad,’ and these are five of the core features it will bring with it. 7-inch square display Originally, Mark Gurman reported that the HomePad would have “a roughly 6-inch screen.” More recently though, in his big overview of 2025 products, Gurman bumped the expected size to 7 inches. The device reportedly looks a lot like a square iPad, and is somewhat equivalent to two iPhone 16 Pro Max models placed side by side. Smaller than some might want, but that should help keep the cost lower. Brand new OS Apple’s HomePad will come with a brand new operating system, likely dubbed ‘homeOS.’ Gurman says the interface “looks like a blend of the Apple Watch operating system and the iPhone’s recently launched StandBy mode.” It will also dynamically adjust based on how close or far away you are. For example, from far away it might display status update-type info, but if you get close it would change to offer controls. Unsurprisingly, based on Gurman’s ‘StandBy’ reference, the HomePad is expected to support widgets. You’ll reportedly be able to add these to the Home Screen, just like you can on iPhone and iPad. It’s unclear yet if third-party widgets will be supported, but there’s a good chance they will be. Apple has already built tech into macOS that lets you run iPhone widgets there, so extending that feature to the HomePad seems like a no-brainer. Accessories for fitting in anywhere On its own, the HomePad will look like a simple smart display. But Apple is building various bases and attachments that will make it fit into the different spaces of your home. Gurman writes: Apple has designed different attachments for the device, including ones that affix the screens onto walls like a classic home-security panel. There will be bases with additional speakers that can be placed in the kitchen, on a nightstand or on a desk. These additional accessories seem like another attempt at keeping the HomePad’s cost from ballooning (assuming they’re separate purchases). Importantly though, they also acknowledge the reality that different home spaces will require different HomePad setups. AI-infused Siri Apple’s current Home products, the HomePod, HomePod mini, and Apple TV 4K, don’t support any Apple Intelligence features. This means they’re all stuck with the ‘old’ Siri. But HomePad will offer the new, AI-infused Siri that’s more powerful already (mainly thanks to ChatGPT) but will get substantially more so this fall with iOS 19. Siri will be able to perform hundreds of new in-app and cross-app actions, understand your personal context, and more. If it works as expected, there should be far fewer times Siri has to say, “I’m sorry” on the HomePad—a too-frequent occurrence on HomePod. Apple’s HomePad is expected to offer a lot more functionality than what’s outlined above, including serving as a smart home controller, FaceTime device, and more. But the five features I’ve mentioned provide a solid overview of what to expect from HomePad’s core functionality. Are you interested in buying the HomePad? Why or why not? Let us know in the comments. Best HomeKit smart home accessories Add 9to5Mac to your Google News feed.  FTC: We use income earning auto affiliate links. More.You’re reading 9to5Mac — experts who break news about Apple and its surrounding ecosystem, day after day. Be sure to check out our homepage for all the latest news, and follow 9to5Mac on Twitter, Facebook, and LinkedIn to stay in the loop. Don’t know where to start? Check out our exclusive stories, reviews, how-tos, and subscribe to our YouTube channel