Base imagery © 2023 Hexagon and data partners

A rising demand for high-quality imaging

In today’s imaging industry, the demand for ever-higher image quality is relentless, yet constraints on cost, weight and size continue to challenge engineers. Meeting these requirements often means more complex or larger hardware, pricier components, and increasingly demanding testing, all of which push projects toward higher budgets and longer development cycles.

Meanwhile, AI-driven image enhancement tools are rapidly advancing and becoming more widespread, from image deblurring features in Lightroom and Photoshop to AI upscalers and generative models like Stable Diffusion. This raises a compelling question: Can artificial intelligence remove blur and improve usable image resolution without altering the existing optical hardware? In other words, could we now finally achieve the kind of dramatic ‘image enhancement’ often depicted in popular TV shows in real-world applications?

Rewriting image quality with AI

The answer? Kind of. The big breakthrough is that artificial intelligence can now act as a kind of ‘virtual upgrade’ for optical systems. By training models on images from high-end cameras or sensors, developers can teach AI to reconstruct what a better system would have seen. In practice, this means cheaper, smaller and lighter optics can produce images that look like they came from far more expensive hardware, dramatically reducing overall system costs.

There are two especially promising approaches. The first is to use AI to boost the image quality of cheaper cameras and sensors, allowing low-cost, lightweight hardware to produce results closer to premium systems. This can also be used in constellations of sensors or satellites: a few units carry premium optics, while many others use simpler hardware. The second is retrospective enhancement: applying AI for image deblurring in hardware already deployed in the field, improving quality without any physical upgrades. Together, these approaches show how AI can extend the capabilities of both future and existing optical systems, purely through software.

At its core, this technique is generative: the AI starts with a degraded image and generates a sharper, higher-quality version. This only works well if the distortions and noise of the real optical system are accurately modeled during training, which requires deep technical expertise. A strong starting model is also crucial, using an AI already trained to deblur regular images makes adaptation faster, cheaper and more reliable. This is how you can reuse and fine-tune existing models to turn image enhancement from a research challenge into a practical engineering tool.

Base imagery © 2023 Hexagon and data partners

AI in action: From space to surgery

AI-powered image enhancement is starting to improve real-world systems across the board. Any product that relies on optics can benefit: from drones monitoring traffic and satellites mapping crops and deforestation changes, to smartphones capturing sharper photos without bigger cameras. Even weather alerts and environmental policy enforcement can be more accurate thanks to clearer, AI-enhanced imagery.

In optical systems for life sciences, AI image deblurring can boost standard microscopes by leveraging training on high-resolution systems. This approach offers high-resolution capabilities to millions of conventional microscopes in the field, improving live-cell imaging, digital pathology and microbial detection. It also supports low-cost diagnostic devices and high-throughput drug screening. This will help researchers and clinicians gain clearer, more reliable insights without upgrading hardware, making advanced imaging more accessible, scalable and cost-effective.

Generative constraints and reliability

AI-based image deblurring is powerful, but it isn’t magic. It can’t recover information that was completely lost in the original image. Since it’s a generative process, it can also introduce artifacts. This is inevitable, but in most cases, not a dealbreaker. For example, imagine a blurred photo of a pedestrian holding a small object. After deblurring, the pedestrian becomes clearer, but the artificial intelligence can only make an educated guess about the object. They might reconstruct it as a phone, a coffee cup, or something else entirely, but for tasks like counting pedestrians or detecting presence, this isn’t a problem.

The AI works much like humans do: it interprets shapes and context to reconstruct likely features, but the accuracy depends heavily on the quality and variety of training data. That’s why high-quality images across diverse conditions and blur types are crucial. By understanding these limits, AI deblurring can be used responsibly, enhancing images without creating misleading or false information.

A new era of optics and AI

AI-driven image enhancement is already proving to change the game tremendously. It can boost the performance of cheaper cameras and sensors, and even improve images from hardware that’s already out in the field. Better-quality imagery at lower cost could transform Earth observation, from tracking traffic and deforestation to responding faster to disasters. While the impact will be huge, broad adoption will likely happen step by step. As artificial intelligence continues to get smarter, the line between what optics alone can achieve and what software can deliver will keep blurring, opening new possibilities for how we capture and use images across industries.

The post Blurring the line between optical hardware and AI appeared first on Verhaert Masters in Innovation.

The post Blurring the line between optical hardware and AI appeared first on Verhaert Masters in Innovation.

Rethinking the narrative

It’s not so strange that many people see large language models as a threat. For most, ChatGPT was their first and only encounter with this technology. It arrived overnight and it looked like a powerful chatbot that could easily replace human input. But LLMs are much more than conversational tools. Their true strength lies in decision making: the ability to interpret, connect and act on complex information at scale.

