Connect with us

AI Research

Rushing into genAI? Prepare for budget blowouts and broken promises – Computerworld

Published

2 months ago

on


“What you’re going to find on the input and output tokens is a dramatic difference in price,” Suda said. “The build cost is quite low to get into the game, but once you begin using it, the costs go up. Say you have 400 users to start. By year four, you may have 2,000 users.”

“So, what happens? You’re consuming more and your costs goes up four times by year four,” he said.

“GenAI is not like Google, but some organizations use it like Google — you go into it and ask a question and get an answer. That doesn’t really happen,” Suda continued. “You get an answer and often think, ‘That’s not quite what I wanted.’ And so that makes them want to ask another question. That can multiply your cost quickly.”

The hidden costs that can add up

Thermo Fisher Scientific’s Kwiecien said one cost that hasn’t been considered involves testing. “Every time you ask a question and test it, that’s a cost,” she said. “I’m not just going to load that 500 time,s because that will cost me every time.

“We need to test how often AI gives good answers to common questions like ‘What’s the recruiting process?’ or ‘Where’s my 401(k) info?’” she said. “But each test costs money, so we have to balance accuracy with cost and decide how many times to test to be confident in the results.”

Thermo Fisher is currently using a virtual chatbot from ServiceNow, and hopes to make it more intuitive by adding a genAI layer. As a result, it’s currently eyeing genAI solutions from Microsoft, IBM and others.

Another cost can come with efforts to use genAI in hiring. Amy Ritter, vice president for Talent Acquisition at Thermo Fisher, said the company implemented a genAI-powered hiring app from Phenom to automate parts of its global manufacturing hiring platform. The company then had to invest in job preview videos to show candidates what it’s like to work at Thermo Fisher — covering the environment, required PPE, and key skills — since recruiters weren’t involved early in the process.

The cost of change management is also often overlooked, Ritter said. “We invested time and money visiting sites, engaging leaders, and building buy-in, which paid off with strong adoption at launch,” she said.

Injecting Phenom’s genAI into its HR hiring platform, however, netted big returns, Ritter said. It cut candidate screening time from 16 days to a just 7 minutes. Along with automating interview scheduling, cumulatively Thermo Fisher is saving over 8,000 hours a year in candidate screening, 12,000 hours in scheduling time and filling roles 10% faster, Ritter said.

And there are infrastructure costs — the cost of building out, running and maintaining server farms, including managed service, is also often underestimated, according to AWS’s Hennesey. “One insurance customer had 200 [proofs of concept] running, but couldn’t articulate the expected value — most were just experiments. Our advice: clearly define the problem, align it with organizational goals, and measure expected returns,” he said.

Moving from pilot to production can also be a soft spot for costs, as can shifting from on-prem to the cloud; the latter means new services and pricing models that need to be understood and forecast.

AWS’s Bedrock, Microsoft’s Azure AI Studio, Google’s Cloud Vertex AI, IBM’s Watson.ai and Cohere’s Platform are all fully managed service offerings that allow AI developers to build apps using top foundation models via a single API — no infrastructure management needed. “You pay on a per model, on a per region basis,” Hennesey said. “And, then you have to think about tokens.”

Making a “capacity commitment” to a vendor can cuts costs. So instread of buying capacity “on demand” organizations can make an LLM capacity commitment for a specific amount of time – whether one month or six months – and deliver up to a 60% savings, Hennesey said.

The bottom line: there’s still a lot of uncertainty around the cost of genAI projects because the technology is still in its early days — and still evolving.

“I feel like we’re not getting great answers, because people are unsure how it’s going to be used,” Kwiecien said. “And so it’s hard to understand what your usage may look like in the future, because we can’t tell how fast it’s going to take people to flip to that.

“How fast are we going to get the solutions to really answer the way that we want it to answer?” she said.



