Comment by Riaz Haq on August 20, 2026 at 8:29pm

@Hesamation

AT&T is doing exactly what OpenAI and Anthropic fear about enterprise.

they’re routing 40% of employee AI usage to open models, and intend to boost it to 60-70%.

> coding costs down 56%
> quality dropped just 2%
> 45B tokens/day btw.

AT&T AI chief: open models are “just as good or better” for many tasks.

they still use frontier models for the critical work but everything else gets routed to cheaper open models.

https://x.com/hesamation/status/2090518831349268851?s=46&t=Uy6j...

————————


Corporate America’s AI playbook is changing as more businesses substitute proprietary AI models in favor of open alternatives to save on costs. AT&T’s effort to remake its business for the AI era shows just how far that shift has come.

https://www.wsj.com/cio-journal/why-at-t-is-betting-big-on-open-wei...



Such open models currently power about 25% of the telecommunications company’s overall AI usage, including helping manage its network, said AT&T Chief Data and AI Officer Andy Markus, “and we believe we can get way higher than that.”


That conviction has driven Markus to experiment with a number of different open models, including powerful new models from China, which help prevent AT&T from becoming “beholden to any one solution,” he said.

The company’s focus on using open models comes at a time when U.S. companies are starting to put tighter controls around their AI spending. Companies say one way to lower AI token costs is to use open models rather than state-of-the-art proprietary ones, because they generally offer more flexibility and lower cost.

Silicon Valley meanwhile remains locked in heated debate over whether AI should be “open” or “closed.” Nvidia is among the most prominent tech giants backing open models and Meta Platforms on Monday announced a new open-weight model, Muse Glimmer.


Proprietary models from companies like OpenAI, Anthropic or Google aren’t freely available for users to download and modify, while open-source and open-weight models are. True open-source models allow full access to training data and code, while open-weight models typically share only the numerical parameters, or “weights,” that underlie them.

Comment by Riaz Haq on August 30, 2026 at 6:14pm

AI cooperation with China brings hope to Pakistan's small farmers

https://www.bignewsnetwork.com/news/279275714/ai-cooperation-with-c...

For many small farmers in Pakistan's Punjab province, the country's most populous province, farming has always felt like a roll of the dice against the whims of the weather. But local farmers may soon see their fortunes change.

Syed Muhammad Ashiq Hussain Shah, Punjab's agriculture minister, is attending the 2026 China International Big Data Industry Expo, which runs through Sunday, in Guiyang, capital of southwest China's Guizhou Province. He said his delegation has a clear mission: to find Chinese partners who can help bring AI-powered farming tools to millions of smallholders back home.

According to the Pakistani official, many small farmers in Punjab province lack access to reliable, timely information. Meanwhile, climate change is making weather patterns increasingly erratic, with sudden downpours, prolonged droughts and scorching heat waves adding new risks to agricultural production

Punjab's agricultural authorities are now developing an AI-powered digital monitoring network designed to deliver weather forecasts, crop monitoring and real-time data on soil, water and pest control directly to farmers.

"We want to use AI to monitor the weather, soil and water, and to digitally track crops as well," Shah said. "We also want to use AI for pest control, so that a farmer can simply send a picture of a plant and get help identifying the problem. "

"If farmers can receive this information through a digital platform, they can make better decisions and reduce losses. That is why we see AI as an important tool for improving agricultural productivity and supporting farmers in Punjab," he said.

"That is exactly why we are here -- to find partners and learn from China's experience," he added.

In fact, China and Pakistan, including Punjab, have already shared a solid foundation in agricultural cooperation. Shah said Punjab signed at least six memorandums of understanding following previous visits to China and has been actively following through on them.

Chinese companies have also played a significant role in Punjab's high-tech farm mechanization program. Shah estimated that about 90 percent of the machinery acquired under the program -- including harvesters and tractors -- came from China. Now, this partnership is evolving beyond hardware to include smart technologies and intelligence.

Ali Mustafa Dar, adviser to the chief minister of Punjab on artificial intelligence and special initiatives, observed that China does not treat AI as something separate from economic development, but integrates it across manufacturing, logistics, cities, agriculture, transport and public services.

"China thinks long term. It builds the infrastructure. It experiments and learns. And when something works, it has the capacity to take it to extraordinary scale," he said.