History shows that every technological leap reshapes how we create value, not just through the most obvious applications. The question isn’t whether to adapt, but how to do it in a way that strengthens what already works. For example, one of our clients in the paper industry, a field that might seem distant from digital innovation, is using LLM technology to extend its paper products by building an adjacent digital platform. Features powered by transforming text into structured data capabilities now help transform how users interact with information, turning written material, like notes, into actionable insights. The core product remains the same, but its value expands into new, data-driven territory.

Making AI work for your business

Real business value with AI doesn’t come from following the hype, but from understanding what the technology can actually do. As mentioned earlier, large language models bring a new decision-making capacity to digital systems. Not just processing information, but also interpreting and connecting it. They can add a new layer of intelligence to existing tools, combining capabilities like reading images, extracting data and putting information into context. Even simple implementations can improve workflows, but when these features are combined, they can unlock opportunities that weren’t possible before.

Nevertheless, everyone has access to the same AI platforms, so the key to creating value isn’t the tool itself, but how different models, data sources and methods are combined to address specific challenges. Some companies focus on internal productivity, letting AI handle routine decisions and free up human time. Others make existing products smarter, for instance, a CRM that can understand client messages, automate responses and collect insights for better service.

Choosing the right platform starts with understanding what you need it to do. A handwritten form might require an OCR tool, while light decision-making could work with a smaller, private model. Tailoring a system means orchestrating multiple components, from prompts to context layers to fine-tuning, so the result fits the intended audience. Through this kind of integration, AI becomes a practical partner rather than just another piece of software.

Avoiding scaling, quality and data security pitfalls

Scaling LLM-based tools comes with a unique set of challenges that go beyond the technology itself. Cost and infrastructure are often the first hurdles: some platforms charge per interaction, and storing full conversation histories can quickly become expensive. Deciding what to keep, what to archive and how to link data efficiently requires careful planning.

Cybersecurity and privacy remain critical concerns. As with any digital tool, companies must ensure sensitive information isn’t exposed—especially when using external AI services. Internal tools with proper protocols are safer, but external platforms require encryption and strict usage guidelines. Confidential data should never be fed into systems where control cannot be guaranteed.

Another common worry is “losing control” to the AI. This can be mitigated with two approaches: man in the loop, where humans validate outputs and maintain oversight for critical decisions, and man on the loop, where AI handles routine tasks while humans monitor trends and intervene when necessary. By understanding these pitfalls and planning for them, organizations can scale LLM solutions responsibly, without compromising privacy, quality or operational control.

The next wave of intelligence

The evolution of LLMs has been rapid—from chatbots to retrieval-augmented generation and now more advanced decision-making tools. Companies that wait too long risk missing out: the cost of not experimenting is losing value and falling behind. The next wave of opportunity lies in understanding where these technologies truly fit, combining tools thoughtfully and implementing them so people actually use them. The most exciting potential is the broader impact—how these tools can change workflows, products and even society itself.

The post More than a chatbot: Turning LLMs into real business value appeared first on Verhaert Masters in Innovation.

The post More than a chatbot: Turning LLMs into real business value appeared first on Verhaert Masters in Innovation.

Interactive models: a blessing or a curse in disguise?

While many models can predict performance with incredible precision, the true challenge lies in optimizing them effectively. Though powerful, relying solely on human intuition and interactive interfaces may lead to inconsistent outcomes, chaotic workflows, and missed opportunities. How do you ensure you’re not only finding the ‘best’ solution but exploring the full spectrum of possibilities? The answer lies in embracing PIDO, where rigorous tools take over and guide you through even the most complex parameter landscapes—revealing insights beyond the obvious.

Model interaction is only possible if an interface is in place. The most engaging type is a fully interactive graphical user interface (GUI) linked directly to the model. Typically, there are fields for input parameters and outputs. You can make adjustments, click some buttons, and inspect the matching results generated by the model.

At Verhaert, we’ve developed many such models where interactivity allows users to freely explore and adjust parameters as quickly as new concepts emerge. However, the downside is that the optimization control is in the hands of human intuition, leading to a chaotic process, often where repeated trials yield different results. Thus, while interactivity can be a blessing, it can also be a curse for rigorous optimization crucial for informed business decisions.

So remember: optimization is a rigorous task best left to accurate programs. The most crucial feature of your model isn’t a flashy GUI, but a modest scripting interface that another program can control.

One tool to rule them all

Programs that utilize this scripting interface for design exploration and optimization are referred to as PIDO (Process Integration and Design Optimization) tools. These programs operate at a meta-level and perform three core tasks:

1. Systematically varying parameters within a user-defined parameter space.
2. Running an external model (or a chain of models) using these parameters.
3. Saving the results of each run in a structured manner.

For example, Excel’s “scenario manager” and “data table” functions allow you to evaluate functions across different cases and summarize the results in a table.

