Tech Brew keeps business leaders up-to-date on the latest innovations, automation advances, policy shifts, and more, so they can make informed decisions about tech.

Tech Brew

General Artificial Intelligence

Jordyn Grzelewski

emergingtechbrew.com

Cyclists and drivers are often at loggerheads over who has more rights on the road. Could an AI tool help bring peace to these warring factions?

That’s what a new program in Sacramento aims to do: By keeping parked cars out of bike lanes, the effort aims to improve the transit experience for both parties.

The city of Sacramento and the Sacramento Regional Transit District (SacRT) in December debuted an automated bus stop enforcement program using technology from 

. Now, in what officials say is a first, they’re monitoring bike lanes using camera systems that have been installed on 100 local buses.

“I think it’s a great use of the technology,” Charles Territo, Hayden’s chief growth officer, told Tech Brew. “From a safety perspective, the agency is really getting more bang for the buck. For pedestrians that are having to navigate vehicles parked at bus stops, things are going to improve. And for bike riders who have to navigate people parking in bike lanes, that’s going to change, as well. And ultimately what we’re hoping to see is a change in driver behavior.”

Jessica Gonzalez, a spokesperson for SacRT, told us that the program has garnered overwhelmingly positive feedback.

Data from the early days of this new approach to enforcement shows the scope of the problem. Gonzalez said that when SacRT conducted a pilot program across a portion of its service area, 286 bus stop parking violations were identified in 70 days. Then, when the program officially rolled out, nearly 3,000 warnings were issued during a 60-day grace period. And in a one-month span after the city started issuing violations, 2,370 drivers were cited.

The problem, Gonzalez explained, is that when bus drivers are unable to park at a bus stop, some bus riders—for example, those who use mobility devices—may not be able to board the bus. Others may have to walk a distance. This creates issues around safety, 

, and reliability, Gonzalez said. And even small delays or disruptions can throw off bus routes for the rest of the day.

“We’re hopeful that this can bring about change,” she said.

With the bus stop enforcement program going well so far, the city of Sacramento and SacRT are now expanding the strategy to bike lanes.

The cameras use Hayden AI’s machine learning and AI technology to identify and record illegal parking violations. The system sends a video and photo of the vehicle’s license plate, along with the time and location of the incident, to the city, which then processes violations. The city is issuing warnings for bike lane violations; starting June 13, those violations will come with a fine.

“We’re proud to be the first city in the country to use this technology to help keep our bike lanes clear,” Staci Hovermale, Sacramento’s parking services manager, said in a statement. “This tool helps us enforce existing parking rules more effectively, improving safety for cyclists and ensuring everyone shares the road responsibly.”

Territo said that Hayden would welcome the opportunity to apply this model in other cities, but it’s not legal everywhere. The startup will be rolling out bike lane enforcement in Chicago this summer, and hopes to further expand in the future.

The startup’s tech is being used to enforce parking violations in numerous cities across the US, including Los Angeles, New York City, and Washington, DC.

“What we’re trying to do is to give the agencies and the cities we work with a tool that can be used to solve a number of parking- or curb-related challenges,” Territo said. “And I think Sacramento is a great example of how that would work.”

Sacramento pioneers AI-enabled strategy to keep cars out of bike lanes

 “Ultimately what we’re hoping to see is a change in driver behavior,” a Hayden AI executive tells Tech Brew.

Tricia Crimmins

, a new report from the International Energy Agency (IEA) took a more optimistic outlook on how AI will affect the climate crisis—and even claimed that “concerns that AI could accelerate climate change appear overstated.”

The report, which focuses on AI’s energy usage, underscored how under-equipped the electrical grid is for incoming data centers and provided recommendations for how the large amount of energy they consume could be mitigated.

One of the recommendations focused on where data centers are located: Almost half of US data centers are in “five regional clusters,” so the IEA suggested that new data centers be built “in areas of high power and grid availability.” Plus, like a 

 from Duke University, the IEA recommended that data centers operate using on-site generators and do so “more flexibly.”