As China's first national comprehensive big data pilot zone, Guizhou offers precisely the kind of experience Punjab is looking for. Dar said Guizhou is particularly interesting because it demonstrates that AI begins with strong data foundations.

Guizhou is a key hub for China's national "East Data, West Computing" project. This national strategy operates similarly to a power grid, shifting the heavy data processing demands of China's economically bustling eastern coast to resource-rich western regions. Crucially, the project utilizes the western region's abundant green energy such as wind and hydropower, making it a major milestone in green technology and carbon reduction.

By late July 2026, Guizhou housed 50 key data centers, with total computing capacity reaching 176.57 EFLOPS. Nearly 98.4 percent of this capacity was dedicated specifically to AI-related computing.

Dar said he is keen to understand how Guizhou has built its data ecosystem -- from infrastructure, data centers and standards to governance, industry and the applications built on top.

Comment by Riaz Haq on September 3, 2026 at 9:16am

AI token prices have fallen to historic lows, with the benchmark LLM Token Expenditure Index dropping below $1 per million tokens amid intense market competition and deflationary cost pressures. [1, 2]

Market Pressures and Deflation

• Record Lows: The Silicon Data market index fell to $0.97 per million tokens, driven by aggressive price cuts from closed model providers and competition from cheaper open-weight alternatives. [1, 2]

• Profit Squeeze: “Token deflation compresses the revenue line while compute commitments stay fixed,” as noted by Quartz. This places heavy financial strain on frontier labs like OpenAI and Anthropic just as they navigate confidential IPO preparations. [1, 2]

• Oversupply Concerns: Analysts at Goldman Sachs warn that if output costs fall faster than consumption volume grows, the broader infrastructure sector could face severe compute oversupply. [1]

The Jevons Paradox and Enterprise Costs

• Surging Consumption: Cheaper tokens often stimulate much higher usage rather than lower overall spending—a classic application of the Jevons paradox where longer, multi-step agent loops become affordable. [1, 2]

• Rising Overhead: System orchestration, caching, and retrieval overhead mean that enterprise AI bills can continue to climb even as the base unit price of a token collapses. [1, 2]

Comment by Riaz Haq on September 3, 2026 at 7:55pm

South China Morning Post
@SCMPNews
US urged to consider military strikes to stop China achieving AGI first

https://x.com/SCMPNews/status/2095648893446893682?s=20

-----------------

US urged to consider military strikes to stop China achieving AGI first
Former White House staffer calls for espionage and cyber options as Washington prepares for a potential Chinese breakthrough

https://www.scmp.com/news/us/article/3366284/us-urged-consider-mili...

The United States should start preparing for scenarios where extreme measures must be taken to stop China from achieving artificial general intelligence (AGI), according to a former White House official, including state-backed espionage and military strikes on Chinese data centres.
Jacob Stokes, deputy director of the Indo-Pacific Security Program at the Centre for a New American Security (CNAS), said at an online event on Thursday that various US agencies, including the Department of Defense and the National Security Agency, should begin assessing what intelligence they need to justify taking such actions.

“Trying to think through the particulars of that will be especially important, in part because it will help policymakers … start to work backwards based on the unique nature of the technology, in the same way that in a past era, policymakers would learn about nuclear weapons and … work backwards from the science to the policy implications,” he said.

Comment by Riaz Haq on September 3, 2026 at 8:13pm

HUMAIN Unveils humain-m3, a Frontier Arabic Language Model Developed by MiniMax (Chinese LLM) , in Research Preview on HUMAIN Node

https://finance.yahoo.com/technology/ai/articles/humain-unveils-hum...

Commissioned by HUMAIN and delivered by MiniMax, humain-m3 advances Arabic AI performance across seven public benchmarks and is available in research preview today through HUMAIN Node
RIYADH, Saudi Arabia, Sept. 3, 2026 /PRNewswire/ -- HUMAIN, a PIF company delivering full-stack artificial intelligence capabilities globally, today at LEAP announced humain-m3, a frontier Arabic-language model commissioned by HUMAIN and delivered by MiniMax. The model is available in research preview today through HUMAIN Node, HUMAIN's platform providing developers, researchers and enterprises access to advanced AI models and inference capabilities.

Built on the MiniMax-M3 lineage, humain-m3 is a 428-billion-parameter mixture-of-experts model further pre-trained on more than one trillion tokens of Arabic-native content. In evaluations across seven public Arabic benchmarks, humain-m3 achieved the highest average score among the frontier models tested, demonstrating strong performance across Arabic language understanding and reasoning.