Dedicated PIDO packages such as optiSLang (Ansys), HEEDS (Siemens), Isight (Simulia), and Optimus (3rd party) offer even more advanced features handling everything from spreadsheets and Matlab scripts to finite element (FE) simulations or a combination of them. These packages also come with optimization algorithms and extensive data visualization suites right out of the box.

If you would rather have full control in developing a PIDO script, you can quickly assemble it in Python using open-source libraries. Our recommended libraries include:

• Systematically varying parameters and saving the results: psweep or pypet
• Interfacing with models written in other programs: Python Subprocess
• Optimizing objective functions: DEAP (evolutionary)
• Visualizing data: matplotlib or seaborn (mostly static) / Plotly (interactive)

Once created, a custom PIDO script can be reused for any model with minimal modifications.

In short, if your model has a scripting interface, PIDO software can leverage it to optimize designs with minimal manual input. If you have a license for FE software, chances are a PIDO program comes prepackaged with it; and if you don’t, it’s easy to cook up in a language like Python. So why not give it a try? 

The post ‘Pareto Front’ and beyond: discovering the art of smart simulation & modeling appeared first on Verhaert Masters in Innovation.

The post ‘Pareto Front’ and beyond: discovering the art of smart simulation & modeling appeared first on Verhaert Masters in Innovation.

Key trends in our focus industries

The initial wave of the pandemic resulted in widespread factory shutdowns and reduced demand for goods, leading to a decrease in production and a reduction in the supply chain. In addition, the pandemic has led to a shift in consumer behavior, with a growing focus on essential goods and a decrease in demand for non-essential goods.

In the wake of the pandemic, the manufacturing industry has had to adapt to new challenges and find new ways to operate efficiently while ensuring the safety of employees. This has led to a greater emphasis on automation and digitalization, as companies seek to reduce the need for manual labor and minimize the risk of virus transmission.

The conflict between Ukraine and Russia has also had an impact on the manufacturing industry in Europe. The ongoing conflict has led to disruptions in the supply chain and reduced access to raw materials and goods. In addition, the conflict has led to increased uncertainty and volatility in the market, making it difficult for companies to plan and operate effectively.

How to address the trends

To address these challenges, many companies in the manufacturing industry are seeking to diversify their supply chains and reduce their dependence on a single region or supplier. This includes investing in local production facilities and seeking out new suppliers in different regions. The use of artificial intelligence (AI) and machine learning algorithms (MLAs) has become increasingly widespread, as companies seek to optimize production processes, improve decision-making, and enhance the customer experience.

One of the major trends in the machinery industry during the pandemic has been the shift toward automation and digitalization. AI and MLAs have played a significant role in this trend by enabling the development of advanced automation systems and digital solutions that help companies to increase efficiency, reduce costs, and minimize the impact of disruptions caused by the pandemic.

AI algorithms have been used to optimize production processes and reduce downtime, improving the overall efficiency of the machinery industry. To illustrate with a few examples:

1. Predictive Maintenance: Predictive maintenance is an AI algorithm that uses machine learning techniques to analyze data from machinery and predict when it is likely to fail. By detecting potential failures before they occur, companies can schedule maintenance during periods of low usage, reducing downtime and increasing efficiency.
2. Process Optimization: AI algorithms can be used to optimize production processes by analyzing data from machinery to identify bottlenecks and inefficiencies. This can include identifying the optimal production speed, reducing waste and improving overall production time.
3. Quality Control: AI algorithms can be used to monitor production processes in real-time to identify and flag any quality issues. By using image recognition, machine learning algorithms can detect deviations from product specifications, reducing the need for manual inspections and improving the accuracy of quality control.
4. Supply Chain Optimization: AI algorithms can be used to optimize supply chains by analyzing data on inventory levels, production schedules, and shipping routes to reduce waste and increase efficiency.

In addition, the use of AI and MLAs has also contributed to the development of predictive maintenance solutions that help companies to reduce unscheduled downtime and reduce the need for inventory and spare parts. This has helped companies to remain competitive and maintain profitability even in the face of the challenges posed by the pandemic.
Another trend in the machinery industry has been the increased focus on supply chain resilience and security. The conflict between Ukraine and Russia has added to the complexity of the global supply chain, making it more important for companies to have access to reliable and secure supply chains. AI and MLAs have contributed to the development of advanced supply chain management solutions.

The post Digital innovation: the game changer in uncertain times appeared first on Verhaert Masters in Innovation.

The post Digital innovation: the game changer in uncertain times appeared first on Verhaert Masters in Innovation.

Let’s start with an example of the training process of a large language model. Don’t fall off your chair reading the rest of this sentence, but that process can produce up to five times the amount of carbon emissions of an average car over its entire lifetime. Yes, you read that right: up to 5 times! Take a moment to let that sink in because this issue will only grow over time as more and more data is being collected annually and more and larger (smarter) models are being trained on this data. So what can we do to deal with this growing problem?

