Posted on

CAPTURING STYLE

In a game of chess, the outsider often gauges the quality of the gameplay by the entertainment and surprise factor offered by the players’ choice of moves. A good chess match is a mind-bending dance of strategy and power: a true spectator sport.

The advent of artificial intelligence (AI) — specifically, software chess engines — seems to have painted this vibrant tableau in shades of monotony.

LISTEN TO THE DEEP DIVE

After chess world champion Garry Kasparov’s defeat against IBM’s Deep Blue in the late ‘90s, human chess players have grudgingly accepted that chess engines can beat them. However, they found solace in the fact that chess engines were utilitarian in their style: A classic chess engine may well consider tens of millions of alternative moves per second, but its playing style — especially with a limited lookahead — is boring. Unadorned and calculated. There’s none of the dynamism or creativity. None of the humanness. Developers did add in some randomness to introduce a semblance of unpredictability, but this resulted in predictable sequences of generally good moves interspersed with occasional mistakes.

As the AI meticulously generates all possible move sequences, it assigns an evaluation score to each resulting game state. This score takes into account the relative power balance between the players after a move. For example, if white gains a pawn, the score is +1, but if black gains a knight, the score for white is -3. More sophisticated engines incorporate positional information, such as the location of the pieces on the board and even factors like piece mobility and the safety of the king.

The addition of randomness forces chess engines to choose a move sequence at random from those with similar scores, or even occasionally play a randomly generated move, just to spice things up.

But these chess engines still spit out long sequences of unimaginative good moves, peppered with the odd bad one.

And, as human players know all too well, blind randomization is unlikely to land a player on a winning streak.

CAPTION: ERNESTO DAMIANI is senior director of the Robotics and Intelligent Systems Institute at Khalifa University

The subsequent generation of chess programs, powered by artificial neural networks (ANNs), learn from gameplay examples rather than predetermined formulas. These neural networks encode a binary representation of each position on the board, piece type and player color.

The outputs are the evaluation function values, leading to rapid and unpredictable gameplay as the bulk of computations are conducted during training, not live games. Developers use millions of online chess games played between humans or by humans against machines to set up training data. As the outcome of each game is known, it’s possible to more finely tune the models, selecting the best move for each scenario. Once trained, the ANN can be used by the chess engine to evaluate quickly and efficiently the score for each possible move — ANN-based chess engines become even more positional. And boring.

Google’s AlphaZero changed the landscape. It implements a new ANN-based approach that can be trained to play not just chess, but other board games. Given a game state, the AlphaZero engine computes a policy that maps the game state to the probability distribution of making each of the possible moves. In chess, that’s 4,672 possible moves — for white’s first move. For human players, this is ridiculous: There are 20 legal moves for white’s first move. And that’s the point.

DESIGN & PROMPTS: : Anas Albounni, KUST Review IMAGES: AI Generated, KUST Review.

AlphaZero includes all sorts of moves that are illegal, like selecting empty squares, selecting opponent’s pieces, making knight moves for rooks, or making long diagonal moves for pawns. It also includes moves that pass through other blocking pieces.

During training, nothing is learned or imposed about avoiding non-valid moves. The engine just post-processes the ANN output, filtering out illegal or impossible moves by setting their effective probability to zero. Then it re-normalizes the probabilities across the remaining valid moves. A philosopher could argue that this engine is devoid of ethics as it does not distinguish illegal from impossible, but the resulting ANN structure is simpler than one expressing only valid moves. On a modern processer, AlphaZero needs just a few tens of milliseconds to make a move.

This speed enabled AlphaZero to play against itself in millions of games, completing its training with reinforcement learning, which privileges moves that lie on a sequence that led to victory in the past. Of course, to know whether a move is on a winning sequence, one must complete the game, so AlphaZero reinforcement was performed by playing “fast and dumb,” i.e., using a very shallow search depth. Playing dumb in the reinforcement training phase maximizes the number of games that end in victories and defeats rather than in uninformative draws. As a result of this, AlphaZero considers fewer positions than the algorithmic chess engines of the past.

