State-of-the-art artificial intelligence systems like OpenAI’s ChatGPT, Google’s Gemini and Anthropic’s Claude have captured the public imagination by producing fluent text in multiple languages in response to user prompts. Those companies have also captured headlines with the huge sums they’ve invested to build ever more powerful models.
An AI startup from China, DeepSeek, has upset expectations about how much money is needed to build the latest and greatest AIs. In the process, they’ve cast doubt on the billions of dollars of investment by the big AI players.
I study machine learning. DeepSeek’s disruptive debut comes down not to any stunning technological breakthrough but to a time-honored practice: finding efficiencies. In a field that consumes vast computing resources, that has proved to be significant.
Where the costs are
Developing such powerful AI systems begins with building a large language model. A large language model predicts the next word given previous words. For example, if the beginning of a sentence is “The theory of relativity was discovered by Albert,” a large language model might predict that the next word is “Einstein.” Large language models are trained to become good at such predictions in a process called pretraining.
Pretraining requires a lot of data and computing power. The companies collect data by crawling the web and scanning books. Computing is usually powered by graphics processing units, or GPUs. Why graphics? It turns out that both computer graphics and the artificial neural networks that underlie large language models rely on the same area of mathematics known as linear algebra. Large language models internally store hundreds of billions of numbers called parameters or weights. It is these weights that are modified during pretraining. https://www.youtube.com/embed/MJQIQJYxey4?wmode=transparent&start=0 Large language models consume huge amounts of computing resources, which in turn means lots of energy.
Pretraining is, however, not enough to yield a consumer product like ChatGPT. A pretrained large language model is usually not good at following human instructions. It might also not be aligned with human preferences. For example, it might output harmful or abusive language, both of which are present in text on the web.
The pretrained model therefore usually goes through additional stages of training. One such stage is instruction tuning where the model is shown examples of human instructions and expected responses. After instruction tuning comes a stage called reinforcement learning from human feedback. In this stage, human annotators are shown multiple large language model responses to the same prompt. The annotators are then asked to point out which response they prefer.
It is easy to see how costs add up when building an AI model: hiring top-quality AI talent, building a data center with thousands of GPUs, collecting data for pretraining, and running pretraining on GPUs. Additionally, there are costs involved in data collection and computation in the instruction tuning and reinforcement learning from human feedback stages.
All included, costs for building a cutting edge AI model can soar up to US$100 million. GPU training is a significant component of the total cost.
The expenditure does not stop when the model is ready. When the model is deployed and responds to user prompts, it uses more computation known as test time or inference time compute. Test time compute also needs GPUs. In December 2024, OpenAI announced a new phenomenon they saw with their latest model o1: as test time compute increased, the model got better at logical reasoning tasks such as math olympiad and competitive coding problems.
Advertisement
Slimming down resource consumption
Thus it seemed that the path to building the best AI models in the world was to invest in more computation during both training and inference. But then DeepSeek entered the fray and bucked this trend.
DeepSeek sent shockwaves through the tech financial ecosystem.
Their V-series models, culminating in the V3 model, used a series of optimizations to make training cutting edge AI models significantly more economical. Their technical report states that it took them less than $6 million dollars to train V3. They admit that this cost does not include costs of hiring the team, doing the research, trying out various ideas and data collection. But $6 million is still an impressively small figure for training a model that rivals leading AI models developed with much higher costs.
The reduction in costs was not due to a single magic bullet. It was a combination of many smart engineering choices including using fewer bits to represent model weights, innovation in the neural network architecture, and reducing communication overhead as data is passed around between GPUs.
It is interesting to note that due to U.S. export restrictions on China, the DeepSeek team did not have access to high performance GPUs like the Nvidia H100. Instead they used Nvidia H800 GPUs, which Nvidia designed to be lower performance so that they comply with U.S. export restrictions. Working with this limitation seems to have unleashed even more ingenuity from the DeepSeek team.
DeepSeek also innovated to make inference cheaper, reducing the cost of running the model. Moreover, they released a model called R1 that is comparable to OpenAI’s o1 model on reasoning tasks.
They released all the model weights for V3 and R1 publicly. Anyone can download and further improve or customize their models. Furthermore, DeepSeek released their models under the permissive MIT license, which allows others to use the models for personal, academic or commercial purposes with minimal restrictions.
Resetting expectations
DeepSeek has fundamentally altered the landscape of large AI models. An open weights model trained economically is now on par with more expensive and closed models that require paid subscription plans.
The research community and the stock market will need some time to adjust to this new reality.
