Nvidia plans new AI chips every year after logging  billion in profits

Nvidia plans new AI chips every year after logging $15 billion in profits


Representative Image: Nvidia H100 chip

American chip maker Nvidia would likely make new artificial intelligence chips every year instead of once every two years. According to a report by The Verge, during the company’s Q1 earnings call, Nvidia CEO Jensen Huang said that there is another chip which is currently under the works and the company is on a “one-year rhythm”. 


Nvidia logged profits worth $14 billion in a single quarter, thanks to AI chips.


Additionally, Huang also said that the company is working on new central processing units (CPUs) and graphic processing units (GPUs). “New CPUs, new GPUs, new networking NICs, new switches… a mountain of chips are coming,” he said. 


Nvidia has been releasing a new AI chip platform once every two years with Ampere platform in 2020, Hooper in 2022 and Blackwell in 2024. Nvidia’s H-series chips that are used for training AI models and other AI applications are based on the company’s Hooper architecture. These AI chips are currently used by multiple big technology companies such as Microsoft, Google, Meta, Amazon and more to train large language models (LLMs) and to deploy AI-powered services on their platforms. 


In march, Nvidia revealed its new B-series chip that is based on the Blackwell platform, however, the chips are not currently available for deployment as Nvida said that “Blackwell-based products will be available from partners starting later this year.”


Earlier this month, analyst Ming-Chi Kuo in a post on social media platform Medium, reported that Nvidia’s next generation chip for AI-application will be called “R100”. He said that the R100 chip will be based on TSMC’s N3 process for 3nm architecture. For comparison, the current generation Blackwell chips are based on TSMC’s N4P process for 5nm design. Kuo said that these R100 AI chips will enter mass production in the fourth quarter of 2025.

First Published: May 23 2024 | 11:38 AM IST



Source link

Microsoft's UAE deal could transfer key US chips, AI technology abroad

Microsoft's UAE deal could transfer key US chips, AI technology abroad


Microsoft President Brad Smith said the tech company’s high profile deal with the United Arab Emirates-backed AI firm G42 could eventually involve the transfer of sophisticated chips and tools. Photographer: Krisztian Bocsi/Bloomberg


Microsoft President Brad Smith said the tech company’s high profile deal with the United Arab Emirates-backed AI firm G42 could eventually involve the transfer of sophisticated chips and tools – a move that a senior Republican congressman warned could have national security implications.

 


In an interview with Reuters this week, Smith said the sales accord, many details of which are being reported here for the first time, could progress to a second phase that entails the export of crucial components of AI technology such as model weights, a crown jewel of AI systems that determine how powerful they are. Smith said there is no firm timeline for the second phase.

 


US officials have said that AI systems could pose national security risks, for example by making it easier to engineer


chemical, biological and nuclear weapons . The Biden administration in October required the makers of the largest AI systems


to share details about them with the US government.

 


To move forward, the deal would require the approval of the US Department of Commerce. Microsoft executives said the agreement has safeguards to protect Microsoft’s technology and prevent it from being used by Chinese entities to train AI systems.

 


But those measures have not been made public, and some USlawmakers question whether they are adequate.


The closed-door nature of the negotiations between two private companies over the terms and safeguards on transfers of US technology have alarmed some lawmakers.

 


“Despite the significant national security implications, Congress still has not received a comprehensive briefing from the executive branch about this agreement,” Michael McCaul, the Republican chairman of the foreign affairs committee in the US

 


House of Representatives, told Reuters. “I am concerned the right guardrails are not in place to protect sensitive US-origin technology from Chinese espionage given the (Chinese Communist Party’s) interests in the UAE.” The Commerce Department already requires notifications and, in several regions, export licenses to send AI chips abroad. But the Microsoft-G42 deal highlights gaps in US laws as regulators rush to keep up with fast-moving technology.

 


At present, for example, there is no regulation restricting the export of AI models, though McCaul and a bipartisan group of lawmakers this week advanced legislation that would give US officials more explicit power to do so.

 


Microsoft executives said the company welcomes a debate on a new legal framework governing the transfer of AI technology and that the deal with G42 requires the UAE firm to comply with US regulations as they evolve.