And the stakes are high: “If the electricity sector does not step up,” the report said, “there is a risk that meeting data center load growth could entail trade-offs with other goals such as electrification, manufacturing growth, or affordability,” a warning 

The report was, however, more reassuring about the effects AI will have on the environment. Though it acknowledged that optimizing AI to stabilize electricity networks won’t be adequate to combat the climate crisis, the report said that “the widespread adoption of existing AI applications could lead to emissions reductions that are far larger than emissions from data centers.”

“AI can be a tool in reducing emissions,” the report said, “but it is not a silver bullet and does not remove the need for proactive policy.”

Data center emissions could be further curtailed—or at least delayed—by the recent news that Microsoft has paused data center construction projects in 

“Data center planning is a multi-year and capital-intensive program we plan for years in advance to ensure we have sufficient infrastructure in the right places,” Microsoft President of Cloud Operations and Innovation Noelle Walsh said in a LinkedIn post. “We are slowing or pausing some early-stage projects.”

Amazon data centers near houses in the Loudoun Meadows neighborhood

New IEA report suggests concerns that AI could exacerbate climate crisis ‘appear overstated’

The report also recommends that data centers be built outside areas where they’re concentrated.

The auto industry has historically operated under long development cycles.

But amid rapid technological advancements and growing competition from Chinese manufacturers that are bringing new products to market at a lightning pace, automakers are under more pressure than ever to speed things up—without sacrificing quality or cost.

Enter the go-to solution for seemingly every modern-day conundrum: artificial intelligence.

JP Clausen, EVP of global manufacturing at General Motors, recently shared with Tech Brew some of the automaker’s strategies for making the best use of AI tools in manufacturing—and he was quick to caution that AI isn’t a silver bullet.

“Not every problem in this world can be solved by AI. There are some problems that AI and machine learning are good at, but it’s not everything,” said Clausen, an alum of Google, Tesla, and Lego who stepped into the lead manufacturing role at GM last year. “Understanding the problem you’re going to solve first is probably the most important thing of the whole equation.

“Knowing your parameters…is the key point for making an AI system that is successful,” he added. “Because otherwise you’ll just boil the ocean.”

GM’s implementation of AI in manufacturing is focused on three areas: safety, quality, and efficiency.

The automaker, for example, is putting automated guided vehicles to work in plants, Clausen said, to automate processes that could be cumbersome or unsafe for employees. GM’s leaders have emphasized that they view AI and robots as complementary to workers, not as replacements.

“We want everyone to feel comfortable with the technology. It’s an opportunity, in our mind, to continue to develop our employees and make them a part of the process,” Clausen said. “Robots, AI, and people should work together and make the workplace better for everyone.”

GM is also using robotics and a proprietary AI tool to inspect the welds and paint coats on vehicles to ensure they’re done properly.

 to simulate production lines before they’re up and running, as part of an effort to scale up manufacturing faster and more cost effectively.

“You can build a whole factory. You can build all the production lines. You can build all the different work cells,” Clausen said. “And you can go down and actually simulate all the movements.”

Speed is crucial, he said, because it’s become especially important to bring new products to market quickly in the incredibly competitive EV space.

Clausen said that GM’s approach of offering a wide variety of electric models across its brands (about a dozen are slated to be on the market by the end of this year), and releasing them at a relatively quick pace, has helped boost its competitiveness. The automaker is now the No. 2 EV maker in the US.

“You can actually go out and scale your production faster if you do it with AI,” Clausen said, “simply because you have less surprises.”

 a new partnership under which the automaker will use Nvidia’s AI chips and software for its autonomous vehicle tech as well as in its plants. Nvidia’s platforms will underpin AI manufacturing models for factory planning, among other initiatives.