STM Daily News is a vibrant news blog dedicated to sharing the brighter side of human experiences. Emphasizing positive, uplifting stories, the site focuses on delivering inspiring, informative, and well-researched content. With a commitment to accurate, fair, and responsible journalism, STM Daily News aims to foster a community of readers passionate about positive change and engaged in meaningful conversations. Join the movement and explore stories that celebrate the positive impacts shaping our world.
AI Spacecraft Propulsion: Machine Learning’s Role in Space Travel
AI Spacecraft Propulsion: Discover how AI and machine learning are transforming spacecraft propulsion systems, from nuclear thermal engines to fusion technology, making interplanetary travel faster and more efficient.
Machine learning is a branch of AI that identifies patterns in data that it has not explicitly been trained on. It is a vast field with its own branches, with a lot of applications. Each branch emulates intelligence in different ways: by recognizing patterns, parsing and generating language, or learning from experience. This last subset in particular, commonly known as reinforcement learning, teaches machines to perform their tasks by rating their performance, enabling them to continuously improve through experience. As a simple example, imagine a chess player. The player does not calculate every move but rather recognizes patterns from playing a thousand matches. Reinforcement learning creates similar intuitive expertise in machines and systems, but at a computational speed and scale impossible for humans. It learns through experiences and iterations by observing its environment. These observations allows the machine to correctly interpret each outcome and deploy the best strategies for the system to reach its goal. Reinforcement learning can improve human understanding of deeply complex systems – those that challenge the limits of human intuition. It can help determine the most efficient trajectory for a spacecraft heading anywhere in space, and it does so by optimizing the propulsion necessary to send the craft there. It can also potentially design better propulsion systems, from selecting the best materials to coming up with configurations that transfer heat between parts in the engine more efficiently.In reinforcement learning, you can train an AI model to complete tasks that are too complex for humans to complete themselves.
Reinforcement learning for propulsion systems
In regard to space propulsion, reinforcement learning generally falls into two categories: those that assist during the design phase – when engineers define mission needs and system capabilities – and those that support real-time operation once the spacecraft is in flight. Among the most exotic and promising propulsion concepts is nuclear propulsion, which harnesses the same forces that power atomic bombs and fuel the Sun: nuclear fission and nuclear fusion. Fission works by splitting heavy atoms such as uranium or plutonium to release energy – a principle used in most terrestrial nuclear reactors. Fusion, on the other hand, merges lighter atoms such as hydrogen to produce even more energy, though it requires far more extreme conditions to initiate.Fission splits atoms, while fusion combines atoms.Sarah Harman/U.S. Department of Energy Fission is a more mature technology that has been tested in some space propulsion prototypes. It has even been used in space in the form of radioisotope thermoelectric generators, like those that powered the Voyager probes. But fusion remains a tantalizing frontier. Nuclear thermal propulsion could one day take spacecraft to Mars and beyond at a lower cost than that of simply burning fuel. It would get a craft there faster than electric propulsion, which uses a heated gas made of charged particles called plasma. Unlike these systems, nuclear propulsion relies on heat generated from atomic reactions. That heat is transferred to a propellant, typically hydrogen, which expands and exits through a nozzle to produce thrust and shoot the craft forward. So how can reinforcement learning help engineers develop and operate these powerful technologies? Let’s begin with design.The nuclear heat source for the Mars Curiosity rover, part of a radioisotope thermoelectric generator, is encased in a graphite shell. The fuel glows red hot because of the radioactive decay of plutonium-238.Idaho National Laboratory, CC BY
Reinforcement learning’s role in design
Early nuclear thermal propulsion designs from the 1960s, such as those in NASA’s NERVA program, used solid uranium fuel molded into prism-shaped blocks. Since then, engineers have explored alternative configurations – from beds of ceramic pebbles to grooved rings with intricate channels.The first nuclear thermal rocket was built in 1967 and is seen in the background. In the foreground is the protective casing that would hold the reactor.NASA/Wikipedia Why has there been so much experimentation? Because the more efficiently a reactor can transfer heat from the fuel to the hydrogen, the more thrust it generates. This area is where reinforcement learning has proved to be essential. Optimizing the geometry and heat flow between fuel and propellant is a complex problem, involving countless variables – from the material properties to the amount of hydrogen that flows across the reactor at any given moment. Reinforcement learning can analyze these design variations and identify configurations that maximize heat transfer. Imagine it as a smart thermostat but for a rocket engine – one you definitely don’t want to stand too close to, given the extreme temperatures involved.
