Artificial Intelligence News
AI world news.
OpenAI has acquired a company, Rockset, building tools to drive real-time search and data analytics. In a post on its official blog, OpenAI said that it would integrate Rockset’s technology to “power [its] infrastructure across products.” Members of Rockset’s team will join OpenAI, and Rockset’s existing customers will be transitioned off of Rockset’s platform “gradually.” […]
© 2024 TechCrunch. All rights reserved. For personal use only.
Summer is officially here — and with the season changing, naturally, you are thinking about your next hairstyle. Will it be bangs? Dyed hair? A haircut? Before committing to an expensive and potentially regrettable change, you can use artificial intelligence (AI) to see exactly how you would look with the new hairstyle.
In the past, I have dyed my hair nearly every shade from the black to bleach-blonde spectrum. Now that I have the hair-change bug again, I consulted AI before making the same mistake I have made in the past of picking a style or color that does not suit me.
Also: The best AI image generators
I used FaceApp, an AI photo-editing app that can transform your face into different looks, including using makeup, accessories, hairstyles, and more. This application stands out because the results are realistic, unlike other apps I've tried.
Although the app, available for iPhone and Android, is free to download, it requires a FaceApp Pro subscription to access most of the functions, which costs $10 per month. Even though that cost may seem steep, $10 is less risky than a failed $300 hair-styling session you can't easily undo.
So, just how easy is FaceApp to use? Keep reading to find out.
First, you will need to install the application on your device. FaceApp can be downloaded for free on the Apple App Store and Google Play Store.
Then, you can upload a photo of yourself to start the AI magic. If you are using the app to test out different hairstyles, I recommend using a well-lit photo that frames your face and displays all your facial features. This means no hats, sunglasses, hair on your face, etc. You can use my original photo above as a reference.
Although many generative AI applications take user inputs and use them to train their models, FaceApp assures users in its privacy policy posted on its website that the company does not "use photographs or videos you provide when you use the Apps for any reason other than to provide you with the editing functionality of the Apps."
Also: Two ways you can build custom AI assistants with GPT-4o — and one is free!
However, the company has faced some controversy regarding its security and privacy practices, so investigate to see if you feel comfortable uploading your images into the app.
Once you select your photo, you will see a bar at the bottom of the app with all the edits you can make, including impressions (one-click makeovers), hairstyles, sizes, skin, makeup, smiles, hair colors, age, and more.
You will likely use the "hairstyles" and "hair color" options (or a combination of both) for hair makeovers. Because you have a free account, those options will be locked, and you must subscribe to the FaceApp Pro $10.00 subscription for access. Even though the exact process will depend on whether you use an Apple or Android phone, you can click on the locked style, opt into the FaceApp Pro popup, and follow the instructions to subscribe.
The app is intuitive. You can browse the different options, selecting new hairstyles, colors, haircuts, textures, volume, and more.
Also: How my 4 favorite AI tools help me get more done at work
When you click on some styles, numbered options pop up; those control the filter intensity. You will also see a transfer option allowing you to upload a reference image, such as a picture of a celebrity, and have FaceApp recreate the style on your photo.
Beware: the results are so realistic that after trying the tool in the office, my colleague is ready to finally take the plunge on a hairstyle she has wanted for about a year. Happy makeovers!
Hewlett Packard Enterprise (HPE) and Danfoss today announced a collaboration to deliver HPE IT Sustainability Services – Data Center Heat Recovery, an off-the-shelf heat recovery module that helps organisations manage excess heat as they transition to more sustainable IT facilities.
The solution combines HPE’s scalable Modular Data Center with Danfoss’ innovative heat reuse and cooling technologies.
The rapid integration of AI is expected to dramatically increase the power demand of IT infrastructure. The AI industry is projected to consume at least ten times more electricity in 2026 compared to 2023.