Source link

Related Topics:
Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

AI Research

Artificial Intelligence Of Things (AIoT) Guide

Published

52 minutes ago

on

September 4, 2025

By


For organizations exploring connected technologies, the conversation around the Internet of Things (IoT) has shifted. They are looking to make devices smarter, responsive and capable of operating with greater insight.

The Artificial Intelligence of Things (AIoT) combines the sensor-driven, networked structure of IoT with the decision-making power of artificial intelligence (AI) at the edge to collect and act on data. This generates insights to help businesses improve efficiency and reliability and streamlines user experience.

Explore what AIoT is, its benefits and use cases below.

What is AIoT?

AIoT is the convergence of AI with IoT. The technology brings intelligence to devices that connect and transmit data, transforming them into systems that interpret information, learn from patterns and support decision-making.

The key components of AIoT systems include:

  • Sensors and actuators: Sensors interact with the surroundings by collecting and measuring data, while actuators translate instructions from the AI system into actions.
  • Connectivity: Connectivity solutions enable communication between devices.
  • Edge computing: Devices equipped with embedded AI chips or processors handle data processing to improve response times.
  • AI models: AI algorithms analyze patterns, detect anomalies, forecast trends and support automated decision-making. These models are trained on historical or real-time data and deployed at the edge or in the cloud.
  • Cloud infrastructure: The cloud stores large volumes of data and supports remote device management and software updates.

The difference between IoT and AIoT

IoT systems collect data from sensors, transfer it through networks and store it in the cloud for later analysis. These systems may need oversight to make decisions based on the data collected. This creates delays and increases dependence on bandwidth.

AIoT embeds intelligence closer to where data is generated. It reduces latency and the bandwidth required to move data. By analyzing and interpreting data at the edge, it streamlines decision-making, allowing businesses to react faster to changing conditions.

AIoT architecture

Explore how AIoT’s architecture allows it to collect, analyze and share information.

Edge computing AIoT

Edge-based computing has the following layers:

  • Collection terminal layer: This includes sensors, cameras and mobility devices that gather information from surrounding environments.
  • Connectivity layer: This layer consists of hardware and software that transmits data to nearby edge devices.
  • Edge layer: Edge AI processors handle real-time analytics, pattern recognition and insight generation.

Cloud-based AIoT

Cloud-based AIoT layers include:

  • Device layer: This layer includes beacons, sensors, embedded devices and tags that collect and transmit data.
  • Connectivity layer: The connectivity layer comprises software and hardware that facilitate secure and reliable data transfer to the cloud.
  • Cloud layer: AI workloads are run here for analysis and insights.
  • User communication layer: The user communication layer interfaces with dashboards, apps or web portals for the end user.

Benefits of integrating AI in IoT devices

Implementing a new technology should drive measurable outcomes in performance, cost and improve user experience. Here are the benefits of connecting devices to AIoT.

1. Boosts operational efficiency

AIoT increases efficiency by enabling systems to respond to conditions in real time. When decisions are made at the edge, devices can automatically adjust behaviors.

This means:

  • Heating, ventilation, and air conditioning (HVAC) systems regulate airflow based on occupancy patterns.
  • Supply chains reroute based on traffic or inventory signals.
  • Machinery modulates speed or power use in response to production demand.
  • The result is smarter resource use, better coordination and fewer interventions.

2. Enhances safety

AIoT improves safety by enabling instant decision-making, which helps prevent accidents.

For example:

  • Edge-enabled cameras detect unauthorized access or unusual motion patterns and trigger alarms immediately.
  • Equipment shuts down or enters safe mode if abnormal vibration or heat is detected.
  • Smart infrastructure manages traffic flows or alerts authorities to emergencies before a hazard escalates.

3. Reduces human error

By automating data interpretation and decision-making, AIoT reduces the room for human error in routine or high-risk tasks. In sectors where conditions change rapidly, AI-enabled devices adjust irrigation or voltage levels automatically.

In administrative settings, AIoT monitor systems and flag inconsistencies or risks that might be overlooked. It supports teams by handling the repetitive, high-precision tasks where mistakes may slip through.