Reinforcement learning and fusion technology
Reinforcement learning also plays a key role in developing nuclear fusion technology. Large-scale experiments such as the JT-60SA tokamak in Japan are pushing the boundaries of fusion energy, but their massive size makes them impractical for spaceflight. That’s why researchers are exploring compact designs such as polywells. These exotic devices look like hollow cubes, about a few inches across, and they confine plasma in magnetic fields to create the conditions necessary for fusion. Controlling magnetic fields within a polywell is no small feat. The magnetic fields must be strong enough to keep hydrogen atoms bouncing around until they fuse – a process that demands immense energy to start but can become self-sustaining once underway. Overcoming this challenge is necessary for scaling this technology for nuclear thermal propulsion.
Reinforcement learning and energy generation
However, reinforcement learning’s role doesn’t end with design. It can help manage fuel consumption – a critical task for missions that must adapt on the fly. In today’s space industry, there’s growing interest in spacecraft that can serve different roles depending on the mission’s needs and how they adapt to priority changes through time. Military applications, for instance, must respond rapidly to shifting geopolitical scenarios. An example of a technology adapted to fast changes is Lockheed Martin’s LM400 satellite, which has varied capabilities such as missile warning or remote sensing. But this flexibility introduces uncertainty. How much fuel will a mission require? And when will it need it? Reinforcement learning can help with these calculations. From bicycles to rockets, learning through experience – whether human or machine – is shaping the future of space exploration. As scientists push the boundaries of propulsion and intelligence, AI is playing a growing role in space travel. It may help scientists explore within and beyond our solar system and open the gates for new discoveries. Marcos Fernandez Tous, Assistant Professor of Space Studies, University of North Dakota; Preeti Nair, Master’s Student in Aerospace Sciences, University of North Dakota; Sai Susmitha Guddanti, Ph.D. Student in Aerospace Sciences, University of North Dakota, and Sreejith Vidhyadharan Nair, Research Assistant Professor of Aviation, University of North Dakota This article is republished from The Conversation under a Creative Commons license. Read the original article.
Dive into “The Knowledge,” where curiosity meets clarity. This playlist, in collaboration with STMDailyNews.com, is designed for viewers who value historical accuracy and insightful learning. Our short videos, ranging from 30 seconds to a minute and a half, make complex subjects easy to grasp in no time. Covering everything from historical events to contemporary processes and entertainment, “The Knowledge” bridges the past with the present. In a world where information is abundant yet often misused, our series aims to guide you through the noise, preserving vital knowledge and truths that shape our lives today. Perfect for curious minds eager to discover the ‘why’ and ‘how’ of everything around us. Subscribe and join in as we explore the facts that matter. https://stmdailynews.com/the-knowledge/
Now, that futuristic vision has gained some serious thrust. Archer Aviation — one of the leading players in electric vertical take-off and landing (eVTOL) aircraft — has announced a major move that could change how the city thinks about air mobility.
Archer Takes Control of Hawthorne Airport
In a landmark deal, Archer announced plans to acquire control of Hawthorne Airport — just three miles from LAX — for approximately $126 million in cash.
The 80-acre site, home to 190,000 square feet of hangars and terminal facilities, will become the company’s operational hub for its Los Angeles air-taxi network and a testbed for AI-driven aviation technology.
Alongside the purchase, Archer raised an additional $650 million in new equity funding, bringing its liquidity to more than $2 billion — a strong signal that the company is serious about turning concept into concrete.
What This Means for LA’s Mobility Future
This isn’t just a real estate move. It’s a strategic infrastructure play.
If Los Angeles is to handle Olympic crowds and long-term congestion, new vertical mobility hubs are essential. Hawthorne could serve as the first of several vertiports forming a network across the metro area.
It also puts Archer in a prime position to work alongside city planners and mobility partners preparing for the LA28 Games — potentially transforming how visitors move between venues, airports, and downtown.
Caution: Not Quite “Jetsons” Yet
While this progress looks promising, it’s not smooth skies ahead just yet.
FAA certification remains the biggest hurdle; only about 15% of compliance documentation has been approved. Production and scaling still pose risks — building and maintaining a fleet of electric aircraft at commercial levels isn’t cheap. Public acceptance will matter too. Even the quietest aircraft need to earn the city’s trust for noise, cost, and safety.
Advertisement
Still, compared to even a year ago, the vision of air taxis over Los Angeles feels far less like science fiction.
A Step Toward the Olympic Future
Archer’s move aligns perfectly with the question we raised earlier:
Can Los Angeles turn the 2028 Olympics into a showcase for sustainable, futuristic transportation?
