Less is more: How ‘chain of draft’ could cut AI costs by 90% while improving performance - Related to cut, i've, draft’, investment, us

Less is more: How ‘chain of draft’ could cut AI costs by 90% while improving performance

A team of researchers at Zoom Communications has developed a breakthrough technique that could dramatically reduce the cost and computational resources needed for AI systems to tackle complex reasoning problems, potentially transforming how enterprises deploy AI at scale.

The method. Called chain of draft (CoD), enables large language models (LLMs) to solve problems with minimal words — using as little as of the text required by current methods while maintaining or even improving accuracy. The findings were .

“By reducing verbosity and focusing on critical insights, CoD matches or surpasses CoT (chain-of-thought) in accuracy while using as little as only of the tokens, significantly reducing cost and latency across various reasoning tasks,” write the authors. Led by Silei Xu, a researcher at Zoom.

Chain of draft (red) maintains or exceeds the accuracy of chain-of-thought (yellow) while using dramatically fewer tokens across four reasoning tasks, demonstrating how concise AI reasoning can cut costs without sacrificing performance. (Credit: .

How ‘less is more’ transforms AI reasoning without sacrificing accuracy.

COD draws inspiration from how humans solve complex problems. Rather than articulating every detail when working through a math problem or logical puzzle, people typically jot down only essential information in abbreviated form.

“When solving complex tasks — whether mathematical problems. Drafting essays or coding — we often jot down only the critical pieces of information that help us progress,” the researchers explain. “By emulating this behavior, LLMs can focus on advancing toward solutions without the overhead of verbose reasoning.”.

The team tested their approach on numerous benchmarks, including arithmetic reasoning (GSM8k), commonsense reasoning (date understanding and sports understanding) and symbolic reasoning (coin flip tasks).

In one striking example in which Claude Sonnet processed sports-related questions, the COD approach reduced the average output from tokens to just tokens — a reduction — while simultaneously improving accuracy from to .

Slashing enterprise AI costs: The business case for concise machine reasoning.

“For an enterprise processing 1 million reasoning queries monthly. CoD could cut costs from $3,800 (CoT) to $760, saving over $3,000 per month,” AI researcher Ajith Vallath Prabhakar writes in an analysis of the paper.

The research comes at a critical time for enterprise AI deployment. As companies increasingly integrate sophisticated AI systems into their operations, computational costs and response times have emerged as significant barriers to widespread adoption.

Current state-of-the-art reasoning techniques like (CoT), which was introduced in 2022. Have dramatically improved AI’s ability to solve complex problems by breaking them down into step-by-step reasoning. But this approach generates lengthy explanations that consume substantial computational resources and increase response latency.

“The verbose nature of CoT prompting results in substantial computational overhead, increased latency and. Higher operational expenses,” writes Prabhakar.

Implementing AI efficiency: No retraining required for immediate business impact.

What makes COD particularly noteworthy for enterprises is its simplicity of implementation. Unlike many AI advancements that require expensive model retraining or architectural changes, CoD can be deployed immediately with existing models through a simple prompt modification.

“Organizations already using CoT can switch to CoD with a simple prompt modification,” Prabhakar explains.

Building on these developments, the technique could prove especially valuable for latency-sensitive applications like real-time customer support, mobile AI, educational tools and. Financial services, where even small delays can significantly impact user experience.

Industry experts suggest that the implications extend beyond cost savings, however. By making advanced AI reasoning more accessible and affordable, COD could democratize access to sophisticated AI capabilities for smaller organizations and resource-constrained environments.

As AI systems continue to evolve. Techniques like COD highlight a growing emphasis on efficiency alongside raw capability. For enterprises navigating the rapidly changing AI landscape, such optimizations could prove as valuable as improvements in the underlying models themselves.

“As AI models continue to evolve, optimizing reasoning efficiency will be as critical as improving their raw capabilities,” Prabhakar concluded.

The research code and data have been made publicly available on GitHub, allowing organizations to implement and test the approach with their own AI systems.

Co-founder and CEO of Perplexity AI. Aravind Srinivas, has joined Astra as an angel investor, lending his expertise and resources to the AI startup. F...