“This new collaboration is absolutely key for us to crank more data faster,” Clausen said. “And we definitely believe everything we are doing in manufacturing with robots, simulations, automatic traffic systems, and detecting deviations in quality systems will be further developed with this partnership.”

GM is using another AI tool in plants that integrates data from battery tests, vehicle operations, and service work to train an AI model that’s used to detect voltage anomalies in battery packs before they make their way into an EV, according to Jie Du, senior scientist manager of applied analytics and insights. It’s now in use at all of GM’s battery assembly plants.

The goal is to improve the customer experience, provide more detailed insights to ensure EV battery quality, and help bring down warranty costs.

The idea is that the algorithm can predict voltage anomalies much sooner in the battery pack assembly process, ensuring that packs with defects in them don’t end up in a vehicle before they leave the factory. Since introducing the tool last year, GM has analyzed hundreds of thousands of battery packs and found an anomaly rate of less than 0.1%.

“AI is only as good as the application of it,” Jon Francis, GM’s chief data and analytics officer, whose team leads GM’s battery manufacturing AI work, said in a statement. “At GM, we’re intentional about applying AI in ways that will have a positive impact on the business or our customers. Applying AI to battery manufacturing does exactly that.”

Correction 4/07/2025: This story has been updated to correct JP Clausen's title.

How GM is using AI to boost quality, safety, and efficiency in auto plants

“Knowing your parameters…is the key point for making an AI system that is successful,” JP Clausen, GM’s EVP of global manufacturing, tells Tech Brew.

Patrick Kulp

 brought on by ever-growing data centers, tech giants are looking for a nuclear option.

Google and Amazon both announced last week that they’ve cut deals involving the use of small modular reactors (SMRs) to eventually power data centers. In Google’s case, the company agreed to buy power from small reactors slated to be developed by Kairos Power. Amazon said days later it would invest in small reactor startup X-energy and work with two utilities on development plans in Washington and Virginia.

The news is the latest sign of Big Tech’s embrace of nuclear power as the rush to build the data centers needed for generative AI has companies getting creative with energy sources. Last month, Microsoft signed a purchase agreement that will reopen the 

 Three Mile Island nuclear power plant in Pennsylvania, and Amazon and Google both inked other nuclear energy deals this year.

While tech companies have also poured money into renewable energy sources like wind and solar, nuclear has the advantage of providing around-the-clock energy with no carbon emissions, a boon for always-on data centers. But it faces more regulatory hurdles, 

“Nuclear power is one part of that mix—it can be brought online at scale, and has a decades-long record of providing a reliable source of safe carbon-free energy,” Amazon wrote in its announcement.

The SMRs in question have yet to be approved by the Nuclear Regulatory Commission, meaning that it will likely be years before Google or Amazon’s announcements last week amount to anything concrete, according to Peter Kelly-Detwiler, co-founder of consulting firm NorthBridge Energy Partners.

“Once they do cross that [regulatory] hurdle, then the next question is, ‘Well, how quickly can you build a factory?’ And that’s going to take a while. And then once you build the factory, it’s…a question of scaling,” Kelly-Detwiler said. “The race for nuclear—the earliest that probably gets done is between 2028 and 2030.”

 The Department of Energy says that small nuclear reactors could have the advantage of being cheaper, easier to build, and more flexible location-wise. But there could also be an issue of managing the dangerous waste from so many disparate locations, Kelly-Detwiler said. The nation’s first SMR plant, operated by NuScale and backed by the DOE, ended in a high-profile failure last year after failing to 

The Biden administration has largely been an ally of the nuclear energy industry, promoting the power source as a key part of its “all of the above” energy policy and investing heavily in SMRs, despite the NuScale project. Energy Secretary Jennifer Granholm spoke at the event announcing Amazon’s SMR deals at the company’s second headquarters in Virginia last week, according to 

 The “glass half-full” view of the situation is that tech companies’ willingness to win the AI arms race at any cost could provide a much-needed push forward for modular nuclear energy, a tech that might otherwise have a hard time finding commercial investors up for such a long and uncertain ride, Kelly-Detwiler said.