A classic chess engine may well consider tens of millions of alternative moves per second, but its playing style is boring.

AlphaZero and successors like Deep Chess have a distinctive style, steering away from merely seeking positional or material advantage. Their style is alien — best likened to atonal music: difficult to appreciate for anyone but the chess elite, and certainly useless for the average human amateur to learn or improve their game.

It is interesting that we humans still describe an intelligent chess engine’s style in positional terms.

Like the ‘90s Deep Blue victory, today’s post-AlphaZero scenario highlights some general problems that we will have to solve to be able to work together with the super-human AI engines of the future. AI decision-making must be intelligible to humans for us to accept its decisions. This interpretability needs to be wired into the AI training. Plus, interacting with humans is a crucial step for AI engine evolution: Playing against all possible competitors makes them stronger than any human can individually hope to become.

The game of chess, always a metaphor for life, suggests that controlling the evolution of future AI engines may become more akin to taming a tiger than training a pet.

Ernesto Damiani is senior director of the Robotics and Intelligent Systems Institute at Khalifa University.

Posted on

Build your own robot

The first step in building any robot is to decide what you want it to do. While most of the robot’s abilities will be unlocked with clever machine learning and artificial intelligence algorithms, you need to set your robot up for success with the right mechanical features.

LISTEN TO THE DEEP DIVE

For a human eyeball, nice and round, turn to embedding light-sensitive receptors directly onto the surface of a 3D sphere like the team from the Hong Kong University of Science and Technology, UC-Berkeley and the Lawrence Berkeley National Laboratory.

You could also add a narrow bandgap semiconductor as a photosensing material — then your robot could see in the dark with infrared light sensing. In lieu of realism, you could turn to any number of sensors to have your robot “see”:

Distance sensors and gauges – maybe an ultrasonic range finder or laser measurement sensor. Positioning sensor – room navigation or indoor localization might come in handy. A GPS system or other live tracking devices will help your robot find its way around.

Thermal imaging sensors or pressure sensors are also an option.

Facial recognition – that’s some machine learning pre-programming.

LEGS
Want to jump? Forget biomimicry. Researchers at the UC Santa Barbara use an actuator system based on elasticity. It’s a spring with rubber bands and carbon fiber slats used to shoot the bot into the air.

Or keep the biomimicry but add hydraulic systems and electric motors a la Boston Dynamics’ Atlas.

You could leave humanity behind and go the marsupial route. German engineering firm Festo took it one further and developed the BionicKangaroo.

A “tendon” in its robotic leg drives it forward and captures energy on landing. The impact drives the legs into position for the next leap on its spring-loaded legs.

GRAPHICS: Abjad Design

Stanford University engineers developed a “stereotyped nature-inspired aerial grasper” or SNAG, bird-shaped feet that can perch on any branch.

WINGS
Go classic with drone design and choose rotary wings that spin to create lift and thrust like a helicopter. These are best for hovering, vertical takeoff and changing direction quickly.

Maybe you’d rather the classic plane look and have room for a runway or launcher. Fixed wings generate lift by moving through the air and offer higher speed, longer endurance and greater stability, though your robot will be at the mercy of the weather conditions.

You could even turn to the flapping wings of insects and birds. There are complex transmission systems using gears and motors available from the Harvard team that developed a solar-powered tiny robot styled after a honey bee. A team at the University of Bristol developed a tiny flying robot that flaps its wings more efficiently than an insect, using an electrostatic “zipping” mechanism (their words).

HANDS
What kind of hand does your robot need? Do you want the classic gripper, optimized for delicacy or accuracy? Or is a suction cup plenty?

How many joints does your robot arm need? You’re not limited by human anatomy here.

Many robot hands come with sensors packed into their fingertips only, but an MIT team built a robotic finger with sensors providing continuous sensing along the finger’s entire length, allowing it to accurately identify an object after grasping it just one time.

Researchers at Columbia Engineering developed a highly dexterous robot hand that can operate in the dark. It uses tactile sensors rather than vision to manipulate objects.