To mitigate this, the new energy-efficient data center solution from HPE and Danfoss incorporates direct liquid cooling to reduce energy consumption by 20%, and Danfoss heat reuse modules and oil-free compressors to capture excess data center heat and enhance cooling efficiency by up to 30%.
HPE’s Modular Data Center offers a power usage effectiveness (PUE) of 1.1, compared to 1.3-1.4 for traditional data centers. It enables faster deployment in as few as 6 months and supports compute-intensive AI workloads. Danfoss is leveraging its energy-efficient solutions to drive decarbonization, inspired by its use of recovered data center heat at its own carbon-neutral headquarters campus in Denmark.
“Our strategic partnership with HPE is a great example of how we revolutionize building and decarbonizing the data center industry together with customers,” said Jürgen Fischer, President, Danfoss Climate Solutions. “At HPE, we believe in the power of collaboration to create transformative solutions,” added Sue Preston, VP at HPE.
HPE IT Sustainability Services – Data Center Heat Recovery is available to order now as a modular solution with seamlessly integrated components from both companies. As part of a holistic sustainability approach, Danfoss is also partnering with HPE to retire and refurbish end-of-use IT assets through HPE Asset Upcycling Services.
A recent research paper “Finding Blind Spots in Evaluator LLMs with Interpretable Checklists” was released by the Indian Institute of Madras and AI4Bharat, an initiative for spearheading AI research in India. The paper reveals significant flaws in the current methods used by LLMs to evaluate text generation tasks.
Authored by researchers Sumanth Doddapaneni, Mohammed Safi Ur Rahman Khan, Sshubam Verma, and Mitesh M Khapra, FBI is a novel framework that is designed to assess how well Evaluator LLMs can gauge four critical abilities in other LLMs: factual accuracy, adherence to instructions, coherence in long-form writing, and reasoning proficiency.
The study involved introducing targeted alterations in answers generated by LLMs that impact these key capabilities, aiming to determine if Evaluator LLMs could detect drops in quality. A total of 2400 modified answers spanning 22 perturbation categories were created for the comprehensive study. Different evaluation strategies were applied to five prominent Evaluator LLMs frequently referenced in the literature.
Source: arxiv.org
Findings from the research revealed significant deficiencies in current Evaluator LLMs, which failed to identify declines in quality in over 50% of cases on average. Single-answer and pairwise evaluations exhibited notable limitations, while evaluations based on references demonstrated relatively better performance.
The study underscores the unreliable nature of current Evaluator LLMs and emphasises the necessity for cautious implementation in evaluating text generation capabilities. It is to be noted that Evaluator LLMs consistently missed basic errors, such as spelling and grammar mistakes.
Systems that require high-stakes decision-making, the reliability of their evaluations must be scrutinised. The study underscores the need for improved evaluation strategies and the potential risks of over-reliance on current LLM evaluators.
The FBI framework offers a path forward by providing a more interpretable and comprehensive method for testing evaluator capabilities. By revealing the prevalent failure modes and blind spots of existing models, this framework can guide the development of more robust and reliable AI evaluators.
As AI continues to reach into every domain of life, the question remains as to what kind of software these tools will run on. The choice in software stacks – or collections of software components that work together to enable specific functionality on a computing system – is becoming even more relevant in the GPU-centric computing needs of AI tasks.
With AI and HPC applications pushing the limits of computational power, the choice of software stack can significantly impact performance, efficiency, and developer productivity.
Currently, there are three major players in the software stack competition: Nvidia's Compute Unified Device Architecture (CUDA), Intel's oneAPI, and AMD's Radeon Open Compute (ROCm). While each has pros and cons, Nvidia's CUDA continues to dominate largely because its hardware has led the way in HPC and now AI.
Here, we will delve into the intricacies of each of these software stacks – exploring their capabilities, hardware support, and integration with the popular AI framework PyTorch. In addition, we will conclude with a quick look at two higher-level HPC languages: Chapel and Julia.
