Designfax weekly eMagazine

Subscribe Today!

Archives

View Archives

Partners

Manufacturing Center
Product Spotlight

Modern Applications News
Metalworking Ideas For
Today's Job Shops

Tooling and Production
Strategies for large
metalworking plants

Rugged photoelectric sensors see up to 4 meters

Automation-Direct has added AchieVe FDM series 12-mm tubular photoelectric sensors that offer a rugged metal construction, high IP67 protection ratings, and sensing distances up to 4 m. These sensors feature selectable light-on/dark-on operation, a 10- to 30-VDC operating voltage range, potentiometer or teach-in button sensitivity adjustment, and a fast 1-kHz switching frequency. Highly visible red LED models are offered with the polarized reflective sensing style, while infrared models are available in diffuse and through-beam styles. Lots of applications. Three-year warranty.
Learn more.

Engineer's Toolbox: Critical inspection of airplane parts with a SVS-Vistek 10GigE camera

Manufacturers of aviation engine components are being impacted by Industry 4.0's emphasis on quality control, which is challenging them to rethink outdated processes and to embrace new technologies. A new system developed by researchers in Italy uses a Kuka robot, a SVS-Vistek 61-megapixel 10GigE camera, and AI to detect defects in honeycomb aerospace parts faster and with more accuracy.
Read the full article.

What's new in MATLAB and Simulink?

Release 2024b from MathWorks offers hundreds of new and updated features and functions in MATLAB and Simulink including several major updates -- including 5G Toolbox, Simulink Control Design, System Composer, and more -- that streamline the workflows of engineers and researchers working on wireless communications systems, control systems, and digital signal processing applications.
View the video.

COTS-based space-ready orbital systems

Aitech Systems' solutions can meet the growing demands for shorter development times and lower costs among satellite buses, subsystems, and payloads. Using a Space Digital Backbone (DBB) approach, which provides a flexible, scalable communication pathway for the increasing number of Internet of Things technologies being implemented into space missions, the company provides a selection of space-rated subsystems for common space platforms including: Earth observation, communications, power control, navigation, and robotics.
Learn more.

Circuit breakers have magnetic module option

SCHURTER has upgraded its 2-pole classic TA35 and TA36 thermal circuit breaker models with an additional, optional magnetic module. From now on, no additional fuse is required when using a thermal-magnetic type. Depending on the application, the magnetic modules are available either with a slow- or a fast-acting characteristic. Both models are designed for snap-in mounting and with finely graduated rated currents. A variety of colors and lighting options make the designer's choice easier.
Learn more.

All about magnetic rotary encoder

The precision and reliability offered by modern rotary encoders are essential in many product categories. These include robotics, machine tools, printing presses, motion control systems, medical equipment, aerospace, gaming and entertainment, and automotive. Learn all about magnetic rotary encoders -- and important developments in the technology's future.
Read the full Avnet article.

High-res image sensor for automotive ADAS and AD

OMNIVISION has expanded its TheiaCel™ product portfolio with a new OX12A10 12-MP high-res image sensor for automotive cameras. This sensor, with the highest resolution in its line, improves automotive safety by eliminating LED flicker regardless of lighting conditions. It is ideal for high-performance front machine vision cameras for advanced driver assistance systems (ADAS) and autonomous driving (AD).
Learn more.

Durable, full redundant angle sensors for automotive and off-highway

Novotechnik's new RSK-3200 Series angle sensors are designed for harsh automotive and off-highway applications. Measurement range is 0 to 360 degrees, and the temperature range is -40 to 125 C. This unit's built-in coupling accepts D-Shaft, with shaft customization available. The sensors are sealed to IP 67 or IP 69k depending on version. RSK-3200 Series sensors are extremely durable with MTTF of 285 years for each of the two channels! Applications include throttle control and EGR valves, transmission gear position, and accelerator position. Very competitive pricing.
Learn more.