Since the launch of the Gemini Flash model family, developers are discovering new use cases for this highly efficient family of models. Gemini .

What TSMC's $165 billion investment in the US may mean for the chip industry

During a press conference at the White House Monday, US President Donald Trump and Taiwan Semiconductor Manufacturing firm (TSMC), the world's largest chip manufacturer, showcased that TSMC will spend $100 billion in the US in coming years to build multiple chip factories, on top of $65 billion already committed to US investment.

TSMC, in a press release. Billed the combined $165 billion investment as the "largest single foreign direct investment in US history."

The spending is expected to focus on "advanced technologies," which could be taken to include chips for artificial intelligence, which has largely been done by TSMC in its Taiwan factories until now. TSMC serves just about every chip maker in the world, including producing the most powerful chips from Nvidia for AI, the Hopper and Blackwell GPU chips.

Also: Intel touts new Xeon chip's AI power in bid to fend off AMD, ARM advances.

Trump stated the move means "The most powerful AI chips in the world will be made right here in America," .

"Through this expansion. TSMC expects to create hundreds of billions of dollars in semiconductor value for AI and other cutting-edge applications," said the company. "TSMC's expanded investment is expected to support 40,000 construction jobs over the next four years and create tens of thousands of high-paying, high-tech jobs in advanced chip manufacturing and R&D," it said.

TSMC already has a factory in Phoenix. Arizona, that began producing chips last year and which employs more than 3,000 people on 1,100 acres of land. The company plans to add three more US factories and an R&D center, said TSMC.

TSMC's announcement comes as chip-maker Intel, which has struggled for years with declining sales and lost market share, has been seeking clients for its own factories in the US.

A Reuters research Monday stated that two of the world's biggest AI chip makers, Nvidia and Broadcom, both competitors to Intel, are nevertheless testing out Intel's factories to manufacture their chips, citing two unnamed data.

Also: Best of MWC 2025: The 7 most impressive products you don't want to miss.

The tests, the article states. Suggest the companies are "moving closer to determining whether they will commit hundreds of millions of dollars' worth of manufacturing contracts to Intel." Another Intel rival, Advanced Micro Devices, is also expressed to be considering using Intel's factories, though it's unclear if the firm has conducted tests.

Intel's deals could be impacted, however, by continued delays in the business's manufacturing process, which has lagged TSMC's for years. The Reuters findings notes that the so-called 18A chip manufacturing method, Intel's most cutting-edge, in the recent past suffered yet another six-month delay, citing two unnamed data and. Documents.

The 18A process is "taking longer than anticipated," write Reuters's Max Cherney and Fanny Potkin. As a result, "Without the qualified fundamental building blocks of intellectual property that small and mid-size chip designers rely on, a swath of potential clients would be unable to produce chips on 18A until at least mid-2026," they write.

Amidst Intel's struggles, TSMC has had discussions with the Trump administration about absorbing some of Intel's US factories. Multiple news information have reported. Broadcom has reportedly also considered purchasing some of Intel's chip-design assets.

Of course there is randomness in GPT-4o’s outputs. After all, the model samples from a probability distribution when choosing each token. But what I d...

Si l’Apple Intelligence débarque enfin en France, disponible via la version bêta de l’iOS . Apple semble toujours faire face à un retard au niveau...

Electronic Arts (EA) donne un nouveau souffle à la franchise Command & Conquer en publiant son code source sous licence GPL. Cette initiative vise à r...

One of the best budget laptops I've tested also has a battery that lasts days on end

ZDNET's key takeaways Acer's Aspire 14 AI is available now at Costco for $699.

It has a fantastic battery and some of the latest hardware: all for a very approachable price point.

I suggest opting for the OLED. Since the low-end display isn't the best. View now at Costco.

I went hands-on with Acer's Aspire Go 15 last year, and. Praised it as one of the best cheap laptops you can get, priced at around $280. This year, one of the latest additions to Acer's Aspire line, the Aspire 14 AI, is a few steps up in terms of hardware and elements, but maintains a competitive price point of $699.

Acer's brand messaging has positioned itself as at the forefront of eco-friendly components and post-consumer recycled materials, and the aesthetic of its laptops from the past few years reflects this.