The premise above might sound familiar if you’ve already read the blog by Jente or seen the talk by Bart De Vos and Niels on this topic. If not, we definitely recommend doing so first to get a general scope. In this perspective we dive deeper into this issue, shifting the scope to techniques focused on reducing the energy consumption of deep learning algorithms.

Decision-making AI

Let’s start by asking a very obvious question: “Do we really need a prediction that frequently?”. A prediction is not always needed, sometimes it’s perfectly fine to wait a bit or to even skip it completely. Reducing the number of inference tasks means a reduction of computations and subsequently also the energy consumption.
For connected devices, a second question comes up: ”Where should the inference be performed?”. Both cloud and edge have their advantages and disadvantages that need to be balanced against each other to determine the best choice.

The post Designing energy-efficient deep learning algorithms appeared first on Verhaert Masters in Innovation.

The post Designing energy-efficient deep learning algorithms appeared first on Verhaert Masters in Innovation.

More data means more (and sometimes even bigger) models, which in turn increases energy consumption. We’re already aware of this rising problem, so we need to act NOW before it is too late. How do we do this?

1. We can start by optimizing the energy efficiency:

Connect the device to the cloud

In the cloud, hardware gets shared, meaning the power consumption gets centralized in one location. This gives us the opportunity to optimally use the hardware, optimize the power consumption and get large gains in return. Additionally, the necessary cooling for these large data centers can be optimized and the produced heat can even be recycled!

However, this poses scaling problems when the fleet of connected devices grows. To circumvent this, we can move the intelligence to the edge of the devices.

Reduce the energy consumption on the edge

For embedded or edge devices, the hardware doesn’t get shared. In this case, you need to maximize the ‘idle task’. Within the idle task, put the system in low power by using the low power level features of the hardware. Other possibilities are to reduce the complexity of the code or the number of computations.

2. Take into account the network quality:

Network-wise, not only the availability but also the quality (especially the latency) of the network should be considered with regard to power consumption. When the network connection is bad it takes the hardware longer to send the same amount of data, thus consuming more power. Research has even shown this relation is exponential!

So how can you tackle this latency problem? One possibility is to combine the data in batches before sending it to the cloud. An even better approach is to wait until the network connection is better, to reduce the required energy, to send these batches. You can also do parallel computing which is more efficient. But this ‘caching of data’ also brings problems along with it, i.e. you have to wait longer to get an answer back, so the Quality of Service (QoS) might be lower.

Both approaches discussed above come with advantages but also disadvantages. Each approach has its possibilities to help minimize energy usage, influencing the actual energy consumption.

3. If you can measure it, you can manage it!

So how do you decide which task to perform on either the edge or in the cloud? It might seem contradictory, but you can train an AI system with reinforcement learning to maximize the energy efficiency of each device (individually) throughout its lifetime. But in order to optimize something, you first need qualitative data and to get that data you need to measure whatever you want to optimize.

To make this concrete, let’s again look at optimizing the energy efficiency of a connected device. After deciding where to perform a task (edge or cloud), you can use the measurements of the power consumption to teach the model how good the decision was. This, however, also requires energy and time since you need to collect data and train the network. In the end, it’s a balancing exercise worth the effort, not only taking into account the technological but also the business side!

Key takeaways

1. The energy efficiency can be optimized, either in the cloud or on the edge.
2. The network quality, especially the latency, is also a key factor for energy consumption.
3. With the use of AI, the optimal load balancing can be determined for each connected device (individually) to optimally use its battery.

Any questions or want to know more about this topic? Watch the complete presentation during our InnoDays webinar or get in touch!

The post The future of energy-efficient AI systems appeared first on Verhaert Masters in Innovation.

The post The future of energy-efficient AI systems appeared first on Verhaert Masters in Innovation.

Personalization and optimization of consumer interaction

Engaging customers and increasing convenience, user experience, and personalization are essential to staying competitive in today’s market. However, this can be difficult to achieve without the right tools and technology. AI and data can help resolve this pain point by providing personalized recommendations, detecting individuals, demographics, or behavior patterns, and adjusting to them in real time. This allows companies to create a more engaging and personalized experience for their customers, which can in turn lead to increased customer satisfaction and loyalty.

Another pain point for FMCG companies is the guarantee of optimal quality and proper usage of their appliances. When appliances are deployed in the field, it can be challenging to ensure that they are functioning properly and delivering the best possible experience to customers. AI can help resolve this pain point by fine-tuning and optimizing local process parameters in real time. This allows appliances to automatically check and adjust the dispensed product, ensuring the best possible quality for customers.

Real-time appliance optimization

In addition to these specific pain points, AI and data can also help FMCG companies in a number of other areas. For example, advanced user interfaces can help increase customer engagement through more intuitive and fun user experiences. AI can also be used to drive value in a fleet of smart appliances, providing usage pattern detection, predictive maintenance services, and automated refilling services.

