Connect with us


Apple Launches Eight Small AI Language Models for On-Device Use



Within the field of artificial intelligence, “small language models” have gained significant traction lately due to their ability to operate locally on a device rather than requiring cloud-based data center-grade computers. On Wednesday, Apple unveiled OpenELM, a collection of minuscule AI language models that are available as open source and small enough to run on a smartphone. For now, they’re primarily proof-of-concept research models, but they might serve as the foundation for Apple’s on-device AI products in the future.

Apple’s new AI models, collectively named OpenELM for “Open-source Efficient Language Models,” are currently available on the Hugging Face under an Apple Sample Code License. Since there are some restrictions in the license, it may not fit the commonly accepted definition of “open source,” but the source code for OpenELM is available.

A similar goal is pursued by Microsoft’s Phi-3 models, which we discussed on Tuesday. These models are small, locally executable AI models that can comprehend and process language to a reasonable degree. Although Apple’s OpenELM models range in size from 270 million to 3 billion parameters across eight different models, Phi-3-mini has 3.8 billion parameters.

By contrast, OpenAI’s GPT-3 from 2020 shipped with 175 billion parameters, and Meta’s largest model to date, the Llama 3 family, has 70 billion parameters (a 400 billion version is on the way). Although parameter count is a useful indicator of the complexity and capability of AI models, recent work has concentrated on making smaller AI language models just as capable as larger ones were a few years ago.

Eight OpenELM models are available in two flavors: four that are “pretrained,” or essentially a next-token version of the model in its raw form, and four that are “instructional-tuned,” or optimized for instruction following, which is more suitable for creating chatbots and AI assistants:

The maximum context window in OpenELM is 2048 tokens. The models were trained using datasets that are publicly available, including RefinedWeb, a subset of RedPajama, a version of PILE that has had duplications removed, and a subset of Dolma v1.6, which contains, according to Apple, roughly 1.8 trillion tokens of data. AI language models process data using tokens, which are broken representations of the data.

According to Apple, part of its OpenELM approach is a “layer-wise scaling strategy” that distributes parameters among layers more effectively, supposedly saving computational resources and enhancing the model’s performance even with fewer tokens used for training. This approach has allowed OpenELM to achieve 2.36 percent accuracy gain over Allen AI’s OLMo 1B (another small language model) with half as many pre-training tokens needed, according to Apple’s published white paper.

In addition, Apple made the code for CoreNet, the library it used to train OpenELM, publicly available. Notably, this code includes reproducible training recipes that make it possible to duplicate the weights, or neural network files—something that has not been seen in a major tech company before. Transparency, according to Apple, is a major objective for the organization: “The reproducibility and transparency of large language models are crucial for advancing open research, ensuring the trustworthiness of results, and enabling investigations into data and model biases, as well as potential risks.”

By releasing the source code, model weights, and training materials, Apple says it aims to “empower and enrich the open research community.” However, it also cautions that since the models were trained on publicly sourced datasets, “there exists the possibility of these models producing outputs that are biased, or objectionable in response to user prompts.”

Though the company may hire Google or OpenAI to handle more complex, off-device AI processing to give Siri a much-needed boost, Apple has not yet integrated this new wave of AI language model capabilities into its consumer devices. It is anticipated that the upcoming iOS 18 update—which is expected to be revealed in June at WWDC—will include new AI features that use on-device processing to ensure user privacy.


Coforge and Microsoft Establish Copilot Innovation Hub to hasten the Deployment of Generative AI



Global supplier of digital services and solutions Coforge Limited recently announced a partnership with Microsoft to launch the Coforge Copilot Innovation Hub. In addition to working closely with Microsoft to integrate these solutions with Microsoft’s generative AI products and technologies, such as Microsoft Azure OpenAI Service, Microsoft Power Platform, and Microsoft Copilot, the Hub will concentrate on building a pipeline of new, industry-specific cognitive analytics solutions.

Coforge announced the launch of two new copilots as part of the Copilot Innovation Hub: Underwriter Copilot for insurance companies and Advisor Copilot for financial services firms. An innovative technique to improve ROI and streamline the process of navigating the complexity of underwriting, the Underwriter Copilot for Insurance gives insurance underwriters more authority and the ability to make informed decisions. The goal of the solution is to increase carriers’ combined ratios by two to three percent in order to open up new income streams. Insurance carriers can achieve a 30- to 35 percent boost in underwriter productivity and efficiency by implementing Underwriter Copilot.