Posted on

Protecting your produce

When we purchase fresh produce from our local grocery store, we aren’t usually consumed with worry about whether it contains fecal matter. But the threat is real, and current testing methods are tedious and expensive. That’s why a team of researchers from Purdue University in the United States has developed a reliable and quick method to ensure the produce on our table isn’t contaminated.

Listen to the Deep Dive

But what is the threat?

A cattle farm in Arizona’s Yuma County in the United States produces 115,000 cows annually. Just three miles from the facility is a lettuce farm that is threatened by dust or irrigation water contaminated with feces. An investigation found E. coli bacteria in a nearby canal, and because Yuma County produces 90 percent of the USA’s winter lettuce, these risks need to be mitigated.

In fact, a 2018 outbreak of the same strain of E. coli killed five people after they consumed produce from the Yuma Valley.

“With changes in climate and emergence of new threats (e.g., most recently highly pathogenic avian influenza), maintaining the status quo will increase the burden of these threats,” the Purdue team’s Mohit Verma tells KUST Review.

IMAGE: Pixabay

His team’s new biosensor aims to mitigate these threats.

For a method of detection to be easily integrated, it needs to be accurate, cost effective and simple.

The biosensor detects DNA using loop-mediated isothermal amplification (LAMP), which is simpler than the polymerase chain reaction (PCR) method because it operates at a constant temperature, rather than requiring temperature changes. And to detect fecal contamination, the team uses Bacteriodales, which is an order of bacteria found in animal feces and intestines, but not usually in the surrounding environment. This makes Bacteriodales the best measure of fecal-matter presence.

CAPTION: Verma lab’s molecular tests(using loop-mediated isothermal amplification or LAMP), that can be completed with just some warm water incubation. Results can be read within one hour. IMAGE: Purdue Agricultural Communications

Small plastic sheets on wooden skewers, called collection flags, are placed around the farm and left for a week to collect samples. The flags are then collected and swabbed to transfer bioaerosols — small particles from any nearby animal operations — to the team’s biosensor, which use LAMP to amplify Bacteriodales DNA. The presence and amount of this DNA will cause a color change that can be measured instantly, detecting any level of fecal contamination ranging from safe to high-risk.

The current method of detection is typically lab-based. The biosensor, however, when compared with lab results of the lab-based quantitative polymerase chain reaction results, proved 100 percent accurate.

The team does admit, however, that the testing was done in extreme conditions (very high and very low levels in the field), but still anticipates more than 90 percent sensitivity and specificity at intermediate-level testing.

“These biosensors have the potential to serve as a site-specific risk-assessment tool. They can provide a faster response and thus help in curbing problems before they become too large. They can also help in guiding decisions quickly compared to current lab-based approaches,” Verma says.

Traditional methods of testing also require expensive equipment, expert staffing, take 24-48 hours or more to produce results, and each test runs about U.S.$50. The new biosensor, however, requires simple equipment costing about U.S.$200 and provides equally accurate results within one hour at U.S.$10 per test.

Verma says the collection flags will help producers make important decisions about where to plant and the type of crops based on biosensor results. Also, it can help farmers determine if harvest timings should be adapted due to environmental risk or changing weather patterns by providing site-specific data.

“The biosensor is designed with the end user in mind. Thus, it is meant for use by producers and food safety professionals. The biosensors come with an operation manual and the user can be trained within an hour to run the assays.”

— Mohit Verma, associate professor of agricultural and biological engineering — Purdue University

The applications are not limited to fecal detection on produce farms, however.

“These biosensors are broadly applicable because they can detect DNA or RNA. Specifically, when detecting Bacteroidales, they could be applied for measuring water quality as well. In addition, Bacteroidales can be used for microbial source tracking, i.e., determining where fecal contamination might be coming from. Thus, it applies to water safety as well,” Verma says.

Verma’s new start-up company, Krishi Inc., will develop the biosensor technology commercially and work to enhance its versatility and ease of distribution. Verma says he hopes to also target the health market for companion animals such as cats and dogs, developing biosensors to detect antimicrobial resistance in urinary-tract infections and skin and ear infections.