Nvidia's CUDA is the company's proprietary parallel computing platform and software stack meant for general-purpose computing on their GPUs. CUDA provides an application programming interface (API) that enables software to leverage the parallel processing capabilities of Nvidia GPUs for accelerated computation.
CUDA must be mentioned first because it dominates the software stack space for AI and GPU-heavy HPC tasks – and for good reason. CUDA has been around since 2006, which gives it a long history of third-party support and a mature ecosystem. Many libraries, frameworks, and other tools have been optimized specifically for CUDA and Nvidia GPUs. This long-held support for the CUDA stack is one of its key advantages over other stacks.
Nvidia provides a comprehensive toolset as part of the CUDA platform, including CUDA compilers like Nvidia CUDA Compiler (NVCC). There are also many debuggers and profilers for debugging and optimizing CUDA applications and development tools for distributing CUDA applications. Additionally, CUDA's long history has given rise to extensive documentation, tutorials, and community resources.
CUDA's support for the PyTorch framework is also essential when discussing AI tasks. This package is an open-source machine learning library based on the Torch library, and it is primarily used for applications in computer vision and natural language processing. PyTorch has extensive and well-established support for CUDA. CUDA integration in PyTorch is highly optimized, which enables efficient training and inference on Nvidia GPUs. Again, CUDA's maturity means access to numerous libraries and tools that PyTorch can use.
In addition to a raft of accelerated libraries, Nvidia also offers a complete deep-learning software stack for AI researchers and software developers. This stack includes the popular CUDA Deep Neural Network library (cuDNN), a GPU-accelerated library of primitives for deep neural networks. CuDNN accelerates widely used deep learning frameworks, including Caffe2, Chainer, Keras, MATLAB, MxNet, PaddlePaddle, PyTorch, and TensorFlow.
What's more, CUDA is designed to work with all Nvidia GPUs, from consumer-grade GeForce video cards to high-end data center GPUs – giving users a wide range of versatility within the hardware they can use.
That said, CUDA could be better, and Nvidia's software stack has some drawbacks that users must consider. To begin, though freely available, CUDA is a proprietary technology owned by Nvidia and is, therefore, not open source. This situation locks developers into Nvidia's ecosystem and hardware, as applications developed on CUDA cannot run on non-Nvidia GPUs without significant code changes or using compatibility layers. In a similar vein, the proprietary nature of CUDA means that the software stack's development roadmap is controlled solely by Nvidia. Developers have limited ability to contribute to or modify the CUDA codebase.
Developers must also consider CUDA's licensing costs. CUDA itself is free for non-commercial use, but commercial applications may require purchasing expensive Nvidia hardware and software licenses.
AMD's ROCm is another software stack that many developers choose. While CUDA may dominate the space, ROCm is distinct because it is an open-source software stack for GPU computing. This feature allows developers to customize and contribute to the codebase, fostering collaboration and innovation within the community. One of the critical advantages of ROCm is its support for both AMD and Nvidia GPUs, which allows for cross-platform development.
This unique feature is enabled by the Heterogeneous Computing Interface for Portability (HIP), which gives developers the ability to create portable applications that can run on different GPU platforms. While ROCm supports both consumer and professional AMD GPUs, its major focus is on AMD's high-end Radeon Instinct and Radeon Pro GPUs designed for professional workloads.
Like CUDA, ROCm provides a range of tools for GPU programming. These include C/C++ compilers like the ROCm Compiler Collection, AOMP, and AMD Optimizing C/C++ Compiler, as well as Fortran Compilers like Flang. There are also libraries for a variety of domains, such as linear algebra, FFT, and deep learning.
That said, ROCm's ecosystem is relatively young compared to CUDA and needs to catch up regarding third-party support, libraries, and tools. Being late to the game also translates to more limited documentation and community resources compared to the extensive documentation, tutorials, and support available for CUDA. This situation is especially true for PyTorch, which supports the ROCm platform but needs to catch up to CUDA in terms of performance, optimization, and third-party support due to its shorter history and maturity. Documentation and community resources for PyTorch on ROCm are more limited than those for CUDA. However, AMD is making progress on this front.