 


“Fundamentally, what we’re focused on is trying to ensure that American technology can move around the world safely and securely,” Smith said.

 


Beyond UAE

 


When Microsoft and G42 announced the deal last month, it was billed as drawing G42 closer to the US and spreading US


technology influence amid strategic competition with China.

 


Microsoft is investing $1.5 billion in G42 with Microsoft’s president, Smith, taking a seat on its board.

 


The companies did not give details about which technologies might be transferred to the UAE or other countries or which specific security safeguards would be put in place.

 


Some of those details are being reported here for the first time.

 


The broad intent of the deal is for Microsoft and G42 to jointly take AI technology into regions where neither could do so as effectively alone. An early example is a deal in Kenya announced by the two companies on Wednesday.

 


The Microsoft-G42 deal is an agreement between the two companies that requires each to give security assurances to their respective home governments, but there is no direct agreement between the US and UAE governing the transfer of sensitive technologies. The two companies could seek to transfer those technologies to other markets beyond the UAE, including places like Turkey and Egypt, Microsoft executives said.

 


Smith said many of the details of the deal remain to be worked out, including how to protect what are known as AI “model weights,” which is the critical part of an AI model that defines how it responds to questions or prompts. Those weights are obtained by training an AI model with huge amounts of data, often at great expense.

 


Model weights currently cannot be encrypted while in use, and Smith estimated the promising technical approaches for doing so remain at least a year away.

 


Smith said Microsoft has considered several alternative options to protect its technology, including a “vault within a vault” that would involve physically separating parts of data centers where AI chips and model weights are housed and restricting physical access.

 


“I suspect by the time we’re done, we’re going to end up with a regulatory regime or trade export control approach that will be applicable broadly and not just to Microsoft and G42,” Smith said.

 


Under the Microsoft deal, G42 will also follow a “know your customer” rule to determine who is using Microsoft’s technology and will not allow Chinese firms to use it to train AI models, Microsoft executives said. US regulators have proposed a similar rule, but they have not yet enacted it.

 


“We adopted a strategic commercial decision to partner with US companies when it comes to advanced technologies. And we’re very clear on the fact that in order to do so, we will need to adhere to the requirements and our partners and government regulatory requirements or export control regulations,” Talal Al Kaissi, an executive who handles partnerships for G42’s AI work, told Reuters.

 


Under the deal, Microsoft would have the ability to impose financial penalties on G42 and enforce them in arbitration courts in London, Microsoft said. That means Microsoft would not be forced to work through the UAE legal system to ensure G42 complies with its obligations and could seize assets in many countries if G42 is found in violation of the agreement, Microsoft said.

 


Precisely how US Commerce Secretary Gina Raimondo will allow the deal to move forward remains unclear. Smith said the provisions are “informal” and that “certainly with this Secretary of Commerce, one knows pretty clearly whether she approves or rejects something.”

 

In a statement, a Commerce Department spokesperson said any technology transfers would be governed by export controls, “including currently in force licensing requirements” for AI chips and “potential future controls.”


(Only the headline and picture of this report may have been reworked by the Business Standard staff; the rest of the content is auto-generated from a syndicated feed.)



Source link

Artificial intelligence black boxes just got a little less mysterious

Artificial intelligence black boxes just got a little less mysterious



One of the weirder, more unnerving things about today’s leading artificial intelligence (AI) systems is that nobody — not even the people who build them — really knows how the systems work.


That’s because large language models, the type of AI systems that power ChatGPT and other popular chatbots, are not programmed line by line by human engineers, as conventional computer programs are.


Instead, these systems essentially learn on their own, by ingesting massive amounts of data and identifying patterns and relationships in language, then using that knowledge to predict the next words in a sequence.


One consequence of building AI systems this way is that it’s difficult to reverse-engineer them or to fix problems by identifying specific bugs in the code. Right now, if a user types “Which American city has the best food?” and a chatbot responds with “Tokyo,” there’s no real way of understanding why the model made that error, or why the next person who asks may receive a different answer.


And when large language models do misbehave or go off the rails, nobody can really explain why. The inscrutability of large language models is a major reason some researchers fear that powerful AI systems could eventually become a threat to humanity.