Great design: Handle with integrated lighting/signaling

Signaling and indicator lights, switches, and buttons -- elements that hardly any machine can do without. The new JW Winco cabinet U-handle EN 6284 integrates all these functions into a single, compact element. The new U-handle is designed to enhance the operation of systems and machines. It features an integrated button and a large, colored, backlit area on the handle. These elements can be used individually or in combination, providing a versatile tool for system control and process monitoring that can be seen from across the room.
Learn more.

World's most popular 3D multisensor metrology systems get next-gen addition

Offered in two benchtop and two floor-model options to handle nearly any size part, the SmartScope M-Series systems from Optical Gaging Products usher in the next generation of enhancements in image accuracy, optics, and throughput to the world's most popular 3D multisensor video measurement platform. SmartScope M-Series features fixed optics with a 20-megapixel camera and proprietary Virtual Zoom, combined with advanced sensors, illumination, and accessories, to achieve class-leading optical measurement speeds. Lots more features.
Learn more.

SOLIDWORKS Tips: 3 easy ways to focus on your model

SOLIDWORKS Elite Applications Engineer Alin Vargatu demonstrates his top tips for focusing on your model: finding planes the easy way inside your assembly with the Q key, breadcrumbs, and a better way to use the component preview window. Very helpful. Lots more tips on the SOLIDWORKS YouTube channel.
View the video.

Push-pull transformer drivers for automotive power supplies

Nexperia's AEC-Q100 qualified, push-pull transformer drivers (NXF6501-Q100, NXF6505A-Q100, and NXF6505B-Q100) enable the design of small, low-noise, and low-EMI isolated power supplies for a range of automotive applications such as traction inverters and motor control, DC-DC converters, battery management systems, and on-board chargers in EVs. Also suitable for industrial applications such as telecommunications, medical, instrumentation, and automation equipment.
Learn more.

Mini linear position sensor for drones, robots, aero, more

H. G. Schaevitz LLC, Alliance Sensors Group is now offering a miniature, lightweight LZ SERIES linear position sensor product line utilizing LVIT Technology™. These sensors are designed for tight spaces that require excellent stroke-to-length ratio. They are contactless devices for use by drones, OEMs, aerospace, robotics, factory automation, or assembly machinery applications where precision in position sensing is crucial.
Learn all the specs.

What is a Heatric Printed Circuit Heat Exchanger?

According to Parker Hannifin, "A Printed Circuit Heat Exchanger is a robust, corrosion-resistant, high-integrity plate-type heat exchanger manufactured using diffusion bonding." Learn about the technology and why Heatric, a Parker brand, "can manufacture a unit up to 85% smaller and lighter than traditional technologies such as shell and tube heat exchangers."
Read this informative Parker blog.

Tech Tip: Mastering sheet metal bend calculations in Onshape

Mastering bend calculations in sheet metal design is a key skill that can impact the accuracy and manufactur-ability of your designs significantly. Explore the various options available to become a pro in this Onshape Tech Tip: K Factor, bend allowance, and bend deduction, with guidance on when each should be used. You may learn something even if you don't use this software.
Read the Onshape blog.

It may have impressive output, but generative AI doesn't have a coherent understanding of the world

Researchers show that even the best-performing large language models don't form a true model of the world and its rules -- and can thus fail unexpectedly on similar tasks.

By Adam Zewe, MIT

Large language models (LLMs) can do impressive things, like write poetry or generate viable computer programs, even though these models are trained to predict words that come next in a piece of text. Such surprising capabilities can make it seem like the models are implicitly learning some general truths about the world.

That isn't necessarily the case, according to a new study. The researchers found that a popular type of generative artificial intelligence (AI) model can provide turn-by-turn driving directions in New York City with near-perfect accuracy -- without having formed an accurate internal map of the city.

Despite the model's uncanny ability to navigate effectively, when the researchers closed some streets and added detours, its performance plummeted. When they dug deeper, the researchers found that the New York maps the model implicitly generated had many nonexistent streets curving between the grid and connecting far-away intersections.