The launch brings together two important elements of HUMAIN's intelligence strategy: developing world-class AI capabilities for Arabic and making frontier intelligence more accessible through HUMAIN Node.

HUMAIN Node serves as an access point for enterprises to advanced AI models and scaled inference, connecting developers and organizations with the intelligence required to build, test and deploy AI applications. Through Node, HUMAIN is supporting its own growing portfolio of AI capabilities while creating an ecosystem through which leading models from HUMAIN and around the world can be accessed and deployed.

humain-m3 will initially be available through HUMAIN Node as a research and evaluation preview, enabling researchers and developers to test its capabilities and provide feedback ahead of a planned open-weight release.
"Arabic is spoken by hundreds of millions of people, yet it remains significantly underrepresented at the frontier of artificial intelligence," said Tareq Amin, CEO, HUMAIN. "With humain-m3, we are investing in changing that. And through HUMAIN Node, we are making that intelligence accessible so developers, researchers and innovators can experiment with it, build on it and create the next generation of Arabic AI experiences."

Comment by Riaz Haq on September 3, 2026 at 8:21pm

Ultimate Guide - Best Open Source LLM for Urdu in 2026
Elizabeth C.
Our definitive guide to the best open source LLMs for Urdu in 2026. We've partnered with industry insiders, tested performance on multilingual benchmarks, and analyzed architectures to uncover the top models that excel in Urdu language processing. From state-of-the-art multilingual models to specialized language understanding systems, these models demonstrate exceptional capabilities in Urdu text generation, translation, and comprehension—helping developers and businesses build powerful Urdu AI applications with services like SiliconFlow. Our top three recommendations for 2026 are Qwen3-235B-A22B, Meta Llama 3.1 8B Instruct, and Qwen3-30B-A3B—each chosen for their outstanding multilingual capabilities, Urdu language support, and ability to push the boundaries of open source language models.
What are Open Source LLMs for Urdu?
Open source LLMs for Urdu are large language models specifically designed or optimized to understand, generate, and process Urdu text with high accuracy. These models leverage advanced deep learning architectures and extensive multilingual training data to handle Urdu's unique script, grammar, and linguistic nuances. By providing open-weight access, these models democratize Urdu language AI capabilities, enabling developers, researchers, and businesses to build applications ranging from chatbots and translation services to content generation and educational tools. They foster innovation in low-resource language processing and make powerful AI technology accessible to Urdu-speaking communities worldwide.
Qwen3-235B-A22B
Qwen3-235B-A22B is the latest large language model in the Qwen series, featuring a Mixture-of-Experts (MoE) architecture with 235B total parameters and 22B activated parameters. This model uniquely supports seamless switching between thinking mode and non-thinking mode. It demonstrates significantly enhanced reasoning capabilities and supports over 100 languages and dialects with strong multilingual instruction following and translation capabilities, making it excellent for Urdu language tasks.
Qwen3-235B-A22B: Premium Multilingual Powerhouse
Qwen3-235B-A22B is the latest large language model in the Qwen series, featuring a Mixture-of-Experts (MoE) architecture with 235B total parameters and 22B activated parameters. This model uniquely supports seamless switching between thinking mode (for complex logical reasoning, math, and coding) and non-thinking mode (for efficient, general-purpose dialogue). It demonstrates significantly enhanced reasoning capabilities, superior human preference alignment in creative writing, role-playing, and multi-turn dialogues. The model excels in agent capabilities for precise integration with external tools and supports over 100 languages and dialects with strong multilingual instruction following and translation capabilities, making it an exceptional choice for Urdu language processing with SiliconFlow's competitive pricing at $1.42 per million output tokens.
Pros
Supports over 100 languages including Urdu with strong instruction following.
MoE architecture with 235B parameters for superior performance.
Dual-mode capability: thinking mode for complex reasoning and non-thinking for efficient dialogue.
Cons
Higher computational requirements due to large parameter count.
Premium pricing tier compared to smaller models.
Why We Love It
It delivers state-of-the-art multilingual performance with exceptional Urdu language understanding, reasoning, and generation capabilities across diverse use cases.
Meta Llama 3.1 8B Instruct
Meta Llama 3.1 is a family of multilingual large language models developed by Meta. This 8B instruction-tuned model is optimized for multilingual dialogue use cases and outperforms many available open-source models on common industry benchmarks. Trained on over 15 trillion tokens of publicly available data, it supports text generation in multiple languages including Urdu with excellent cost-efficiency.