Bigger picture, Verma hopes to alleviate the current limitation to lab-based methods for surveillance and diagnostics. “The biosensors have the potential to overcome this bottleneck by becoming widely available, providing a rapid response and enabling use in the field,” he tells KUST Review. “Currently, our response time to microbial threats is very slow.”

Funding from the Center for Produce Safety and several other industrial partners supported the team’s work on Bacteroidales.

The 2024 paper was published in Science Direct.

Posted on

Data-driven energy

You might think AI’s role in the energy industry is restricted to exploration — namely, finding the goods. But that’s only a small part of what AI can do.

LISTEN TO THE DEEP DIVE

“AI is impacting almost every single aspect of the energy industry, and as the industry begins rolling out AI-based solutions and realizes the potential of this technology, we expect AI to expand into every area of the industry in some way or another,” says Chris Cooper, CEO of AIQ, an Abu Dhabi-based technology company taking on sustainability and efficiency across the energy sector.

AIQ is a joint venture between the Abu Dhabi National Oil Co. (ADNOC) and AI specialist G42.

Much of the efficiency AI offers in the industry stems from prevention — in all stages of processes.

AI AS RISK MANAGER

AIs are being used in risk aversion: Keeping track of regular machinery maintenance prevents equipment malfunctions and breakdowns.

Though preventative maintenance has been around for a long time, AI can constantly analyze the current condition of equipment and use historical and current data to determine deviations from the norm. This can bring attention to issues before they happen, reducing or eliminating failures and therefore saving millions of dollars in down time.

Not only does this save money, but AI’s preventive elements are also in charge of finding indications that environmental disasters are on the horizon.

Oil spills are one of the most impactful risks to the environment, especially the oceans. In 2022 alone, 700 metric tons of oil leaked in four oil spills. This damage is catastrophic to the ocean’s wildlife and ecosystems: Animals can die from hypothermia when the fur coats that insulate or the feathers that repel water become coated in oil. Plus, while cleaning themselves these animals ingest poison that can impact reproduction and growth rates in offspring.

Aside from things we cannot control like natural disasters, the main causes of oil spills are equipment failure and human error. AI can manage these.

The cost of human error is high. The 2010 Deepwater Horizon explosion, for example, killed 11 people and is the most expensive water oil spill in history. It cost BP and partners about U.S.$71 billion in legal fees and cleanup costs.

“Many of the AI solutions that have been developed to address efficiency and productivity will also bring improvements to overall sustainability by reducing waste, eliminating unnecessary processes and so on.”

— Chris Cooper, CEO of AIQ

AI is used to monitor pipeline conditions for small amounts of deterioration, cracks, etc., offering critical real-time information. Identifying imperfections can help operators address the issues before leaks occur, accidents happen or machines fail. Predictive maintenance contributes to safety and savings across the industry and reduces unpredicted downtime, typically averaging 27 days at a price of U.S.$38 million.

ADNOC initiated pilot testing of its Centralized Predictive Analytics Diagnostics (CPAD) in 2017, and it remains at the center of the company’s digital transformation path.

“CPAD’s predictive maintenance capability can track any mechanical degradation as well as variations in performance to help maintenance teams to plan required work well in advance with consideration to any production constraints,” the website reads.

“AI technology is contributing to monitoring and maintenance again in many different areas, from computer vision being used to monitor pipelines, to sensor monitoring to detect variations in machine operations, through to chemical analysis to detect corrosion in pipelines and monitoring of remote sites or hard-to-reach locations using drones,” AIQ’s Cooper says.

AI can also monitor pressure and oil-flow rates to identify issues before they become problems, mitigating leaks and ensuring worker safety.

GRAPHICS: Abjad Design

The benefits are ample, protecting the environment, workers and the bottom line.

AI’s ability to predict demand means it is also poised to provide risk assessment for investors.

Investment analysis incorporates political and economic events, trends in oil products across the value chain, historical data, public inclination and so on. AI can collate all the information that contributes to price fluctuations to help investors make well-informed decisions.

EXACTING THE EXTRACTION

Energy companies are also harnessing AI capabilities to process data from seismic surveys, improving the accuracy of drilling locations and the drilling plans themselves.