Like Nvidia, AMD also provides a hefty load of ROCm libraries. AMD offers an equivalent to cuDNN called MIOpen for deep learning, which is used in the ROCm version of PyTorch (and other popular tools).
Additionally, while ROCm supports both AMD and Nvidia GPUs, its performance may not match CUDA when running on Nvidia hardware due to driver overhead and optimization challenges.
Intel's oneAPI is a unified, cross-platform programming model that enables development for a wide range of hardware architectures and accelerators. It supports multiple architectures, including CPUs, GPUs, FPGAs, and AI accelerators from various vendors. It aims to provide a vendor-agnostic solution for heterogeneous computing and leverages industry standards like SYCL. This feature means that it can run on architectures from outside vendors like AMD and Nvidia as well as on Intel's hardware.
Like ROCm, oneAPI is an open-source platform. As such, there is more community involvement and contribution to the codebase compared to CUDA. Embracing open-source development, oneAPI supports a range of programming languages and frameworks, including C/C++ with SYCL, Fortran, Python, and TensorFlow. Additionally, oneAPI provides a unified programming model for heterogeneous computing, simplifying development across diverse hardware.
Again, like ROCm, oneAPI has some disadvantages related to the stack's maturity. As a younger platform, oneAPI needs to catch up to CUDA regarding third-party software support and optimization for specific hardware architectures.
When looking at specific use cases within PyTorch, oneAPI is still in its early stages compared to the well-established CUDA integration. PyTorch can leverage oneAPI's Data Parallel Python (DPPy) library for distributed training on Intel CPUs and GPUs, but native PyTorch support for oneAPI GPUs is still in development and is not ready for production.
That said, it's important to note that oneAPI's strength lies in its open standards-based approach and potential for cross-platform portability. oneAPI could be a viable option if vendor lock-in is a concern and the ability to run PyTorch models on different hardware architectures is a priority.
For now, if maximum performance on Nvidia GPUs is the primary goal for developers with PyTorch workloads, CUDA remains the preferred choice due to its well-established ecosystem. That said, developers seeking vendor-agnostic solutions or those primarily using AMD or Intel hardware may wish to rely on ROCm or oneAPI, respectively.
While CUDA has a head start regarding ecosystem development, its proprietary nature and hardware specificity may make ROCm and oneAPI more advantageous solutions for certain developers. Also, as time passes, community support and documentation for these stacks will continue to grow. CUDA may be dominating the landscape now, but that could change in the years to come.
In general, many developers prefer to create hardware-independent applications. Within HPC, hardware optimizations can be justified for performance reasons, but many modern-day coders prefer to focus more on their application than on the nuances of the underlying hardware.
PyTorch is a good example of this trend. Python is not known as a particularly fast language, yet 92% of models on Hugging Face are PyTorch exclusive. As long as the hardware vendor has a PyTorch version built on their libraries, users can focus on the model, not the underlying hardware differences. While this portability is nice, it does not guarantee performance, which is where the underlying hardware architecture may enter the conversation.
Of course, Pytorch is based on Python, the beloved first language of many programmers. This language often trades ease of use for performance (particularly high-performance features like parallel programming). When HPC projects are started with Python, they tend to migrate to scalable high-performance codes based on distributed C/C++ and MPI or threaded applications that use OpenMP. These choices often result in the "two language" problem, where developers must manage two versions of their code.
Currently, two "newer" languages, Chapel and Julia, offer one easy-to-use language solution that provides a high-performance coding environment. These languages, among other things, attempt to "abstract away" many of the details required to write applications for parallel HPC clusters, multi-core processors, and GPU/accelerator environments. At their base, they still rely on vendor GPU libraries mentioned above but often make it easy to build applications that can recognize and adapt to the underlying hardware environment at run time.