After all, if we can’t understand what’s happening inside these models, how will we know if they can be used to create novel bioweapons, spread political propaganda or write malicious computer code for cyberattacks? If powerful AI systems start to disobey or deceive us, how can we stop them if we can’t understand what’s causing that behavior in the first place?


But this week, a team of researchers at the AI company Anthropic announced what they called a major breakthrough — one they hope will give us the ability to understand more about how AI language models actually work, and to possibly prevent them from becoming harmful. The team summarised its findings in a blog post called “Mapping the Mind of a Large Language Model.”


They looked inside one of Anthropic’s AI models — Claude 3 Sonnet, a version of the company’s Claude 3 language model — and used a technique known as “dictionary learning” to uncover patterns in how combinations of neurons, the mathematical units inside the AI model, were activated when Claude was prompted to talk about certain topics. They identified roughly 10 million of these patterns, which they call “features.” They found that one feature, for example, was active whenever Claude was asked to talk about San Francisco. Other features were active whenever topics like immunology or specific scientific terms, such as the chemical element lithium, were mentioned. And some features were linked to more abstract concepts, like deception or gender bias.


They also found that manually turning certain features on or off could change how the AI system behaved. For example, they discovered that if they forced a feature linked to the concept of sycophancy to activate more strongly, Claude would respond with flowery, over-the-top praise for the user, including in situations where flattery was inappropriate.


Chris Olah, who led the Anthropic interpretability research team, said these findings could allow AI firms to control their models more effectively.

“We’re discovering features that may shed light on concerns about bias, safety risks and autonomy,” he said. “I’m feeling really excited that we might be able to turn these controversial questions that people argue about into things we can actually have more productive discourse on.” Other researchers have found similar phenomena in these language models. But Anthropic’s team is among the first to apply these techniques . Jacob Andreas, an associate professor of computer science at MIT, who reviewed a summary of the research, characterised it as a hopeful sign that large-scale interpretability might be possible.


CRACKING THE CODE




> Black box Problem can be defined as an inability to fully understand an AI’s decision-making process


> Large language models, such as ChatGPT, are not programmed line by line by human engineers


> So, when they misbehave or go off the rails, nobody can really explain why


> Researchers looked inside one of Anthropic’s AI models — Claude 3 Sonnet


> They used a technique known as “dictionary learning” to uncover patterns in how combinations of neurons were activated when Claude was prompted to talk about certain topics


> They found that one feature, for example, was active whenever Claude was asked to talk about San Francisco


> They also found that manually turning certain features on or off could change how the AI system behaved


> Researchers believe these findings could allow AI firms to control their models more effectively

First Published: May 22 2024 | 11:18 PM IST



Source link

Tech wrap May 22: WhatsApp's new feature, Realme GT 6T launch, and more

Tech wrap May 22: WhatsApp's new feature, Realme GT 6T launch, and more



WhatsApp could soon let you clear all unread chats in one go. Reportedly, Meta’s instant-messaging platform is testing this feature in a beta version for Android. Part of the notification settings, the “clear unread when app opens” option will allow users to clear their unread message count whenever they open the app.


Realme on May 22 launched the Realme GT 6T smartphone in India at Rs 30,999 onwards. Alongside, the Chinese smartphone brand launched the Buds Air 6 wireless earbuds. Powered by Qualcomm Snapdragon 7+ Gen 3 system-on-chip, the smartphone boasts a dual vapour chamber (VC) cooling system that the company said helps the performance-centric phone maintain thermal balance. Moreover, the phone boasts a 5,500 mAh battery and support for 120W fast wired charging. The smartphone is offered in Fluid Silver and Razor Green colours.


Microsoft debuted a new category of personal computers, called Copilot Plus PCs, at an event on May 20. The Surface Pro and Surface Laptop are the first Windows PCs to launch under the Copilot Plus PCs platform, powered by Qualcomm’s Snapdragon X Elite and Snapdragon X Plus. Other original equipment manufacturers (OEMs) such as Acer, ASUS, Dell, HP, Lenovo, and Samsung have also announced Copilot Plus PCs powered by Qualcomm Snapdragon X Elite and Plus chipsets.