Furthermore, AI can improve churn, demand prediction, logistics, and inventory management. According to a McKinsey report, AI systems can reduce 65% of lost sales due to out-of-stock situations. The report also highlights that AI can reduce 10 to 50% of warehousing costs. Thus, AI forecasting can help organizations increase revenue by enhancing sales and cutting down inventory costs.

AI can also be used for market analytics and pattern discovery. Companies are often sitting on mountains of data, but it can be difficult to find the hidden patterns that can increase product placement, evaluate new product launches, or determine when to delist a product. AI systems can help by categorizing both distribution and retail stores based on location, seasonality, and demographics.

Source: Beer Hawk – PerfectDraft app

Finally, customer sentiment analysis is another area where AI and data can help FMCG companies. Understanding the true feelings of customers is invaluable, but dealing with large amounts of webshop reviews or sifting through customer complaints can be tedious. AI can help automatically detect sentiment in written and spoken language, as well as facial and emotional expressions, providing valuable insights into customer sentiment.

In conclusion, the use of AI and data can help FMCG companies resolve a number of pain points, from personalization and gamification of customer interactions to real-time appliance optimization and smart appliances. By leveraging the power of AI and data, FMCG companies can create a more engaging and personalized experience for their customers, leading to increased satisfaction and loyalty.

Curious about what else AI can do? Check out our services page!

The post Maximizing customer engagement: AI and data in FMCG appeared first on Verhaert Masters in Innovation.

The post Maximizing customer engagement: AI and data in FMCG appeared first on Verhaert Masters in Innovation.

Different approaches

Let’s start with the most basic approach, reactive maintenance, in which you only react and carry out maintenance when something breaks. This approach maximizes the usage of a machine but can result in unforeseen breakdowns. In a production line reactive maintenance typically isn’t preferred, as an unforeseen breakdown can bring production to a halt.

An approach that deals with this is preventive maintenance, in which you’ll carry out maintenance on a schedule. The idea is to ensure nominal machine performance by doing maintenance before machines break. The great benefit here is that you can schedule when your machine will be unusable. The cost, however, is that you are performing maintenance early.

If you remember the title, you already know where this is going. To prevent the added cost of preventive maintenance, you can use machine learning to give insights in the state of the machine. Based on these insights, the predictive maintenance approach allows you to schedule maintenance only when it’s necessary.

Possible machine insights

There are many different ways to gain useful information for maintenance. One approach is based on failure classifications. Depending on how you frame this task, you can not only predict if a machine will fail within a time frame but also what type of failure it’ll be and what caused it.

But what if you want to have an exact prediction of a machine that could break down? Look no further. Failure regression allows you to predict when the machine will most likely fail and know well in advance when maintenance is necessary or how long you can postpone halting the machine.

Not all failures are predictable though. What if someone uses a machine the wrong way or something hits the machine in a vulnerable place? For this, you need to somehow know if your machine is working correctly right now. This task can be done through anomaly detection algorithms which will tell you at any moment if a machine is still working correctly or not.

A final category is survival analysis, which models the machine’s degradation process. This way you can gain insights in how the machine degrades and which activities or environments make it degrade faster or slower. This approach allows you to plan maintenance depending on the degradation level, as well as learn how to optimally use the machine. And by doing so, you increase the time between maintenance and make the machine more profitable.

Predictive maintenance requirements

Predictive maintenance all starts with data and the infrastructure to get this data. Any relevant data that indicates the correct functioning is interesting, including generic information from the machine, the gauges and other sensory equipment that is already available. People can see, hear or even feel if something is wrong or if the machine requires maintenance, like rust, creaking, or vibrations. These are all things that can be measured either directly or through virtual sensors, for instance by placing cameras with algorithms to detect rust or microphones for sounds and vibrations. This data corresponds to the state of the machine.

To learn how machines degrade, you also need to know how they have been used. Through work order and inventory usage data, for example, algorithms can learn that a period of heavy usage has a greater effect on the degradation than the holidays. More often than not, data from your CRM and ERP systems contain more useful information than the sensory equipment on the machines themselves, especially if you combine the data of all identical machines. That way you can reuse the same models on multiple machines and acquire more data to train and improve your models.

So how will you collect all that data? A good IoT infrastructure is going to be crucial. Installing such an IoT system would involve placing additional sensors, cameras and microphones, among others, and then connecting these to a network. Through this network, the data from these devices can be collected, cleaned and stored in a server, and then be used to create algorithms for predictive maintenance. Even if you don’t necessarily need or want predictive maintenance, such an infrastructure would be a great investment because the benefits go much further, like for data analyses or enhancing the work of monitoring teams.