“There’s a valley of death that all these modular nuclear companies have to cross, which is, how do they essentially get from where they are today, to scale and commerciality,” Kelly-Detwiler said. “[Tech giants are] relatively inelastic with respect to power, and therefore, I think we can expect them to do what they can to accelerate this new technology, which is great because modular nuclear is probably something the industry and the world needs to continue down a decarbonization path.”

Two nuclear power plant cooling towers with intricate circuit board patterns seamlessly emerging from the stacks, blending technology with industrial design.

Power-hungry data centers drive Google, Amazon to go nuclear

Google and Amazon both recently inked deals with reactor startups.

Power-hungry data centers drive tech giants to go nuclear

That’s one reason General Motors is integrating artificial intelligence into just about every aspect of its motorsports division.

Tech Brew recently got a peek behind the scenes of where the machine learning magic happens: GM’s Charlotte Technical Center in North Carolina. Jonathan Bolenbaugh, manager of data and analytics for GM Motorsports, gave us a virtual tour and explained how AI is changing the way the organization races.

It “took a leap of faith when we started the team,” he said. “Like, ‘Hey, we’re gonna use AI to change racing.’ And then people are like, ‘Why would we do that?’ Well, because it’s cool and it’s fun and I like winning races.”

Bolenbaugh—whose experience spans both racing and production vehicles—formed the data and analytics team two years ago.

The idea is to use technology to assist the teams that GM Motorsports supports across racing series like NASCAR, IndyCar, and IMSA.

Data scientists and engineers use machine learning to create models based on vehicle performance data, track conditions, and competitors’ behaviors to inform race strategy. AI also is used in real-time race analysis to help the team make decisions and improve outcomes.

“It’s literally hundreds of people per car to get that out on the track,” Bolenbaugh said. “This team was founded with the intent of using technology—specifically AI and [machine learning] and some statistical methods—to give those people the tools that they need to have the peak outcome every weekend.”

The command center at GM's Charlotte Technical Center where AI and machine learning are used on race day.

Some of the AI-powered tools may seem basic, like real-time audio transcription of conversations drivers have over the radio, but consider the challenge of getting an accurate transcript amid noisy racetrack conditions.

The team also uses image analysis to get insights into as many as tens of thousands of photos that might be captured during a typical NASCAR event, according to Bolenbaugh.

“All the categorization and derived information is available seconds later,” he said. To determine the size of a wheel, for example, a process that used to take minutes—literally using a ruler to calculate the diameter in a photo—now takes a matter of seconds.

The team employs machine learning to predict competitors’ behaviors, and then, throughout a weeklong design cycle, the team quickly makes performance-boosting tweaks to race cars.

A typical week kicks off with a Monday morning report card on how the previous weekend’s races went. By Tuesday, Bolenbaugh said, the team is already making changes to vehicle models. It all leads up to Sunday, when changes are deployed and the data and analytics team is on hand to support the race team.

“We’re having a pretty dynamic conversation while this race is running about, ‘Hey, we think that the tool is overestimating this parameter or underestimating this parameter,’ and then we can tweak that,” Bolenbaugh said.

He insisted that these tools complement the work of human engineers without replacing them.

“We’re doing these things to cover the basics so the race engineers can do the things that only the humans can do. Obviously, technology is advancing, but sitting on a pit box, making calls, understanding how that works out, that intuition and that subject matter expertise still very much matters,” he said. “It matters more than ever, because a lot of times people will take AI and ML tools and say, ‘Well, AI told me to do this, it’s gotta be right.’”

Logan McLeod, who heads up software engineering at GM Motorsports, in a presentation earlier this year spelled out the demanding nature of racing. The organization participates in over 130 races a year, supporting dozens of race team partners. AI advancements, he said, enable the team to be more efficient by freeing up resources that can be used elsewhere.

