The boom in interest in mainstream models and generative AI is driving a surge in demand for additional computing power, as well as interest in training ever larger models.

According to documents available to Insider, Microsoft is preparing to address this by doubling Azure capacity in the second half of this fiscal year, which runs through the end of June.

Microsoft will triple additional data center capacity in the first half of its fiscal year 2025.

According to the aforementioned documents, the company is deploying a “record number of GPUs” to handle AI workloads, and more than doubled its total installed base of accelerators in the second half of last year.

Recall that Microsoft had intended last year to spend “many billions of dollars” building additional capacity in anticipation of the rapid increase in the number of customers looking to run generative AI projects.

By the end of 2023, Synergy’s data shows that the US accounts for 51% of global capacity as measured by MW of critical IT workload, Europe 17%, and China close behind with 16%.

The companies with the largest footprints are actually the big three cloud providers – Amazon, Microsoft and Google. They own 60% of all hyperscale data center capacity, followed by Meta, Alibaba, Tencent, Apple, ByteDance and other relatively small hyperscale operators.

Read more:
1. From Competitions in Informatics to MIT – Rumen Hristov’s Formula for Success
2. GitLab Releases GitLab Duo Chat with 40+ New Features
3. OpenSSF, CISA and DHS Join Forces in New Open Source Project

The Silicon Valley giant has released the first two models of the Llama 3 family – one with 8B parameters and one with 70B. According to Meta, these models are significantly better than the Llama 2 models, offering a much lower false rejection rate, improved alignment, and greater diversity in model responses. The specific capabilities of the models such as, code generation and instruction following have also been greatly improved.

Llama 3 was pre-trained on more than 15T tokens from publicly available sources, therefore the training dataset of Llama 3 is seven times larger than the training dataset of Llama 2 and contains four times more code.

According to Meta, in developing Llama 3, a new human evaluation set for benchmarking was also developed that contains 1,800 prompts across 12 use cases. These include asking for advice, classification, answering a closed question, writing code, creative writing, extraction, answering an open question, reasoning, rewriting, summarizing, and more.

The 70B parameter model beat Claude Sonnet, Mistral Medium, GPT 3.5, and Llama 2 using this new assessment set.

Llama 3 is available on AWS, Databricks, Google Cloud, Hugging Face, Kaggle, IBM WatsonX, Microsoft Azure, NVIDIA NIM and Snowflake. In addition, some hardware vendors will also offer support for it, including AMD, AWS, Dell, Intel, NVIDIA and Qualcomm.

Over the next few months, the company plans to update Llama 3 with new features, longer context windows, and larger model sizes.

Read more:
1. GitLab Releases GitLab Duo Chat with 40+ New Features
2. From Competitions in Informatics to MIT – Rumen Hristov’s Formula for Success
3. OpenSSF, CISA and DHS Join Forces in New Open Source Project

The project is led by the Linux Foundation’s LF AI and Data organization, which focuses on platform initiatives related to artificial intelligence and data.

LF AI and Data executive director Ibrahim Hadadt states in a press release that OPEA is expected to pave the way for the release of “fortified,” “scalable” generative AI systems that “leverage the best open source innovations from across the ecosystem.”

Other project members include Intel, IBM-owned Red Hat, Hugging Face, Domino Data Lab, MariaDB and VMware.

When asked what these companies can build together, Haddad hinted at several possibilities, such as “optimized” support for AI toolchains and compilers that allow AI workloads to run on different hardware components, as well as “heterogeneous” pipelines for generating augmented data (RAG). RAG is gaining popularity in enterprise applications of generative AI.

RAG models reference this external information – which can be in the form of proprietary company data, a public database, or some combination of the two – before generating a response or performing a task.

OPEA members say they are very interested in creating a toolkit for enterprise generative AI.

Although each of the companies has a different focus – Cloudera focuses on the cloud AI ecosystem, Domino offers generative AI applications for businesses, and VMware handles the infrastructure for AI – all have shown interest in AI development and have launched products and partnerships in this area.

However, the question of truly working together on developing compatible AI tools within OPEA remains open. Although companies may have different interests and strategies, collaboration is possible, especially if they have a common point that rests on the development of AI technologies.

Read more:
1. Acer Noted Consecutive Revenue Growth
2. Tesla Faces Stock Drop, Layoffs and Investor Pessimism
3. Microsoft Invests $1.5 Billion in UAE AI Company

“It’s core to our mission to make tools like ChatGPT broadly available so that people can experience the benefits of AI. More than 100 million people across 185 countries use ChatGPT weekly to learn something new, find creative inspiration, and get answers to their questions. Starting today, you can use ChatGPT instantly, without needing to sign-up,” OpenAI wrote in a blog post.

According to the company, this capability will be introduced gradually, so some users may still need to create an account to use the service until the rollout is complete.

However, some drawbacks of not having an account include the inability to save chat history, share chats, or use certain features like voice and personalized instructions.

In addition to this change, OpenAI introduces new content protections.

Samsung Mobile Executive Vice President Won Choi announced how the company will focus on improving the voice assistant’s features.

The executive vice president did not reveal details on how Samsung will integrate genAI to the voice assistant. He emphasizes that by adding AI, conversations will be more natural, and the company will work diligently on realizing the mission to build AI into the voice assistant.