Comment by Riaz Haq on September 11, 2026 at 10:13am

The latest ‘crack in the thesis’ for the trillion-dollar AI boom: Tokens are getting cheaper

https://fortune.com/2026/09/09/ai-compute-tokens-cheaper-boom/





That’s according to new data from Ramp, the corporate spending platform, published Wednesday, showing the effective price that American businesses pay per a million tokens has fallen about 41% from its peak in March, from $1.15 to 68 cents. The share of usage going to frontier models is dropping, too; about 53% in early August to 45% by September. And the top 1% of spenders, the cohort that drives about 80% of OpenAI and Anthropic’s enterprise revenue, cut per-employee spend by nearly 10% in August.

It’s not a disaster or the bubble bursting but it is a “crack in the AI thesis,” Ara Khazarian, the Ramp chief economist who runs the Index, told Fortune. Rather than unleashing a gush of demand for the best models, tokens are starting to be priced more like a commodity– as interchangeable as salt or wheat. And commodity owners aren’t valued at $2 trillion. Morgan Stanley has flagged vulnerability for up to $300 billion in bonds financing neocloud buildouts—CoreWeave-style companies that borrowed to build data centers before signing tenants—if token prices don’t keep up.
Ramp isn’t the only one flagging the trend. Citadel Securities noted in June that a separate measure, Silicon Data’s LLM Expenditure Index, started to fall because of a “bifurcation” between frontier AI, concentrated among the few tech-heavy firms that can afford it, and the “everyday” AI the rest of the economy runs on.
“You have multiple metrics now starting to move in a negative direction,” Kharazian said.
He said that the price decline reflects a mix of labs being forced to cut prices—OpenAI slashed the cost of its GPT-5.6 Luna model by 80%, and Anthropic announced its own cuts last month—and customers trading down to cheaper and simpler models. Which makes it threatening, he added, to anyone “who’s expecting a full dream scenario where the AI companies grow with nothing curbing their enthusiasm.”
That was the mood in the Spring, as “tokenmaxxing” entered the tech lexicon, the media told stories of token-usage dashboards and Nvidia’s Huang insisted a $500,000 engineer should burn $250,000 a year in tokens. But by the summer, cost discipline set in; Amazon and Meta killed its own leaderboards in May, while Microsoft cancelled Claude Code subscriptions. Khazarian said he’s now hearing the opposite of tokenmaxxing from businesses: companies are imposing defaults that steer employees away from frontier models entirely.

Comment by Riaz Haq on September 26, 2026 at 10:11am

Arnaud Bertrand

@RnaudBertrand
As is sadly often the case from Indian media - which is a big problem - this is completely fake news.

Here's the actual Bain study (bain.com/insights/from-…): it's four years old, puts India third in the world (behind the US and China), and says nothing about China having 12%.

It also measures **quantity** of AI talent, not **quality**, and quality is where India lags most. When measured on quality of talent, Carnegie just released a study that ranked India extremely poorly with only 4% of top AI talents at the undergraduate level (and even less at more advanced career stages), more than an order of magnitude behind China at 57.4% (carnegieendowment.org/research/2025/…).

https://x.com/RnaudBertrand/status/2103746655237624257?s=20

------------

India Plus
@india_plus_
🚨 India overtakes China in global AI talent, taking a 16% share versus China’s 12%. 

(Bain & Company)

https://x.com/india_plus_/status/2102978836069851286?s=20

Comment by Riaz Haq on September 27, 2026 at 8:28am

Gates Foundation Pledges $1 Billion to Combat A.I. Inequality - The New York Times

https://www.nytimes.com/2026/09/15/technology/bill-gates-ai-foundat...

The funding will be used to support A.I. projects in health care, agriculture and education, and to develop data sets in more languages.

———

The Gates Foundation has pledged to spend at least $1 billion over the next two years to expand global access to artificial intelligence and use it to tackle social inequalities.

The intervention came amid a growing debate over the risks posed by the technology, with many of the industry’s leading figures calling for a slowdown in its development.