“Whether it’s in multi-physics simulation or model-based collaboration, AI represents an acceleration opportunity across everything we do in technology,” he said.

One of the goals is to improve results on the track. But it’s also to feed data, lessons, and new tech to the teams that design GM’s production vehicles. Automakers use motorsports to test out tech that may one day make their way into mass-market vehicles.

“This design, build, test, iterate cycle—we get to do that every week. And the learnings from that translate back into production cars,” Bolenbaugh said. “Retooling a plant costs a lot of money. We can retool here and try things out, and the things that work, we can send right back and make our on-the-road cars much better for our customers.”

Finish line checkered flag with binary code in each square

How GM is using AI to boost performance on the racetrack

Why use AI in racing? “Because it’s cool and it’s fun and I like winning races,” the data and analytics manager for GM Motorsports tells Tech Brew.

If a delayed bus has ever messed with your morning commute, one San Francisco-based tech startup might be able to help.

Hayden AI has developed a vision AI system that it’s deploying in numerous US cities with the goal of 

 to enforce violations when vehicles obstruct bus lanes.

Hayden AI’s system uses cameras mounted on buses to monitor bus lanes, capture images, apply algorithms to the images, and collect evidence of parking violations in bus lanes that it then sends to the relevant authorities.

“As we saw more and more cities looking to add things like bus lanes, there became a recognition that without some type of enforcement, the rules around the use of the lanes were likely to be ignored,” Charles Territo, Hayden AI’s chief growth officer, told Tech Brew.

The system is working as designed if it accomplishes four objectives: boosts on-time arrivals of buses, reduces bus collisions, decreases greenhouse gas emissions by cutting down on the amount of time buses spend idling, and increases ridership.

“The more efficient, the more reliable, the safer those routes were, the more people would choose transit,” Territo said.

All of this is particularly relevant to public transit users with disabilities, who, as Territo pointed out, are “uniquely impacted” when drivers block access to buses.

Hayden AI’s system is in use by numerous transit providers, including New York City’s Metropolitan Transportation Authority (MTA), DC’s Metro, and Los Angeles’s Metro.

The MTA reported a 5% improvement in bus lane speeds along enforced routes, as well as 20% fewer vehicle collisions. MTA officials have deemed the technology useful enough to expand its use to bus stops and to enforce double-parking violations. By the end of this year, the agency expects to have over 1,000 buses on 33 routes equipped with the tech.

“Keeping bus stops clear is critical to ensuring all of our customers can safely get on and off the bus,” Quemuel Arroyo, the MTA’s chief accessibility officer, said in a statement. In a news release announcing the expansion, Riders Alliance, a nonprofit group dedicated to improving public transit for riders in New York City, also expressed support for automated camera enforcement.

AC Transit in Oakland, California, also recently announced it was expanding its use of Hayden AI’s system to bus stops on all routes in its service area. The agency reported that the technology resulted in citation efficiency that was over 34 times better during a roughly month-long period this summer, compared to a similar period last year.

“Although our transit district anticipated an increase, the actual number of motorists cited for violating the bus-only lane laws reinforced the safety imperative to expand AI-powered camera use across our bus network,” AC Transit said in a news release.

AC Transit acknowledged some of the privacy concerns the technology could prompt, but noted safeguards—such as ensuring cameras don’t capture video inside buses and the system not having facial recognition capabilities—are in place.

 by providing city planners with more data.

“The more data that can be captured about the transit network and the impediments to that transit network behaving the way it was designed,” he said, “the better that network will be.”

Hayden AI recently completed a $90 million Series C funding round; the startup has raised more than $193 million since launching in 2019, 

. Territo said the company will invest in hardening its platform, launching new products, and expanding its international business.

How this transportation tech startup is using AI to improve public transit

“The more efficient, the more reliable, the safer these routes were, the more people would choose transit,” a Hayden AI exec tells Tech Brew.