By securing its own hub near LAX and backing it with fresh capital, Archer seems determined to make that answer a yes. Whether passengers will be hailing flying taxis in time for LA28 remains uncertain, but the groundwork — both financial and physical — is clearly being laid.
The skies over LA might just get busier — and cleaner — in the years to come.
Dive into “The Knowledge,” where curiosity meets clarity. This playlist, in collaboration with STMDailyNews.com, is designed for viewers who value historical accuracy and insightful learning. Our short videos, ranging from 30 seconds to a minute and a half, make complex subjects easy to grasp in no time. Covering everything from historical events to contemporary processes and entertainment, “The Knowledge” bridges the past with the present. In a world where information is abundant yet often misused, our series aims to guide you through the noise, preserving vital knowledge and truths that shape our lives today. Perfect for curious minds eager to discover the ‘why’ and ‘how’ of everything around us. Subscribe and join in as we explore the facts that matter. https://stmdailynews.com/the-knowledge/
Rod: A creative force, blending words, images, and flavors. Blogger, writer, filmmaker, and photographer. Cooking enthusiast with a sci-fi vision. Passionate about his upcoming series and dedicated to TNC Network. Partnered with Rebecca Washington for a shared journey of love and art. View all posts
Slate Automotive captured national attention earlier this year when it unveiled what many called the most anticipated “budget” electric pickup truck in America. Promising a minimalist design, domestic manufacturing, and a base price under $20,000 (after incentives), the Slate Truck was positioned as the EV industry’s boldest answer to the affordability problem.
But since its April 2025 debut, several developments have reshaped that story — including pricing adjustments, production plans, and questions about whether “affordable” will still apply once federal incentives fade.
🚨 Slate Auto’s $20K Electric Truck Is No More — Here’s Why
⚙️ From Concept to Production
In April, Slate Auto revealed its small two-door electric pickup — a compact, customizable EV designed for simplicity over luxury. The company’s philosophy is centered around what it calls the “Blank Slate” concept: a base model stripped of unnecessary features but built for expansion.
Base range: ~150 miles, with an optional battery upgrade to ~240 miles
Length: ~175 inches (roughly the size of a compact SUV)
Body style: 2-door truck, with a conversion kit planned for a 5-seat SUV variant
Manufacturing site: Warsaw, Indiana — a repurposed 1.4-million-square-foot former printing plant
When Slate’s founders — backed by investors including Jeff Bezos and Mark Walter (Guggenheim Partners) — launched the concept, they confidently pitched a price “under $20,000 after incentives.”
However, recent developments have changed that equation. The loss of a key federal EV tax credit under recent legislation means the base price now sits closer to $27,000 before incentives. Even with state-level rebates, the total cost will likely land in the mid-$20K range for most buyers.
That’s still lower than most EVs on the market, but Slate’s base model is extremely minimal: manual windows, no touchscreen infotainment, and unpainted exterior panels in the entry trim. The company argues that the simplicity keeps prices low and durability high — echoing the utilitarian design of early pickups.
“We don’t believe an affordable EV should start at $60,000,” a Slate spokesperson said during the reveal. “Our truck is for people who want a reliable tool, not a gadget.”
🧩 Reservations and Early Demand
According to TechCrunch, Slate logged over 100,000 $50 refundable reservations within two weeks of launch — an impressive early show of interest.
That figure, however, does not guarantee actual orders. As seen with other EV startups, reservation enthusiasm doesn’t always translate into deliveries. Still, with $700 million in investor funding and a clear U.S. manufacturing plan, Slate’s prospects appear stronger than many early EV challengers.
🏭 Building in America
The company’s decision to set up shop in Indiana is strategic. It provides central U.S. access to suppliers and a lower-cost workforce compared to coastal hubs. The plant conversion is underway, and Slate aims to ramp up to 150,000 units annually by 2027, according to industry reporting.
If successful, the Slate Truck could become the first mass-produced electric pickup under $30K built entirely in the U.S.
🚦 What It Means for Affordable EVs
Slate’s progress comes at a pivotal moment for electric mobility. As other manufacturers focus on high-margin luxury vehicles, the affordable-EV space has thinned out. Slate’s entry signals a renewed interest in accessible electrification — but also highlights the fragile balance between price, policy, and practicality.
Advertisement
If production holds, the Slate Truck could mark the beginning of a new chapter for everyday EV ownership — proof that electric doesn’t have to mean expensive.
Rod: A creative force, blending words, images, and flavors. Blogger, writer, filmmaker, and photographer. Cooking enthusiast with a sci-fi vision. Passionate about his upcoming series and dedicated to TNC Network. Partnered with Rebecca Washington for a shared journey of love and art. View all posts