Initially developed by Cray, Chapel (the Cascade High Productivity Language) is a parallel programming language designed for a higher level of expression than current programming languages (read as "Fortran/C/C++ plus MPI"). Hewlett Packard Enterprise, which acquired Cray, currently supports the development as an open-source project under version 2 of the Apache license. The current release is version 2.0, and the Chapel website posts some impressive parallel performance numbers.
Chapel compiles to binary executables by default, but it can also compile to C code, and the user can select the compiler. Chapel code can be compiled into libraries that can be called from C, Fortran, or Python (and others). Chapel supports GPU programming through code generation for Nvidia and AMD graphics processing units.
There is a growing collection of libraries available for Chapel. A recent neural network library called Chainn is available for Chapel and is tailored to build deep-learning models using parallel programming. The implementation of Chainn in Chapel enables the user to leverage the parallel programming features of the language and to train Deep Learning models at scale from laptops to supercomputers.
Developed at MIT, Julia is intended to be a fast, flexible, and scalable solution to the two-lanague problem mentioned above. Work on Julia began in 2009, when Jeff Bezanson, Stefan Karpinski, Viral B. Shah, and Alan Edelman set out to create an open technical computing language that was both high-level and fast.
Like Python, Julia provides a responsive interpretive programming environment (REPL or read–eval–print loop) using a fast, just-in-time compiler. The language syntax is similar to Matlab and provides many advanced features, including:
Julia also has GPU libraries for CUDA, ROCm, OneAPI, and Apple that can be used with the machine learning library Flux.jl (among others). Flux is written in Julia and provides a lightweight abstraction over Julia's native GPU support.
Both Chapel and Julia offer a high-level and portable approach to GPU programming. As with many languages that hide the underlying hardware details, there can be some performance penalties. However, developers are often fine with trading a few percentage points of performance for ease of portability.
C++, a language that once shone brightly in the late twentieth century, was at the forefront of technological advancements, particularly in space exploration.
However, the emergence of newer, more visually appealing programming languages has shifted the spotlight away from C++.
At the AI+Data Summit 2024, researcher Yejin Choi said that researchers no longer use the language for AI research.
So, is C++ becoming a relic of the past?
Despite its performance benefits and applications in various AI fields, such as speech recognition and computer vision, C++ is not the go-to language for AI development.
Its complexity and steep learning curve pose significant challenges. In contrast, Python’s user-friendly nature, extensive libraries, and large developer communities have propelled it to the forefront of AI programming.
Furthermore, C++ involves manual memory management, which can result in memory leaks and errors if not done correctly. This can be a considerable issue, particularly in large-scale AI programmes.
Microsoft emphasised this issue when it revealed that 70% of its updates in the previous 12 years were solutions for memory safety bugs, owing to Windows being mostly written in C and C++.
Google’s Chrome team released their own research, which revealed that memory management and safety flaws accounted for 70% of all major security bugs in the Chrome codebase. It is largely written in C++.
C++ also lacks built-in support for garbage collection, database access, and threading, which can necessitate extra effort to develop.
This can be particularly challenging in AI applications that require concurrent processing of data and tasks, such as deep learning and neural networks, real-time systems and embedded systems, data processing, and data science.
To overcome these limitations, developers often use third-party libraries and frameworks that provide threading support, such as OpenMP or Boost. However, these libraries can add complexity and overhead to the code, which may only be ideal for some applications.
If you’ve visited a page like the C++ FAQ, you’ll understand how hard C++ can be. A comma in the wrong location might trigger hundreds of compile errors in earlier language versions.
The language has improved since C++ 11, with move semantics for transferring ownership and rvalue references, although there is still a high learning curve.
In recent years, we’ve witnessed the growth of various programming languages that potentially replace C++ for low-level system tasks, like Rust, which provides safety and security by eliminating buffer overflows and memory leaks (and is much easier to learn than C++).