By removing the need for time-consuming searches across several tools, documents, and data sources, the Coforge Advisor Copilot solution gives financial quick access to full fund information and performance data through an intuitive interface. Financial advisers and asset managers should become more productive by more than thirty percent thanks to the solution.

According to Sudhir Singh, Executive Director & CEO of Coforge, “Coforge is leveraging its deep industry strengths and customer partnerships to build industry specific generative AI solutions on the Microsoft platform to drive transformation and enhance productivity.” Our efforts to provide our clients with generative AI solutions that lead the market will go even faster thanks to our partnership with Microsoft. He went on, “We are announcing two new copilots today: Advisor Copilot for financial services businesses and Underwriter Copilot for insurance carriers.

“Our combined commitment to transforming and scaling organizational capabilities of financial services firms globally is demonstrated by the Coforge Copilot Innovation Hub. The 2024 Work Trend Index Annual Report states that 75% of individuals utilize AI at work, and that the use of generative AI has nearly doubled in the last six months. According to David Smith, Vice President, WW Channel Sales, Microsoft, “Coforge and Microsoft are dedicated to spearheading AI adoption, fostering innovation, and unleashing business value for businesses worldwide.”

Through the automation of manual chores, the improvement of decision-making through the creation of suggestions based on corporate data, and the streamlining and optimization of business processes, these copilots will increase operational efficiency by utilizing Microsoft’s generative AI products and technologies. These solutions will help businesses generate new value streams and speed up change.

Microsoft’s generative AI products will be easier to implement with the Coforge Copilot Innovation Hub, leading to increased productivity and better business results.

Continue Reading


Samsung Appoints New Leader for Chip Unit as AI Competition Intensifies



As the race to build artificial intelligence processors heats up, Samsung Electronics has replaced the leader of its semiconductor division.

In an unexpected announcement made by the business on Tuesday, Vice Chairman Jun Young-hyun has been named head of Samsung’s device solutions division. The company’s foundry, memory, and system semiconductor divisions are managed under the device solutions division.

“Vice Chairman Jun Young-hyun is the key player who took Samsung Electronics’ memory semiconductor and battery businesses to the global top-tier level,” the company stated in a news release.

Samsung is making this announcement as it battles to overtake its regional rival SK Hynix in the market for AI memory chips. When it comes to high-bandwidth memory (HBM) chips, which are essential for AI computing, SK Hynix is in the lead.

According to Samsung, if the board and shareholders approve, Jun may also be named as the company’s chief executive. Samsung has two chief executive officers: one leads the company’s semiconductor division, while the other oversees its mobile and visual display businesses.

Before taking on the role of chief executive of Samsung SDI, the company’s battery division, Jun led Samsung’s memory chip business team for three years, from 2014 to 2017. In 2000, he made his debut as a member of Samsung’s memory chip business team.

Kyung Kye-hyun, who oversaw the semiconductor branch since2022, is replaced by Jun. During the memory chip market collapse, the division under his direction reported billion-dollar losses. The 61-year-old Kyung has been posting lengthy and in-depth posts on social media platforms like LinkedIn and Instagram about subjects including technology and climate change.

Continue Reading


Kudos Secures $10.2 Million for Its AI-Powered Smart Wallet



The Four Cities Fund, Samsung Next, SV Angel, Precursor Ventures, The Mini Fund, Newtype Ventures, Patron, and The Points Guy creator Brian Kelly all participated in the funding round.

Kudos, an app and browser extension, was founded in 2001 by a group with prior expertise at Google, PayPal, and Affirm. It functions as a smart wallet assistant by suggesting or choosing the best credit card for customers to use when making payments in order to optimize rewards and cash back.

Recently, the company introduced a number of new features: Dream Wallet, which suggests cards to members based on their spending patterns; MariaGPT, an AI-powered card discovery tool with over 3000 cards in its database; and Kudos Boost, which offers personalized rewards across over 15,000 partner brands, such as Walmart and Sephora.

Since its initial fundraising round, Kudos has raised its annualized checkout Gross Merchandise Value to $200 million and expanded to over 200,000 registered users.

It intends to use the additional funds to develop MariaGPT into a comprehensive personal finance assistant, introduce an AI-powered hub offering expenditure optimization insights, and create a gateway that lets users book flights using points.

As consumers budgets, various credit cards, and sometimes complex rewards programs, they want to know they’re receiving the best value for their money, according to Tikue Anazodo, CEO of Kudos. With just one user-friendly app and extension, Kudos streamlines everything.”

Continue Reading


error: Content is protected !!