Adobe on May 22 announced the addition of “Generative Remove” feature in Lightroom. Powered by its Firefly AI model, the generative remove feature is integrated across mobile app, web client, and desktop application of Lightroom. With generative remove, Adobe said, users will be able to remove unwanted objects from any photo. Moreover, according to Adobe, the feature intelligently matches the removed area with “pixel perfect generations for high-quality, realistic and stunning results”.


A day after announcing Copilot+ PCs platform and artificial intelligence-powered Windows on Arm, Microsoft unveiled an array of new products and features at its Build 2024 event. While most of the announcements at the event were directed towards developers, there were a few that can bring new experiences to Windows for consumers. These include a new multimodal version of Microsoft’s small language model (SLM) Phi-3 and more AI integration into the Microsoft Edge browser.


Motorola Edge 50 Fusion smartphone is set to go on sale in India today, on May 22. Launched in the country on May 16, the Motorola Edge 50 Fusion will be available for purchase in the maiden sale period starting at 12 pm on May 22. The smartphone will be available on e-commerce platform Flipkart with introductory offers.


WorksOnWoA.com website from Linaro, a software development company that works on application deployment on Arm platform, has reportedly tested 1,481 games on the new Arm chip powered Surface Laptop and other devices with Snapdragon X Elite chips. The website allows users to search for a specific game and check how the game will perform on the new PC platform. The website shows if the game of preference falls in “Perfect”. “Playable”, “Runs” or “Unplayable” categories.


Google is reportedly planning to accelerate the connection process for Quick Share with QR codes. According to a report by Android Authority, Google’s alternative to Apple’s AirDrop could soon allow users to scan a generated QR code to establish a Quick Share connection for file sharing between Android devices.


Dixon Technologies, a local contract manufacturer, will produce Alphabet Inc’s high-end Google Pixel 8 smartphones in the country. The company said the initial batch of India-made phones will be available in the market by September, as trial production has recently started, according to a report in the Economic Times (ET).


The government is planning to chart its own course in the field of artificial intelligence (AI) by laying the groundwork for a bespoke foundational model tailored to the needs of Indian companies, entrepreneurs, academics, and researchers, reported The Economic Times citing sources. This ambitious initiative will start with an initial investment of Rs 2,000 crore, slated for launch after the ongoing 2024 Lok Sabha elections.


Chinese tech giants Alibaba and Baidu slashed prices on Tuesday of large-language models (LLMs) used to power generative artificial intelligence products, as a price war in the cloud computing sector heats up in China.


The PC just got an AI makeover, raising hopes that the buzzy technology would help revive an industry that has been on a steady decline over the last few years. Manufacturers say AI PCs process data more swiftly than traditional PCs and can handle a greater volume of AI tasks directly on the device, including chatbots. That means they do not have to rely on cloud data centers that currently power most AI applications, including OpenAI’s ChatGPT.

First Published: May 22 2024 | 8:03 PM IST



Source link

Amazon lining up AI overhaul, paid plans for Alexa voice assistant: Report

Amazon lining up AI overhaul, paid plans for Alexa voice assistant: Report


Amazon introduced Alexa in 2014 but has not found a consistent means to make it profitable, instead driving shoppers to the company’s website for more purchases. (Photo: Bloomberg)


Amazon.com is working on an artificial intelligence-driven overhaul of its Alexa voice assistant and plans to charge a monthly subscription fee to offset technology costs, CNBC reported on Wednesday.

 


The online retail giant will launch a more conversational version of the voice assistant later this year, positioning it to better compete with AI-powered chatbots from Microsoft and Alphabet’s Google, the report said, citing people with knowledge of Amazon’s plans.


Amazon did not immediately respond to a Reuters request for comment.

 


A subscription to Alexa will not be included in the company’s $139 Prime offering and Amazon is yet to nail down the price of the new services, the report said. It added that the company will use its own Titan large language model for the upgrade.

 


Amazon introduced Alexa in 2014 but has not found a consistent means to make it profitable, instead driving shoppers to the company’s website for more purchases.

 


It updated its Alexa voice assistant with generative artificial intelligence last September to attract users. The company cut “several hundred” jobs in its Alexa voice assistant unit in November.

 


Apple is also reportedly looking to revamp its Siri virtual assistant with generative artificial intelligence features.