AI algorithms can come up with new extraction techniques and create reservoir models to anticipate how different extraction approaches will work in multiple conditions. This could mean better operations and more lucrative extractions with less environmental degradation.

“Better reservoir development and planning can result in fewer wells being drilled to extract the same resources,” Cooper says.

The best way to develop and plan is to model. One AI tool that AIQ has in the market to assist with reservoir modeling is the AR360 (Advanced Reservoir 360). The 360-degree model allows for review and all-encompassing digital assessment of existing reservoir simulation models.

“The Reservoir Performance Advisor module takes advantage of machine learning, advanced analytics, petro-technical workflows and business logic,” Cooper adds. The system analyzes the data identifying wells declining in performance then provides solutions. What used to be done manually over months is now completed in minutes, improving strategic production, field development planning, cost reduction and lower emissions.

Once the oil is out of the ground and ready for the refining process, analysis ensures maximum efficiency throughout. Refining the process, not just the oil, reduces offsets and, consequently, the environmental impact.

“Many of the AI solutions that have been developed to address efficiency and productivity will also bring improvements to overall sustainability by reducing waste, eliminating unnecessary processes and so on,” Cooper tells KUST Review.

Now that the oil is produced, it’s time to get it to its final destination.

AI AS SUPPLY-CHAIN MANAGER

Supply chains around the world are improved by machine learning. Behaving proactively, rather than reactively, will not only get goods from A to B efficiently and safely, it can balance out supply and demand, saving money across the entire production process.

There are many factors built into predicting oil prices, and while some believe it’s too complex for AI to accurately predict, that isn’t stopping researchers from trying.

A team from China’s Shenzhen University found that rather than using single-model machine-learning methods for price prediction, combining multiple models coupled with specific Google Trends for the online data shows promise. This structure, the team says, results in more accurate anticipation of crude-oil price fluctuations.

The team’s AI analyzes large amounts of historical data, looks for patterns and trends and then combines this with current market information to predict demand.

Supply-chain improvement also means better navigation for ships — shorter routes, safety monitoring en route and recommendations for changing course, should the need arise. This is also a sustainability issue.

GRAPHICS: Abjad Design

Paul McStay, performance manager for oil giant Shell’s liquefied natural gas fleet, says improving efficiency can help the industry reduce carbon emissions. “If we can improve our efficiency, we can reduce the amount of time that we’re waiting at port. Then we can reduce our fuel usage. And by reducing our fuel usage, we can improve our emissions,” he says on Shell’s website.

Route planning and fleet management using operations research, mathematics and optimization techniques aren’t new, says Mohammed Omar, who chairs management science and engineering at Khalifa University.

But amplified efficiency and accuracy provided by AI across the board reduce risk, save money and lower the carbon footprint

According to AIMMS, an analytics software company that has been optimizing mathematics to help companies become more efficient since 1989, supply chains can no longer live without AI. Conversely, AI cannot live without supply chain planners.

WHAT’S NEXT?

In a 2023 survey from EY, a leading auditor of oil and gas companies, 50 percent of oil companies reported using AI in some way and 92 percent are planning to begin or add AI applications within the next five years.

With those statistics, it might be understandable for industry workers to worry about their jobs. But many developmental reports and articles insist AI will work in conjunction with humans, not replace them. They say a shift will definitely occur in human roles, but it is unlikely to result in significant job loss.

According to a 2019 EY report, “The AI revolution is already here for some, and for others, such as oil and gas, it’s just around the corner. AI and ML techniques applied to the sector have the potential to take large amounts of structured and unstructured data with a processing power far greater than a company’s workforce, creating transformative impact. What’s more, when AI and ML are coupled with human workforce capabilities, the combined collective intelligence impact has the potential to create lasting competitive advantage.”

Five years on, it seems EY’s foresight was on the money.

Cooper uses the example of AIQ’s suite of AI-enabled applications for borehole image data analysis, WellSight. The analytics are a time-saver for petrophysicists, freeing them up for more complex functions, and the system offers information to enhance planning for drilling operation.