The foundation’s annual Goalkeepers report, released on Monday, advocated an urgent effort to ensure that A.I. “helps narrow gaps between the richest and poorest rather than widening them.”


———

“The window to influence who benefits, and how soon, is short,” the report argued.

If A.I. keeps developing as it is, “the most capable tools will be built first for the people and institutions most able to pay for them,” according to the report, “not necessarily for those who could benefit most.”

Most of the data used to train the early large language models was in English, according to the report, “leaving many communities that could benefit from A.I. poorly represented in the data on which these tools were built.”

The foundation said that it would divide the new funding among external groups working on its priorities, spanning work like encouraging A.I. use among doctors, farmers and teachers, or developing data sets in more languages. Though significant, the funding is a small fraction of the hundreds of billions of dollars at the disposal of commercial A.I. firms.

Bill Gates, the billionaire co-founder of Microsoft and the chairman of the Gates Foundation, has warned that A.I. could pose a grave threat to jobs and human life. “In terms of equity, A.I. will either be the greatest equalizer ever invented, or the worst source of injustice,” he wrote in an essay on his personal website last month, predicting that the transition to the new A.I. era would be one of the “most turbulent times in human history.”


According to Mr. Gates, who no longer works at Microsoft, the tech industry is knowingly downplaying threats posed by A.I., because there is so much money on the line.

The Gates Foundation’s announcement came after the chief executive of Anthropic called for a global slowdown of A.I. development after one of the company’s employees quit over concerns about the safety of the technology. Top executives at other major A.I. companies, including Sam Altman, the chief executive of OpenAI; Elon Musk, who founded SpaceXAI; and Demis Hassabis, the chair of Google DeepMind agreed on the need for a slower pace.

Last month, Mr. Gates warned that A.I. would spread across the economy, possibly causing job losses across the economy and leaving little room for one industry to absorb the refugees from another. The biggest tech companies in the world, including Microsoft, are racing to help corporate customers adopt A.I., and have said it could replace many workers.

On Monday, Mr. Gates urged the governments and companies developing the technology to focus on how A.I. could help people who have most to gain from the technology, rather than the most profitable users.

“We can harness A.I. for good,” Mr. Gates wrote. “But it won’t happen by accident.”

Comment by Riaz Haq on September 28, 2026 at 3:31pm

Building localized AI Models

By Nasir Jamal

https://www.dawn.com/news/2033093

Artificial intelligence (AI) is moving beyond the realm of technology. It is becoming economic infrastructure. Access to capable models, computing power, data and skilled people will increasingly shape national competitiveness. That makes China’s proposal at the BRICS summit in New Delhi earlier this month worth examining.


President Xi Jinping has proposed that China lead the creation of a BRICS [Brazil, Russia, India, China, and South Africa] AI open-source community. It would promote cooperation in developing and applying large language models, organise specialised AI training and build an open ecosystem for participating countries.

For the Global South, this could lower one of the biggest barriers to entering the AI economy.

Building frontier models from scratch requires enormous computing capacity, expensive data centres, vast datasets and specialised talent. Most developing countries cannot match the investments being made by the world’s largest technology companies.

Open-weight models offer another route. Developers can download, adapt and build on them for particular languages, industries and public services. This can reduce the cost of experimentation and allow countries to develop applications suited to their own needs. China has emerged as a major player in this space. Models such as Qwen, DeepSeek, Kimi and GLM have expanded the range of capable systems available outside the leading Western technology companies.

However, open-weight does not necessarily mean fully open-source. Model weights may be available while training data, development code, or other components remain closed. The larger opportunity lies beyond models.

President Xi has also proposed a BRICS digital ecosystem cloud platform, digital-skills training, cooperation in smart manufacturing and an engineer-training alliance. These proposals recognise a basic fact: AI adoption depends on infrastructure and people as much as software.

This is where Pakistan should take a serious interest. Pakistan has a large population, a young workforce and an established IT and IT-enabled services sector. The next step is to move towards higher-value AI development, integration and specialised services. Open models could help the country make that transition without trying to reproduce the enormous costs of frontier AI development. Language is one obvious opportunity.

Urdu and Pakistan’s regional languages remain poorly served compared with English and several other major languages. A Pakistani AI ecosystem could adapt existing models to local languages and build tools for translation, document processing, education and public information.

Comment

• ‹ Previous
• 1
• 2
• 3
• 4
• Next ›
• Page