Although the investment in the right infrastructure can be rather large, the potential benefits of predictive maintenance can be even larger. You can boost your overall equipment effectiveness. You can increase availability by reducing unplanned and planned stops. You can monitor the performance of your machines more closely and increase the overall machine performance quality. All while reducing the number of resources used and waste produced, bringing you closer to lean manufacturing.

Of course, an easily replaceable tool isn’t worth the investment of a predictive maintenance algorithm. So where do you start? By determining what kind of predictive maintenance your business could benefit from. If you are looking for guidance in this innovation journey, check out our AI services or get in touch!

The post Predictive maintenance – know when things will break appeared first on Verhaert Masters in Innovation.

The post Predictive maintenance – know when things will break appeared first on Verhaert Masters in Innovation.

Collecting datasets

In our previous blog, we defined different steps to discover your AI opportunities. Once you’ve identified the process you want to automate and the information you hope to obtain, you’ll need data to feed the model. These are the camera images, audio signals, text messages, or sensor measurements the model will analyze to provide you with answers to your questions. Whether you are looking to predict the stock market or develop a medical application, having low-quality, biased or unreliable data makes the task impossible. Take for example a study on blood oxygenation levels that fails to consider the difference in sensor response of the pulse oximeter between patients with different skin colors. This would significantly reduce the probability of detecting occult hypoxemia in black patients compared to white patients.

Problem understanding is indispensable to producing a valuable dataset. Your team should understand the variability relevant to defining the problem in practice. Often, people tend to overly bias the dataset toward the most accessible data. A self-driving car whose algorithms are trained only on roads the developers happen to travel regularly is not robust. Not entirely unlike humans, ML algorithms might find it challenging to assess unknown situations. For Machine Learning models, this results in unpredictable model outcomes because machine learning models can’t learn outside the data. So high volumes of information gathered in various circumstances are crucial to developing a trustworthy algorithm.

Finetuning the labeling process

The importance of data as a crucial building block in a machine learning project is gaining recognition. However, apart from raw, high-quality data, a machine learning project is built upon the data labels. They’re the ground truth of your model and represent the outcome your model should output. Think of it like this, a parent won’t just point at items to show their baby, they will also say the name of the item. This way the baby will learn to recognize and name these items in its surroundings. With a machine learning algorithm, this is exactly the same.

Obtaining labels can be complex and labor-intensive. Machine vision problems often require manual labeling for specific objects in each image. Depending on the application, the human labelers must have the appropriate qualifications to label medical scans, images of technical defects or any other specific image type.

Some things to consider during the labeling process:

• Different labeling requirements come at different prices. Only requesting a classification label for the complete image is a tenth of the cost of delineating all instances in the picture. The figure below illustrates different labeling approaches.
• While developing the model, it pays off to evaluate the current weaknesses so you know which labels you need to improve. Knowing what the model struggles with allows you to maximize the return on new data.
• When you outsource the labeling task to specialized companies, these are critical suppliers. Your team should treat them as such. You should monitor their results adequately. Too often, the perceived simplicity of the task makes people forget to define strict and well-thought-out quality metrics on the results.

Illustration of different label types. Point annotation (top left) costs less than full mask labelling (bottom right).
Squiggles (top right) and bounding box annotation are in between these extremes. (Source)

Maximizing the return

A dataset of delineated images is necessary to build a model to delineate objects. Currently, techniques are being developed to train models based on weakly supervised data. These techniques aim to use latent information in cheaper, low-information labels to prepare models for high-information output. In the classical approach, models require the same level of information in the labels and the desired result. This is expensive, so you’ll need a human to provide you with ‘examples’ of this valuable output.

Whether you are building an algorithm to read text documents or you are building a self-driving car, the message is the same. You don’t just need data, you need a high volume of qualitative data in all relevant circumstances and you should definitely not forget to gather qualitative labels. Do this and you’ll be one step closer to the optimal solution for your next ML project. Interested in learning more? Subscribe to our AI blog mail or visit the AILab page.

This article was co-written by Jan Alexander.

The post ML’s elephant in the room: data labeling appeared first on Verhaert Masters in Innovation.

The post ML’s elephant in the room: data labeling appeared first on Verhaert Masters in Innovation.

How can a machine do that? Well, a complete answer to that is extremely complex and we could hardly explain it to you in detail in a single blog article. Simply put, artificial intelligence (AI) enables machines to learn from various experiences and perform human-like actions/tasks in a wide variety of contexts. You may have heard of chess-playing computers, Alexa and Siri, and self-driving cars as very well-known AI examples. They all rely heavily on deep learning, which allows computers to be trained to perform specific tasks, by processing large amounts of data and detecting patterns in that data. In this article, we will not go into detail about what AI is, as you can already find a lot of good material on the Internet. Rather, we would like to discuss the benefits and risks of artificial intelligence that exceed human imagination and what this could mean for all of humanity.

As per usual, let’s begin with the positive first.