When you compare the feature sets of modern languages like C++, Python, and Rust, the C language begins to look like a dinosaur! The C standard has not had new features introduced since 2011!
The 2017 standard release included technical corrections and clarifications, and the 2023 standard release did not rock the boat either.
Mark Russinovich, the chief technical officer of Microsoft Azure, has stated that developers should stop creating code in the programming languages C and C++ and that the industry should treat these computer languages as “deprecated”.
Ken Thompson, the Bell Labs researcher who designed the original Unix operating system, called it a “bad language” that is “way too big, way too complex” and “obviously built by a committee”.
GitHub compiled a list of the top ten most popular programming languages for machine learning. Python is the most popular language in machine learning repositories, with C++ being sixth.
According to Stack Overflow’s Developer Survey, beginners beginning to code are more likely to prefer Python over C++ than professionals.
While C++ provides advantages regarding speed and memory management, it also has disadvantages, such as a high learning curve and little community assistance.
Despite its challenges, C++ can be a powerful choice for machine learning applications that require high-performance processing and advanced memory management. The choice between C++ and Python for machine learning ultimately depends on the specific needs of the application and the developers’ skill level.
A few days ago, a video posted by comedian Vishal Kal went viral. In the video, Kal acted out a scenario where an Indian-origin techie got laid off while working in the US and was replaced by an Indian working from India. Everyone took the satire seriously, with many covering it with the title – “They’re taking our jobs.”
Well it turns out that though the video was a satire, there is some truth to what the comedian had pointed out. In a recent discussion, a user on Reddit pointed out that a lot of US citizens are hesitant to get into the field of computer science as companies outsource a lot of their work to India, taking advantage of workers who are ready to work for lower wages.
Adding to all of this is the rise of AI. “The field is doomed due to the rise of AI and outsourcing to India, making groups with 100 software developers into 3 with one managing a team in India, and the other two polishing the code to be used,” explained the user.
The explanation is simple – salaries required to survive in the US are higher than those required in India. Therefore, Indian coders that stay in India are cheaper to hire than those in the US. Adding to that is the fact that it has become easier to replace low-level software developers with the help of one person who knows how to use AI.
Outsourcing is not new, though. For example, when it comes to data labelling and keeping AI models running, companies such as OpenAI, Amazon, and Meta have all been using cheaper labour in countries such as India and Kenya, earning less than $2 an hour. But that is a story for another time.
The sentiment that Indians have been stealing jobs from Americans has been longstanding. Now, the issue is that Indians who run away abroad for better prospects are also realising that they are too expensive to be hired by employers. There’s a highly likely chance that a developer with the same skill set, or better, is available for a lesser salary in India.
Now, with coding becoming available in natural languages, including Hindi and other Indian languages, the situation is probably going to get worse for coders in the US.
On the other hand, it has been observed that several of the graduates from Indian institutes are barely employable as they lack basic skills in coding. Meanwhile, there are jobs for 10 times the available developers, specifically for those who are able to handle multiple roles with the help of AI. “We’d rather hire one software engineer who knows how to use AI than five who don’t, even if it’s the same cost,” recalled Allie K Miller recently.
On the Reddit post, another user replied that there is still time to upskill yourself and graduate with higher grades. “Many of the issues you bring up were issues 2 decades ago, and US tech workers did just fine,” they added.
This is the moment to bring up another recent conversation about how India might soon run out of skilled software engineers as most of them don’t stay in the country due to the education system not upskilling them the way it needs to be.
Moreover, there is a common sentiment that Indians abroad do not support each other within the same organisation, instead choosing to compete with each other.
If Indians are easily employed if they stay within the country, what is even the point of going abroad? All one needs to do is upskill themselves and apply for remote jobs, which, quite honestly, is a reasonable thing to do. As for the Indians going to the US for jobs, it might be time to come back and start building for the country.
While universities abroad offer more hands-on education when compared to Indian ones, they are also expensive. This, in the end, forces graduates from those universities to demand a higher salary. Thus, they often seem overvalued by many employers.