TBR-AI and the grid Audio Embed

 to your coworkers, college essay cheats, or images of the pope in a puffy coat?

The answer has to do with the far-flung process that triggers every time someone taps ChatGPT or Google Gemini. It likely involves one or more of hundreds of nondescript buildings scattered throughout the world and loaded with racks of servers. Those behemoth structures have to be powered up 24/7 along with their water-intensive cooling systems, which require a steady flow of massive amounts of electricity.

And as businesses flock to build out a new generation of unprecedentedly compute-hungry AI systems—they’re called 

 language models for a reason—that power consumption threatens to overwhelm the country’s 

 found that the energy needed to power new data centers will grow 160% by 2030. On average, a ChatGPT query requires nearly 10x the electricity of a Google search, Goldman said.

So what can utilities and data center operators do to keep this voracious appetite from hobbling grids? Paradoxically, part of the answer might also lie in AI, albeit a different variety from the chatbots and image generators currently synonymous with the tech.

“There are a lot of opportunities, particularly around AI, to make grid operation more efficient and better,” said Heiko Claussen, co-CTO of industrial software provider Aspen Technology, which works on grid modernization efforts.

Those efforts include more accurately predicting the output of renewable sources based on weather patterns and historical performance to plan an optimal mix of supply to tap at a given time, Claussen said. AI can also be useful in improving grid maintenance: diagnosing and prioritizing alarms around problems, monitoring the condition of transformers within the grid, and dispatching repair teams, he said. There’s also potential for using AI to better manage congestion and route power to different areas, Claussen said.

Claussen said he doesn’t expect AI power demand to continue to grow linearly at its current rate forever; as chips and data centers become more efficient, it could also relieve some of the strain on grids. He also sought to draw a line between the current wave of resource-voracious generative AI and AI technology more broadly.

“AI has an unnecessarily bad rap as being power-hungry [because of generative AI],” Claussen said. “If you’re talking in our specific space, in the industrial AI space, I would argue that AI is actually helping to save power and resources. It’s a tool for optimization.”

 data centers as critical infrastructure, more people have also started to pay more attention to the inner workings of these thousands of obscure-yet-essential buildings, according to Rithika Thomas, a senior analyst at ABI Research and an architect by training who studies how data centers can operate more efficiently.

“When you think of the cloud, everybody thinks it’s an imaginary thing that nobody has to deal with, and with a lot of people talking about it, becomes a real, physical aspect of a building, actually, whether that’s a cement box or whether it’s underground, it is a physical asset that people need to think about,” Thomas said.

One of the biggest power guzzlers within a data center is the cooling system, which can account for up to half of its energy consumption, according to the trade publication 

. Thomas said some of the work to relieve this burden relies on reusing the excess heat created by data centers, whether to warm homes in the winter or even 

Thomas said data center operators have also implemented software that can make cooling operations more efficient, better account for spotty output of renewable sources, and extend the life cycle of parts within the data center.

For instance, Google recently introduced a demand response system for its data centers built around the same tech as its existing carbon-intelligent computing platform. The system is designed to reroute non-urgent compute tasks to other data centers at times of high grid stress. Demand response can also be useful for better capturing renewable sources by, say, rerouting data center activity depending on how much the sun is shining on solar panels at a given time.

Gernot Wagner, a climate economist at Columbia Business School, said demand response programs like these are one example of how AI efficiencies in conjunction with emissions pricing and climate policy can serve to help rein in the power usage of data centers.

“There will always be emissions—as long as we’re burning fossil fuels—but we can save on some of that with energy-efficiency measures, with demand response, with other intelligent investments and ways to organize ourselves as society, and AI can certainly help with that,” Wagner said. “It’s not the solution. It’s not going to get us to zero emissions on its own, but it can certainly help make us more productive, make us more efficient, and, yes, shave off some emissions as well.”

An AI robot arm plugging in a cable to an outlet on a map of the United States