The power

1. Manual tasks can be automated so that AI can perform frequent, large-scale tasks in a computerized manner. In this way, a lot of time is saved, which increases the quality of the final product.
2. AI gives intelligence to the product. Most products you know today can be improved by AI capabilities. Most likely Siri, which you have been using for years, has significantly improved over from 3, 5, or 7 years ago. Its automation and conversation features have become much more accurate. This is also true for many other products/areas such as smart cameras, security intelligence, investment analytics, and many others. Deloitte’s Global artificial intelligence Industry whitepaper predicts that the world will experience $15.7 trillion in AI-driven GDP growth by 2030. The paper primarily outlines the application of AI technology in major cities around the world and the profound changes it will bring to various industries such as finance, education, digital government, healthcare, autonomous driving, retail, manufacturing, and smart cities once it reaches the commercialization stage. AI applications are continuing to develop across organizations and industries, as shown in the figure below.

   
   Source: www.auditboard.com/blog/what-are-risks-artificial-intelligence

3. AI has the potential to adapt through progressive learning algorithms. Its algorithms acquire capabilities through certain structures and regularities in the massive data to which they have access. An algorithm can teach itself to play different games, recommend products to you online based on your tastes, or instantly recognize a threat at home (security) or on the road (autonomous vehicles). It can do all this thanks to the huge amounts of data it consumes, and its improvement comes from the fact that more and more data is being collected each day that passes.
4. Artificial intelligence analyzes data much deeper and penetrates hidden layers. This is very useful when it comes to detecting fraud, which seemed impossible years ago. Detecting fraud in a dynamic global environment with an overwhelming amount of data and traffic to monitor can be extremely difficult. Using AI to detect fraud has enabled organizations to improve internal security, as it has become possible to effectively prevent financial crime. Now, large volumes of transactions can be analyzed to uncover fraud trends, which can lead to real-time fraud detection. If fraud is suspected, AI models can reject transactions altogether or flag them for further investigation.
5. The accuracy AI provides using deep neural networks is incredible. Take your interactions with Alexa and Google. Did you know that they are all based on deep learning? You have noticed that the more you use them, the more accurate they become. When medicine started incorporating AI into its work, it could not have imagined (perhaps only hoped) that it would now be able to detect cancer in medical images.
6. The accuracy that AI provides using deep neural networks is incredible. For example, when medicine began incorporating AI into its work, it could not have imagined (perhaps only hoped) that it would now be able to detect cancer in medical images. AI in medicine uses machine learning models to search for accurate medical data and provide insights that make doctors’ jobs easier. Thanks to current advances, AI has quickly entered the healthcare field and has become a fundamental tool for medical professionals when it comes to making decisions about treatments, medications, and other patient needs.
7. Data alone means nothing, although it is a great asset when algorithms are self-learning. Experts in the field will tell you that you have all the answers you need on the data that is available to you. All you need to do is apply AI to find them. As many have now realized, data is of tremendous importance as it is a major competitive advantage for businesses today. In a competitive industry, data wins when you have the best and most of it.
8. Using AI and machine learning in your business means reducing the workload of your employees and eliminating unnecessary processes that would have slowed down your growth. In this way, companies can respond to demands on a larger scale with greater efficiency. An example of this is the Illinois Department of Innovation and Technology, which implemented AI into its processes during the Covid era. It had a chatbot that had over 20 million conversations with virtual agents, with a success rate of over 90%. Ultimately, this meant that employees were unburdened during such a critical time as the Covid pandemic, when companies were faced with a completely new reality, allowing them to use their human resources more effectively.

It is important to say that as the benefits of AI increase, so do the risks. Let’s explore that.