It all sounds positive, but what’s the catch?

WE’RE STILL LEARNING

We’re still on the cusp of understanding AI’s full capabilities in any industry. That means we need people skilled in AI and machine learning. There will also be significant investment in training — training throughout the digital transformation but also retraining those shifting to different roles.

Plus, big data is what AI needs in order to perform, but if the data isn’t relevant, the AI will not meet expectations, and all that money will be wasted.

“This cutting-edge technology could be used to make our world safer and better, opening up possibilities that seemed like science fiction just a few years ago.”

— Wael William Diab, of the International Organization for Standardization

Finally, the main challenge is getting the buy-in in the first place. Some businesses simply aren’t ready.

Wael William Diab, from the International Organization for Standardization, says it’s all about the mindset: “An AI-positive future is possible, but we need to actively pursue it. If we approach AI with a positive mindset, placing societal needs such as ethics and sustainability at the heart of its development, then we can unlock its full potential,” he says on the non-governmental organization’s website.

“If it is developed ethically and responsibly, AI could help to usher in a new era of innovation and inclusion,” he adds. “This cutting-edge technology could be used to make our world safer and better, opening up possibilities that seemed like science fiction just a few years ago.”

BUT SHOULD WE OR SHOULDN’T WE?

Some of the big concerns surrounding AI rollout globally are ethics and trust. But in the energy industry, buy-in concerns can typically be mitigated with transparency and education.

“If an engineer doesn’t understand how an AI draws the inference it does, then they are less likely to trust the outcomes, and won’t be able to understand any errors that might occur. So, it is important that the users have a good understanding of how the solution works,” says Cooper.

AI understanding is one aspect, but when something goes wrong, the first question is usually: Who did it? Who is accountable when an autonomous AI system is at the helm?

Cooper says it all boils down to finding balance: Balance between rules, framework and AI decision-making and when a human is required to step in.

Posted on

DATA to delivery

Welcome to Industry 4.0, considered by many experts to be the fourth industrial revolution. Artificial intelligence and data analytics are a big part of it and are already changing how supply chains work. Here are just some of the ways they make getting a product from the manufacturer to your home cheaper and more efficient.

IN THE FACTORY

Generative design: An algorithm receives design parameters (such as cost and information on available materials) and generates thousands of options to find the best one.

LISTEN TO THE DEEP DIVE

Order management: AIs handle complicated order information from multiple channels.

Quality control: Sensors inspect products for defects.

Predictive maintenance: AI monitors systems and machines for early signs something is about to break down, preventing expensive factory shutdowns.

Compliance management: AI manages the red tape when the same product is sold in different markets with different regulations.

Customization: AI may be used to create such customized orders as bespoke suits and made-to-order shoes. And in a process called “reshoring” or “nearshoring,” products made far away can be customized closer to the sale point at the last minute.

IN THE WAREHOUSE

Stocking: Digital cameras monitor inventory levels and AI robots pick, sort and pack products.

Finding damaged packages: Machine learning models scan and analyze images to spot damaged objects.

Helping workers with wearable technology: Smart glasses “read” barcodes. Natural language processing helps humans work hands-free to pick items more safely.

THROUGHOUT THE PROCESS

Supply chain visibility: Internet of Things (IoT) devices provide instant information about such conditions as the location and temperature of shipments. Businesses can spot bottlenecks, manage disruptions in real time and make data-driven decisions.

Collaborative supply chains: Multiple companies use data and analytics to work together to plan and execute supply chain operations. The cooperative approach allows the companies to serve similar customers or achieve a common goal.

DELIVERIES

Optimal routes: Vehicle routing algorithms (without problems) use such factors as capacity, delivery priorities and time windows to plot the most efficient routes.

Real-time conditions: AI can monitor weather, traffic and other conditions to reroute as necessary.

Autonomous vehicles: Truck platooning technology can permit a group of vehicles to operate extremely closely, reducing wind resistance and decreasing fuel consumption for transportation between factory and warehouse or retailer. Smaller vehicles will be used for deliveries. Algorithms optimize routes while AI helps vehicles avoid collisions.