Also: Lenovo's solar-powered laptop at MWC stole the show for me - and. It's surprisingly practical.

From the exposed grill on the hinge, to the visible copper heat pipes on the bottom of the device, to the thick bezels surrounding the display, the Aspire 14 AI isn't the sleekest laptop around. Still, it reflects an unpretentious, eco-conscious aesthetic with a solid suite of hardware that presents up ready to work.

It aspects two Thunderbolt 4 USB-C ports that can support an external monitor at an 8K resolution, an HDMI , support for Wi-Fi 6E and. Bluetooth , and a Kensington lock slot. This spread of ports makes it a capable business laptop for remote or hybrid workers, particularly when paired with the battery.

Battery life. However, is the Aspire 14 AI's best feature. I know we've been singing the praises of many laptops with marathon battery life over the past year, but Acer's Aspire 14 AI fills a niche: an AI-ready budget laptop with an absolute marathon battery that will last for days on end.

During my testing, I brought the Aspire 14 AI into the office on a full charge and. Used it intermittently throughout the day. It was still above 90% by the end of the workday, and when I closed it up and brought it back out the next day, it hadn't dropped a single percentage point.

The next day, I brought it back in and used it quite a bit more aggressively: holding multiple videocalls. Doing more sustained multitasking, and keeping the display jacked up. It finally started to dent the battery life, but by the end of the day, it was still just right around 30%.

Like other power-efficient devices in its class, the Aspire 14 AI's power consumption drops to a trickle when it's not being pushed too hard and. Virtually stops depleting altogether when asleep and not in use. This makes it perfect for all-day use.

Also: This is one of the best affordable OLED laptops I've tested - and it's on sale for $668.

The fantastic battery life is enabled thanks partly to the Intel Core Ultra 7 (Series 2) processor, which has performed very well on other machines I've tested. Like the Asus Zenbook S 14 and the Dell XPS 13. With Acer's 14 AI, the trend continues, as this device not only competes with these two laptops but slightly exceeds them.

The device's physical form factor is non-descript, with an aluminum shell on the top and. Bottom and a lightweight plastic frame in a graphite finish. The keyboard is fine, without any noteworthy characteristics, although the keys feature a very short travel distance and are on the mushy side.

Depending on how it's configured. The Aspire 14 AI comes with either a 14-inch WUXGA IPS touch display or an OLED. The IPS display is functional but not particularly noteworthy in terms of brightness or color gamut. The OLED, on the other hand, makes it feel more premium.

The Aspire 14 AI is well-stocked in terms of memory, with up to 32GB of LPDDR5x RAM and. 1TB of PCIe Gen 4 SSD, making it a solid option for any mainstream workflow. As its name indicates, the Aspire 14 AI is expressly designed for AI tasks, delivering up to 40 TOPS and the usual suite of AI-specific capabilities that come with other Copilot+ PCs.

Also: This budget Lenovo 2-in-1 I recommend to students and professionals is cheaper than ever.

Additionally, the Intel Arc GPU supports up to 53 TOPS, with upscaling technology for gaming and. enhanced performance when it comes to graphics rendering and video editing.

That being mentioned, while gaming on the Aspire 14 AI is well-supported by the hardware, the IPS display may hold back the experience. Therefore, I'd suggest opting for the OLED configuration if you have any notions about gaming on this laptop.

Acer's Aspire 14 AI is one of the superior bang-for-your-buck options available right now. Starting at $699, it offers some of the latest high-performing hardware and a battery that will last multiple days in a very competitive package.

If you're looking for a "budget plus" laptop with respectable hardware that opts for a more utilitarian form factor (and approachable price). The Aspire 14 AI is a solid choice. I would suggest opting for the OLED configuration if you're willing to pay for the upgrade, especially if you'll be using it for visually creative work or gaming.

India’s digital landscape is rapidly evolving, and. Application programming interfaces (APIs) are playing a crucial role in this transformation. They s...

A recent survey of 1,050 CIOs revealed that 93% of IT leaders will implement AI agents in the next two years. With IT leaders worki...

Les attaques par deepfakes se multiplient de façon inquiétante. Selon une enquête de Regula, près de 49 % des entreprises ont déjà été confrontées à d...