The risks

1. Developing a machine that simulates and sometimes surpasses human intelligence comes at a high cost. Ensuring you have the best AI requires a lot of time and resources.
2. Unfortunately, much of the data currently available tend to be biased toward men, heterosexuals, and white people due to the injustices in our history. We must balance data in a fair way so that our AI is not fixed on one gender, sexual orientation, ethnicity, etc. Since AI systems will make their decisions based on the data they were trained with, and if that data is biased towards, say, a particular gender, that could mean that a person of the opposite gender will most likely get an unfair outcome. In a fair world, we need to train AI to fight discrimination against people, not further it.
3. Some people argue that AI and machine learning lack creativity. It can learn an unimaginable amount of processes and information, but so far it has not been able to think outside the box. The approach it takes lacks a human touch because it relies only on previously fed data. However, one could argue why a machine should be more creative than a human, but that is a topic for another discussion.
4. One of the biggest concerns discussed by the vast majority of people regarding AI is that it is slowly replacing several repetitive tasks with bots. This automatically means that in the next few years the need for human intervention will greatly diminish, leading to the elimination of many jobs. Truth be told, in every phase of industrialization, people had to learn to adapt to the new rules, and this time will be no different (we hope). What worries people is the fact that this evolution is happening very fast, so people do not have a chance to adapt to this new reality as quickly as they wished. A study by McKinsey predicts that AI will replace at least 30 percent of human labor by 2030. We hope that new jobs will also be created by then.
5. While job loss is one of the most pressing issues related to AI disruption, it is only one of many potential risks. In a paper titled “The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation,” 26 researchers from 14 different institutions came up with a list of other threats that could cause serious harm – or at least a little chaos – in less than five years.“Malicious use of AI could threaten digital security (e.g. through criminals training machines to hack or socially engineer victims at human or superhuman levels of performance), physical security (e.g., non-state actors weaponizing consumer drones), and political security (e.g. through privacy-eliminating surveillance, profiling, and repression, or through automated and targeted disinformation campaigns).” In terms of surveillance, there is no better example than China’s “Orwellian” use of facial recognition technology in offices, schools, and other public places.We can only guess whether AI will one day appear to be a fair trade-off for greater security, even if it is nefariously exploited by malicious actors.The same is true for the so-called audio and video forgeries (deepfakes) that result from the manipulation of voices and likenesses. By using AI, an audio clip of any politician could be manipulated to make it look like that person made racist or sexist remarks, when in fact they said nothing of the sort. If the quality of the clip is high enough to deceive the public and not be detected, it could destroy a person, or a political campaign and end up jeopardizing global security.And all it takes is one success. After that, no one knows what is real and what is not. That leads to not being able to trust your own eyes and ears, and not being able to rely on what we thought was the best possible evidence in the past. And that can be a big problem.
6. As AI automates tedious and repetitive tasks, one would say that it has the power to make people even lazier than they’d otherwise be. Since everything is automated, we no longer have to solve puzzles or memorize information, which means our brains are used less and less. This ease can have great negative effects on future generations. On the other hand, this could be a hidden advantage, because by eliminating repetitive tasks, people can finally have the time and opportunity to devote themselves to more creative tasks, which is precisely what we are good at!
7. The lack of transparency and black box algorithms can be problematic. The main purpose of AI systems is to make predictions, and as such, algorithms can be so immensely complex that even those who developed the algorithm cannot explain exactly how the variables combine to produce the resulting prediction. This lack of transparency is why some algorithms are referred to as “black boxes” and why lawmakers are now beginning to explore what checks and balances need to be put in place. For example, if a bank customer is rejected based on an AI prediction of their creditworthiness, companies run the risk of not being able to explain why.
8. And now, we’ve arrived at the unclear legal responsibility issue. Given the potential risks of AI discussed so far, these concerns lead to the question of legal responsibility. If an AI system is designed with fuzzy algorithms and machine learning refines the decision-making itself, who is legally responsible for the outcome? Is it the company, the development team, or the system? This risk is not theoretical – in 2018, a self-driving car hit and killed a pedestrian. In that case, the car’s human passenger was inattentive and was held responsible for the AI system’s failure. This also leads us to the topic of ethics.
9. Every time a discussion about artificial intelligence comes up, someone raises the issue of ethics and morality. These are indeed very important human characteristics that are very difficult to incorporate into AI. As AI grows and evolves, it may evolve in a direction that is unpredictable to us (as previously mentioned), as we lose control of it and eventually erase human notions. When the late physicist, Stephen Hawking, visited Portugal, he also mentioned that AI’s impact could be cataclysmic unless its rapid development is strictly and ethically controlled. “Unless we learn how to prepare for, and avoid, the potential risks,” he explained, “AI could be the worst event in the history of our civilization.”

Overall

The risks of artificial intelligence are significant, but the use of these technologies and their growth are also inevitable. (Some) of the benefits, we have mentioned in this article go beyond simple efficiencies to include a fairer decision-making scenario when algorithms are trained to avoid bias. It is clear the massive potential it has in creating a fairer, better, and simpler world to live in. As we learn more about artificial intelligence, we should pay attention to the key features of AI systems:

• AI systems should include clear design documentation.
• Machine learning should include testing and refinement.
• AI control and governance should take precedence over algorithms and efficiency.

We all, without exception, have a responsibility to learn more about the risks of AI to make sure it doesn’t get out of hand. The issues and the problems around artificial intelligence will not go away, and the risks will continue to grow and change as the technology becomes more advanced and ubiquitous. Companies that embrace the three points above will be better able to manage the risks of AI systems that could otherwise have devastating legal and reputational consequences. That is why having a specialized team help you take the first steps in implementing AI systems in your organization is critical.

Note: This article was written by Aleksandra Korzh and Pedro Oliveira, our colleagues from Load Digital. A warm thank you to them.

Want to read more on artificial intelligence? Check out our other blogposts!

The post The power and risks of AI appeared first on Verhaert Masters in Innovation.

The post The power and risks of AI appeared first on Verhaert Masters in Innovation.