Meanwhile, Indian techies are also not behind when it comes to demanding higher salaries. The skyrocketing salaries at Silicon Valley startups such as OpenAI and Anthropic, coupled with the confidence of having ‘upskilled’ in GenAI, are pushing Indian software engineers to expect exorbitant salaries.
Eventually, the market will reset. Indians demanding higher salaries will become unattractive to companies in the US, and they may end up employing ones within the country. It’s all part of course correction, leading Indians to start working within the country for Indian companies.
As for the US, it might miss out on Indian tech talent pretty soon.
In a world that’s paranoid about data privacy, Databricks CEO Ali Ghodsi firmly believes that everyone should own their data. Speaking at Databricks’ Data + AI Summit, Ghodsi warned companies against using vendors, or distributing their data to anyone, including themselves.
“The idea is to stop giving your data to vendors. They’ll just lock you in. It doesn’t matter if it’s a proprietary data warehouse in the cloud, or if it’s Snowflake or even Databricks. Don’t give your data to us; don’t trust vendors,” he reiterated.
This is surprising in and of itself since data being deemed the new oil, companies have been rushing to sell their own to reap the benefits.
Ghodsi, however, has a point. With so many companies pivoting towards trying to get the best use of their data, many of them fall into the trap of overcomplicating things.
In his experience, Ghodsi said that executives at several companies have admitted to not knowing or understanding the technology behind how their data is being used, due to the sheer amount of software, including multiple data warehouses, data science and machine learning platforms, and data lakes.
Ultimately, this leads to the company’s data being locked into a silo, removing easy access and increasing costs for the company overall.
Elaborating further, Databricks’ VP of field engineering APJ, Nick Eayrs told AIM that it had always been the company’s strategy to democratise data and AI. “That’s kind of our mission and purpose for being. It starts with ensuring you have control over your data,” Eayrs said.
This is the goal Databricks has been moving towards, in effectively getting company data to a point where vendors can plug their “USB sticks” into a company’s data. This ends up giving the company power over their data and how a vendor uses it, as well as ensuring that their data is being used in the most optimised way.
“They should just plug their USB stick into that data that you have in the cloud and then let the best engine win. Let’s see who’s best,” Ghodsi challenged
This also makes sense when it comes to how rapidly the industry is changing. When comparing models, Ghodsi admitted that Databricks’ DBRX model was the best open-source LLM in the market for a whole two weeks, before being outperformed by LLaMa 3, which released only a few weeks later.
Allowing customers to have the freedom to allow vendors to use their data in a controlled environment means that the companies themselves are able to better access how their data is used.
“That puts them in control of their data, which is ultimately their secret sauce. That’s what’s going to differentiate their products and services. We want them to own and control their data, and we want their data to be in an open format in a cloud of their choice. Even if they choose to take it back on-prem, so be it,” Eayrs asserted.
Obviously, as an AI company, Databricks has also begun working on how to ensure their customers come back to them when they need a vendor. However, while there are several companies already working on this, Databricks stands out with its focus on democratisation.
The company’s recent acquisition of Tabular is a testament to that. In an effort to ensure companies don’t have the problem of being confined to silos yet again, only this time in lakehouse format, the Tabular acquisition solved this problem.
“You don’t have to pick which of the two silos I have to go through, and which of the USB formats I must store this in? We don’t want it to be that way,” Ghodsi said.
While they’re currently focusing on democratising data for their clients, Eayrs said that the next steps are to ensure that customers can get the most out of their data. “Once they have their data there and they have it governed and secured, how do we help them accelerate the time to insight and value? That’s where we want to lean in and break some of the magic,” he told AIM.
upGrad has announced a new certification program in GenAI for Leaders in collaboration with Microsoft and the International Institute of IIIT Bangalore. This unique 4-month course is designed to train mid-career and seasoned professionals with core AI competencies through applied learning concepts.