 


The CNBC report comes more than a week after OpenAI unveiled a new AI model called GPT-4o, which enables users to speak to ChatGPT and obtain real-time responses without lag. It can also interrupt ChatGPT.

 


Both are hallmarks of realistic conversations that older AI assistants struggle to handle.

First Published: May 22 2024 | 8:01 PM IST



Source link

Premium devices to deliver value growth for India smartphone market in 2024

Premium devices to deliver value growth for India smartphone market in 2024


Factors driving this premiumisation trend include affordable financing schemes, better value for trade-ins, and bundled schemes


Indian smartphone sales, which have been dominated by the entry and mid-range segments, are set to change as the high-end and super premium ranges grab consumer attention. The premium category is expected to grow more than 20 per cent in 2024 compared to the previous year.


While the overall smartphone market is expected to witness modest single-digit growth in 2024, most of this growth is anticipated to come from premium devices and 5G upgrades.


“While we expect some growth in the Rs 10,000-20,000 and Rs 20,000-30,000 segments, a majority of the growth will come from the premium category. We expect more than 20 per cent growth in this segment in 2024,” said Shilpi Jain, senior research analyst, Counterpoint.


After flat growth in 2023, India’s smartphone shipments picked up in the first quarter (Q1) of 2024 (January-March). In Q1, while the market grew 8 per cent year-on-year (Y-o-Y) in terms of volume, it grew a substantial 18 per cent in terms of value, according to data from Counterpoint – a market research firm.


“During the quarter, India’s smartphone market reached its highest ever Q1 value. The growth was driven by the strengthening trend of premiumisation, with consumers upgrading to higher-value smartphones across price tiers,” said Jain.


New launches during the quarter, such as the Samsung Galaxy S24 and OnePlus 12 series, allowed the premium segment (>Rs 30,000) to reach a record 20 per cent volume share and 51 per cent value share of the overall smartphone market in Q1.

Jain added that this premiumisation was not limited to consumer demand.


“There is brand push as well. Brands are prioritising profits over higher volumes, and are aiming for higher ASPs,” she said.


Notably, the ultra-premium category (>$800 or roughly >Rs 65,000) recorded the highest growth at 44 per cent, with its share up from 7 per cent in Q1 2023 to 9 per cent a year later, as per data from the International Data Corporation (IDC).


Factors driving this premiumisation trend include affordable financing schemes, better value for trade-ins, and bundled schemes, along with the demand for top-tier features such as AI, gaming, and imaging enhancements.


Upasana Joshi, senior research manager, client devices, IDC India, said that several new launches across multiple price segments happened during the quarter, coupled with increased promotional activities, particularly around premium offerings.


“Brands continued their focus on micro-financing schemes to drive affordability. e-tailers organised several sales events and offered discounts on models nearing their end-of-life (EOL), resulting in increased demand and lean inventory,” Joshi said.


Apple and Samsung cornered most of the market share in the premium category during the quarter and are expected to maintain their lead going forward.


The iPhone 14/15/14 Plus/15 Plus together accounted for as much as 64 per cent of shipments in Q1, followed by the Samsung Galaxy S24/S24 Ultra/S23/S24+ with 25 per cent share. Overall, Apple led the segment with a share of 69 per cent, followed by Samsung at 31 per cent, as per IDC data.


Driven by the increased demand for high-end devices, Apple – with an ASP of $953 – had a record first-quarter shipment, growing by 19 per cent Y-o-Y.


Price discounts, special deals on e-tailer platforms, and attractive financing options led to this consistent growth for Apple beyond the festive season, with the iPhone 14/15 making up 56 per cent of shipments for the Cupertino-based firm, IDC analysts said.


Regardless, device prices across categories are expected to rise this year.


“Vendors are aggressively pricing their products to capitalise on the premiumisation trend fuelled by wider channel access and easy financing options. Price hikes are expected to continue as operational pressures rise due to higher component costs, despite import duty reductions on a few parts. Brands are also prioritising channel incentives and retail investments, driving costs up further,” said Sanyam Chaurasia, senior analyst, Canalys.

First Published: May 22 2024 | 5:30 PM IST



Source link

YouTube
Instagram
WhatsApp