4. Improves decision-making

With access to real-time data, historical trends and contextual cues, AIoT recommends or takes action based on potential outcomes. For example, in manufacturing, AIoT identifies production bottlenecks and suggests changes to scheduling. In smart cities, traffic data combined with weather and event schedules drive predictive rerouting. This capability augments human judgment to streamline the gathering and analysis of insights.

5. Streamlines predictive maintenance

By monitoring equipment health and comparing it to past performance data, AI models detect subtle signs of wear, imbalance or malfunction before failure occurs. This helps prevent costly downtime. Additionally, technicians are alerted to service needs when they are required, rather than relying on fixed schedules or post-failure repair.

6. Provides scalability

As more devices are added, the system becomes more intelligent. New sensors or devices feed data into shared models, enriching the entire system. Firmware and AI models are updated remotely, and because intelligence resides at the edge, the system doesn’t become dependent on a single hub. This flexibility allows businesses to scale while maintaining performance and control.

7. Maximizes data value

AIoT systems extract and turn data into meaningful, actionable outputs. This reduces data overload while uncovering patterns some analytics tools might miss. For example, temperature fluctuations combined with vibration readings in a cold storage unit might signal compressor strain. AIoT flags this before damage occurs. It also enables real-time dashboards for operations teams, providing visibility into system health and performance metrics.

8. Helps manage energy

The system enables low energy use by allowing devices to self-optimize based on context. For example, lighting systems in smart buildings dim or shut off based on occupancy. Industrial machines shift loads to off-peak hours, while electric vehicles and grid systems balance loads in response to demand and supply.

This level of precision reduces energy bills and supports green initiatives, helping business demonstrate their environmental responsibility.

9. Supports personalization

At the consumer and user experience level, AIoT allows for tailored and intuitive engagements. Devices learn user preferences, anticipate needs, and adjust accordingly. For instance, thermostats learn daily routines and adjust temperatures as needed. Digital signage in retail adapts based on customer flow and demographic signals, providing an exceptional user experience. Even medical devices can personalize treatment or therapy based on individual patient concerns.

AIoT brings contextual awareness in every interaction, making devices feel more aligned with consumer needs.

Considerations when implementing AIoT

AIoT has several benefits. However, businesses should keep the following considerations in mind when deploying the technology:

Data privacy and security: AIoT devices collect and process vast amounts of data, some of which may be sensitive, personal or proprietary. This makes privacy and security a priority in industries adhering to strict privacy regulations. Businesses should map out where and how data is being used, and align with security policies, regulatory requirements and industry standards.

Interoperability: Ensuring your systems are interoperable across vendors, protocols and applications can be complex in legacy arrangements. Open standards and flexible connectivity options can reduce integration friction. Businesses can partner with providers who factor in interoperability. They can offer customizable solutions that can plug into existing systems instead of requiring complete overhauls.

System complexity: Building a cohesive and balanced system requires specialized expertise. Businesses can work alongside partners who specialize in edge computing, embedded systems and connectivity integration to reduce friction in design and deployment. With the right support, complexity becomes manageable.

AIoT applications and examples

As AIoT matures, intelligent edge devices may help various sectors meet their needs. Below are the use cases of the system in different industries.

Manufacturing

Embedded AI processors in industrial equipment monitor vibration, pressure and temperature to detect early signs of mechanical failure. This makes it possible to perform maintenance before problems interrupt production. Computer vision systems powered by AI also support inspection on assembly lines, automatically flagging defects or inconsistencies in products.

These tools detect variations invisible to the naked eye, raising quality control standards while reducing manual inspection workloads.

Facilities also use smart sensors to handle energy consumption. By tracking usage patterns, models suggest optimizations or adjust systems during peak hours, helping factories control costs.

Smart home technology

In residential settings, AIoT enables homes to respond to owners’ behavior. For example, far-field voice recognition allows devices to understand commands even in noisy conditions or from across the room.