This could have serious implications for generative AI models deployed in the real world, since a model that seems to be performing well in one context might break down if the task or environment slightly changes.

"One hope is that, because LLMs can accomplish all these amazing things in language, maybe we could use these same tools in other parts of science, as well. But the question of whether LLMs are learning coherent world models is very important if we want to use these techniques to make new discoveries," says senior author Ashesh Rambachan, assistant professor of economics and a principal investigator in the MIT Laboratory for Information and Decision Systems (LIDS).

Rambachan is joined on a paper about the work by lead author Keyon Vafa, a postdoc at Harvard University; Justin Y. Chen, an electrical engineering and computer science (EECS) graduate student at MIT; Jon Kleinberg, Tisch University Professor of Computer Science and Information Science at Cornell University; and Sendhil Mullainathan, an MIT professor in the departments of EECS and of Economics, and a member of LIDS. The research will be presented at the Conference on Neural Information Processing Systems in December 2024.

New metrics
The researchers focused on a type of generative AI model known as a transformer, which forms the backbone of LLMs like GPT-4. Transformers are trained on a massive amount of language-based data to predict the next token in a sequence, such as the next word in a sentence.

However, if scientists want to determine whether an LLM has formed an accurate model of the world, measuring the accuracy of its predictions doesn't go far enough, the researchers say. For example, they found that a transformer can predict valid moves in a game of Connect 4 nearly every time without understanding any of the rules. So, the team developed two new metrics that can test a transformer's world model. The researchers focused their evaluations on a class of problems called deterministic finite automations, or DFAs.

A DFA is a problem with a sequence of states, like intersections one must traverse to reach a destination, and a concrete way of describing the rules one must follow along the way. They chose two problems to formulate as DFAs: navigating on streets in New York City and playing the board game Othello.

"We needed test beds where we know what the world model is. Now, we can rigorously think about what it means to recover that world model," Vafa explains.

The first metric they developed, called sequence distinction, says a model has formed a coherent world model it if sees two different states, like two different Othello boards, and recognizes how they are different. Sequences, that is, ordered lists of data points, are what transformers use to generate outputs.

The second metric, called sequence compression, says a transformer with a coherent world model should know that two identical states, like two identical Othello boards, have the same sequence of possible next steps.

They used these metrics to test two common classes of transformers, one which is trained on data generated from randomly produced sequences and the other on data generated by following strategies.

Incoherent world models
Surprisingly, the researchers found that transformers that made choices randomly formed more accurate world models, perhaps because they saw a wider variety of potential next steps during training.

"In Othello, if you see two random computers playing rather than championship players, in theory you'd see the full set of possible moves, even the bad moves championship players wouldn't make," Vafa explains.

Even though the transformers generated accurate directions and valid Othello moves in nearly every instance, the two metrics revealed that only one generated a coherent world model for Othello moves, and none performed well at forming coherent world models in the wayfinding example.

The researchers demonstrated the implications of this by adding detours to the map of New York City, which caused all the navigation models to fail.

"I was surprised by how quickly the performance deteriorated as soon as we added a detour. If we close just 1 percent of the possible streets, accuracy immediately plummets from nearly 100 percent to just 67 percent," Vafa says.

When they recovered the city maps the models generated, they looked like an imagined New York City with hundreds of streets crisscrossing overlaid on top of the grid. The maps often contained random flyovers above other streets or multiple streets with impossible orientations.

These results show that transformers can perform surprisingly well at certain tasks without understanding the rules. If scientists want to build LLMs that can capture accurate world models, they need to take a different approach, the researchers say.

"Often, we see these models do impressive things and think they must have understood something about the world. I hope we can convince people that this is a question to think very carefully about, and we don't have to rely on our own intuitions to answer it," says Rambachan.

In the future, the researchers want to tackle a more diverse set of problems, such as those where some rules are only partially known. They also want to apply their evaluation metrics to real-world, scientific problems.

Published November 2024

Rate this article