There is information that the company is focusing on building features into Bixby that are similar to ChatGPT. According to Samsung, this will improve the voice assistant’s ability to respond naturally.

The voice assistant will also be able to summarize articles, documents as well as make suggestions for different places.

The president of the mobile division at the company says that Samsung will allow AI features on its devices to be free by 2025.

Researchers in China are helping 17 million blind people by creating guide dogs controlled by artificial intelligence, GIZMOCHINA shares. The project, developed by scientists at the Polytechnic University of China, provides functions such as accompanying and navigation.

The guide dog, controlled by artificial intelligence, is trained to understand the commands given to it. The attendant can also conduct conversations. It facilitates the daily life of visually impaired people by guiding them through the streets, helping them get into elevators or enter different buildings.

“Smart guide dogs using the language model can offer them more convenient and safe navigation, effectively improving their quality of life,” said Sun Zhe, an associate professor at the university.

Compared to the regular guide dogs we are used to, training AI guides is much cheaper. In addition, AI guide dogs can be created faster than regular dogs.

The project to create AI companions is still in the early stages of development, but according to China, it is possible that AI guide dogs will become widely available very soon.

The collaboration between the Polytechnic University of China and China Telecom Artificial Intelligence Research Institute accelerates the implementation of the project aimed at supporting the daily life of visually impaired people.

The first update is the ability to extract information from large amounts of data. For example, it can analyze unstructured data, such as customer feedback, and summarize the most important points.

“If you aggregate thousands of transcripts of customer calls, you can ask Airtable AI to categorize each call according to sentiment (positive, negative, or neutral) and group the data accordingly”, Airtable wrote in a blog post.

The ability to automatically apply categories or tags to data is among the other new features. This helps teams keep their data better organized. Categories can be based on topics, product features, etc.

Users can also create content based on their data, such as generating emails for sales teams or draft campaign assets for marketing teams. This is done using message, product and audience information that has been stored in Airtable.

Other notable features include the ability to automatically translate content into multiple languages.

“Organizations have enormous potential to transform the way they work with AI,” said Howie Liu, co-founder and CEO of Airtable.

He further adds that with the new features, Airtable AI offers a fast way to introduce AI into data and workflows and “unlocks the possibilities of AI for small and large customers.

New features in the platform also include capabilities to improve data routing so it gets into the hands of the right team. This will save managers time spent in determining which teams data requests should go to.

Amazon’s total investment in Anthropic is worth $4 billion.

According to sources, Amazon had until the end of the first quarter to decide whether to increase its investment to a maximum of $4 billion, and a few months later – the investment is already in place.

Anthropic’s AI models are one of very few that compete at the highest levels of capability, but are available at enterprise scale for internal deployment or in consumer-facing applications.

OpenAI’s GPT AI chatbots and Google’s Gemini are well known among tech circles, but newcomers like Mistral could very well threaten the AI dominants.

They identify 2021-founded Anthropic as one of OpenAI’s biggest competitors, attracting billions in funding.

Anthropic has relationships with big tech companies like Google, which joined a $450 million funding round in May led by Spark Capital. Google and Amazon Web Services are partners in Anthropic’s cloud computing.

The company will use the new funding to continue to develop its technology and add new features to the platform.

“It has become a remarkably difficult task for developers to confirm crucial questions like, ‘What dashboards will this schema change affect?,’ ‘Can I deploy this code change safely?’ or ‘What data pipelines do we have that are not working?’” said Alon Nafta, CEO and co-founder of Foundational.

“The inability to validate simple code changes, and the lack of visibility and controls, and the time-consuming nature of data engineering create a dramatic bottleneck on innovation and adoption of data and AI”, he added.

By building partnerships with data organizations around the world, Foundational aims to redefine the way large-scale data platforms are built and maintained. With this funding, the company will continue to evolve its technology and introduce new product capabilities for data developers.

This marks the biggest change to the computer keyboard in decades. Although the tech giant isn’t the biggest seller of PCs – that distinction belongs to Lenovo – it runs the most popular Windows operating system. Lenovo has announced its own computers with Copilot keys, as have Dell and HP.

When users type a few words into Copilot, which uses artificial intelligence models from OpenAI, servers in remote data centers do the necessary computational work to produce a response. Microsoft named the new machines as AI PCs.

The key that allows users to open the Copilot panel at any time on the right side of the screen helps with this. Each computer has an Intel processor
Core Ultra, which contains a dedicated neural processing unit, or NPU.

Surface computers have had NPUs since 2019.

When these NPUs are in a computer, the benefits include faster responses and better security. Recently added artificial intelligence features in Windows 11, such as automatic audio transcription and eye contact simulation during video calls, are run by the NPU, which frees up the rest of the chip for other tasks.

“We’re excited to bring to market devices that deliver great AI experiences for our customers,” Microsoft said.

The company reports that this is just the first part of its quest to provide consumers and organizations with improved devices that will not only boost their productivity, but also save some of their time on certain tasks.

The button’s introduction comes two weeks after Apple unveiled the MacBook Air laptops, which feature an upgraded Neural Engine accelerator in the custom M3 chip.