Occupancy sensors and AI-driven climate control systems work in tandem to adjust temperatures and lighting based on who’s home, what rooms are in use and time of day preferences. These features reduce energy consumption while enhancing comfort.

Biometric authentication methods like fingerprint or presence-based access help control smart locks or safes. These systems learn user routines and flag irregular access attempts or unexpected activity.

Smart cities and public infrastructure

One common use case in smart cities and public infrastructure is traffic optimization. By combining data from cameras and road sensors with AI, traffic lights adjust based on vehicle and pedestrian flow, easing congestion and improving safety.

Sensors across cities monitor air quality, noise pollution and water infrastructure. When they exceed thresholds, relevant departments use these insights to create quality standards, assess law compliance and notify the public about environmental conditions.

Retail

In retail, AIoT is driving a shift toward more responsive, tailored in-store experiences while giving businesses visibility into their operations. Smart cameras and shelf sensors monitor customer movement, identify high-traffic zones and optimize product placement. For example, interactive kiosks, enhanced by touch interfaces, adjust content based on user engagement, providing dynamic recommendations or support. These systems personalize the experience and reduce the need for additional staff.

Smart shelves equipped with weight sensors and computer vision detect when items are low or misplaced, automatically triggering restocking processes or alerting staff. This reduces out-of-stock scenarios and makes inventory manageable.

Health care

Wearable devices equipped with AI monitor vital signs and detect patterns that might indicate a problem before symptoms worsen. These insights are then shared with health care providers, enabling early intervention.

In hospitals, edge computing AI devices monitor patients in intensive care units. These systems process data to flag anomalies, supporting faster response times in intervention.

Diagnostic imaging is another area where AIoT is gaining ground. Devices embedded with AI analyze scans for signs of disease, accelerating diagnosis and supporting radiologists by flagging subtle indicators they might miss.

Agriculture

In agriculture, AIoT systems help farmers operate precisely and sustainably. Smart soil sensors monitor moisture, nutrient content and pH levels, sending this data to processors that determine optimal watering or fertilization schedules. This reduces the overuse of resources and helps crops grow under optimal conditions.

Computer vision applications also scan plants for signs of disease or pest infestation so that farmers use targeted interventions, preventing widespread crop damage.

Weather stations powered by AI models analyze local setting patterns and historical trends to recommend planting schedules or harvesting times. This leads to better crop yield, less waste and efficient processes.

Telecommunications

Telecom providers operate vast and decentralized infrastructure, making AIoT valuable for network visibility and uptime. AI-enabled edge monitoring systems track performance indicators to help provide reliable service to end users. These systems help automate routine diagnostics and preemptively schedule maintenance.

Additionally, smart energy systems manage consumption based on load and demand, supporting sustainability and lowering operational costs.

Energy

In energy and utilities, smart grids collect data and feed it into AI models that analyze consumption patterns, detect anomalies and make recommendations for efficiency improvements.

Distributed energy resources like solar panels or wind turbines use edge-based AI to optimize performance. For instance, processors might adjust the tilt angles of solar panels in response to changing light conditions.

Transportation and logistics

In autonomous or semi-autonomous vehicles, edge AI systems handle decision-making, detect obstacles, adjust routes and improve driver safety. These systems process visual and sensor data to avoid delays associated with sending everything to the cloud.

In warehousing operations, AI-driven systems manage inventory and adjust to supply demands, automate stock replenishment and optimize floor layouts. This boosts throughput without requiring constant human oversight.

Implement IoT AI solutions

Synaptics offers a comprehensive portfolio of advanced human interface solutions designed to help businesses implement AIoT. Our sensing, processing and connectivity systems help you create devices that respond in real time.

With robust connectivity options across wireless standards and protocols, we ensure your devices stay responsive and reliable wherever they’re deployed. Our solutions are flexible and scalable, and are optimized for low-power, high-performance settings.