The industry-first curriculum leverages an inclusive framework, enabling learners to assess, design, and transform business problems by applying acquired knowledge in real-life situations. The program is bolstered by strong tech faculty support from IIIT Bangalore and includes live sessions with Microsoft.
Key elements of the pedagogy include a Build Your Own Business (BYOB) case, engagement with industry stalwarts, six comprehensive modules, curated capstone projects, reverse-engineered content and workshops with Microsoft, and access to cutting-edge AI tools such as ChatGPT and DALL-E.
“As Generative AI revolutionises the workforce, mastering AI-driven productivity tools is crucial for professionals,” said Venkat Krishnan, Executive Director of Public Sector, Healthcare, and Education at Microsoft India. “By joining forces with upGrad and IIIT Bangalore, we offer real-world exposure and seamlessly integrate AI tools into educational frameworks.”
Dr. V Sridhar, Professor In-charge of Continuing Professional Education at IIIT Bangalore, emphasised the program’s alignment with market requirements, stating, “When technologies such as these get embedded into businesses, it is the need of the hour for senior leaders to understand how such technologies can be deployed optimally within their organisations to reap the maximal benefits for all stakeholders.”
“This course is not about teaching leadership in the AI era but empowering leaders to adopt and lead GenAI initiatives in a practically,” added Mayank Kumar, Co-founder & MD of upGrad. “Our collaboration with Microsoft and IIIT Bangalore brings together the best of academia and industry, providing professionals with the skills needed to stay ahead.”
Learners who complete the course and graded projects will be eligible for a completion certification from IIIT Bangalore, along with a 1-day offline immersion at both the Microsoft Development Centre and the IIIT Bangalore campus
Accenture announced over $900 million in new bookings for generative AI, reaching a total of $2 billion fiscal year-to-date during its Q3 FY24 earnings report.
“We have achieved two significant milestones this quarter — with $2 billion in generative AI sales year-to-date and $500 million in revenue year-to-date — demonstrating our early lead in this critical technology,” the company said in a statement.
“We are helping a leading global food and beverage company, which has already built a strong digital core as part of its reinvention journey, to now leverage the power of generative AI to create new value,” said Accenture CEO Julie Sweet during the earnings call.
She added that they have developed a digital shelf console pilot, a GenAI engine that accelerates content creation for e-commerce and optimises it to drive sales.
“We also continue to steadily increase our data and AI workforce, reaching approximately 55,000 skilled data and AI practitioners, against our goal of doubling our data and AI workforce from 40,000 to 80,000 by the end of FY2026,” Sweet stated.
The global IT giant has partnered with National Australia Bank, one of the country’s largest financial institutions, to strategically implement and scale generative AI. “We worked on a methodical build of a secure and robust GenAI platform within the bank’s existing strategic data platform, creating 200 generative AI use cases in backlog. To date, over 20 use cases have been tested across the bank, with eight enterprise-grade pilots underway, several of which are already delivering value,” she added.
Accenture is also assisting Saudia Airlines, the national flag carrier of Saudi Arabia, in launching an innovative digital platform to transform the traveler’s experience. “Powered by GenAI, the platform will provide a one-stop solution enabling customers to seamlessly plan their journeys, book flights, and modify their trips with just a few words, all while providing a personalized and conversational experience,” said Sweet.
Last year, Accenture announced a $3 billion investment in its data and AI practice. At the World Economic Forum in Davos, Sweet reiterated Accenture’s commitment, allocating $1 billion annually to train its workforce in generative AI.
Additionally, Accenture recently collaborated with Anthropic and AWS to train over 1,400 Accenture engineers as specialists in utilizing Anthropic’s models on AWS. The company also partnered with Cohere to accelerate the adoption of generative AI across enterprises, utilizing Cohere’s Command and Embed models, along with its Retrieval Augmented Generation (RAG) capabilities, to help organizations scale the use of generative AI.