Source link

Continue Reading

AI Research

Research has shown that people who do not use AI technology more than those who are well aware of ar..

Published

3 hours ago

on

September 4, 2025

By


According to a study by the University of Southern California in the United States, “The less AI you understand, the more magical AI you feel.”

AI-generated image of a college student who is disappointed with low grades [Production = Gemini]

Research has shown that people who do not use AI technology more than those who are well aware of artificial intelligence (AI) technology. In addition, there are studies showing that excessive use of Generative AI tools such as ChatGPT is linked to lower academic achievement, and it is analyzed that dependence on AI should be vigilant.

According to a study by researchers at the University of Southern California in the United States and the University of Bocconi in Italy on the 4th, the lower the understanding of AI, the more often they accept it as magic and use it.

The researchers evaluated 234 university undergraduate students with their understanding of AI (literacy), then gave them writing tasks on a specific topic and investigated whether to use Generative AI tools.

As a result, students with lower scores in AI understanding showed a stronger tendency to use AI for tasks. The researchers analyzed, “People with low understanding of AI perceive AI like magic,” and “They are likely to be in awe when AI performs tasks that were thought to be a unique attribute of humans.”

Conversely, people with a high understanding of AI know that AI works based on computer algorithms, not magic, so they don’t rely too much on it.

In March this year, a study found that sincere students use fewer Generative AI tools, and that AI dependence can lead to a decrease in self-efficacy and academic achievement. The researchers investigated and analyzed the frequency of AI learning and self-efficacy of learning after the end of the semester in 326 undergraduate students.

Sundas Azim, a professor at SZABIST University in Pakistan who conducted the study, said, “In the case of tasks conducted by students relying on Generative AI, AI produced similar responses, resulting in less classroom participation or discussion activities.” As a result, students with more AI use tended to have relatively lower average GPA.

It is analyzed that services such as ChatGPT can be effective when they need immediate help in their studies, but can have a negative impact on long-term learning and achievement.



Source link

Continue Reading

AI Research

Alberta Follows Up Its Artificial Intelligence Data Centre Strategy with a Levy Framework

Published

3 hours ago

on

September 4, 2025

By


Alberta is introducing a levy framework for data centres powering artificial intelligence technologies, the Province recently announced.

Effective by the end of 2026, a 2% levy on computer hardware will apply to grid-connected data centres of 75 megawatts or greater, according to a statement from Alberta.

The levy will be fully offset against provincial corporate income taxes, the government says. Once a data centre becomes profitable and pays corporate income tax in Alberta, the levy will not result in any additional tax burden.

Data centres of 75MW or greater will be recognized as designated industrial properties, with property values assessed by the province. Land and buildings associated with data centres will be subject to municipal taxation.

The framework was forged through a six-week consultation with industry stakeholders, according to Nate Glubish, Minister of Technology and Innovation.

“Alberta’s government has a duty to ensure Albertans receive a fair deal from data centre investments,” the provincial minister remarked. “This approach strikes a balance that we believe is fair to industry and Albertans, while protecting Alberta’s competitive advantage.”

Glubish added that the Alberta government is also exploring other options. This includes a payment in lieu of taxes program that would allow companies to make predictable annual payments instead of fluctuating levy amounts, as well as a deferral program to ease cash-flow pressures during construction and early years of operation.

“After working closely with industry, we’re introducing a fair, predictable levy that ensures data centres pay their share for the infrastructure and services that support them,” commented Nate Horner, Minister of Finance.

“This approach provides stability for businesses while generating new revenue to support Alberta’s future,” he posits.

The decision builds on the Alberta Artificial Intelligence Data Centre Strategy, introduced in 2024.

The strategy aims to capture a larger share of the global AI data centre market, which is expected to exceed $820 billion by 2030 as Alberta becomes a data centre powerhouse within Canada.

However, the Province’s tactics have not gone uncriticized.



Source link

Continue Reading

Trending