AMD turns to AI startups to inform chip, software design
FILE PHOTO: A smartphone with a displayed AMD logo is placed on a computer motherboard in this illustration taken March 6, 2023. REUTERS/Dado Ruvic/Illustration/File Photo
SAN JOSE (Reuters) -Advanced Micro Devices has forged close ties to a batch of artificial intelligence startups as part of the company's effort to bolster its software and forge superior chip designs.
As AI companies seek alternatives to Nvidia's chips, AMD has begun to expand its plans to build a viable competing line of hardware, acquiring companies such as server maker ZT Systems in its quest to achieve that goal.
But to build a successfulline of chips also requires a powerful set of software to efficiently run the programs built by AI developers. AMD has acquired several small software companies in recent weeks in a bid to boost its talent, and it has been working to beef up its set of software, broadly known as ROCm.
"This will be a very thoughtful, deliberate, multi-generational journey for us," said Vamsi Boppana, senior vice president of AI at AMD.
AMD has committed to improve its ROCm and other software, which is a boon to customers such as AI enterprise startup Cohere, as it results in speedy changes and the addition of new features.
Cohere is focused on building AI models that are tailored for large businesses versus the foundational AI models that companies like OpenAI and others target.
AMD has made important strides in improving its software, Cohere CEO Aidan Gomez said in an interview with Reuters. Changing Cohere's software to run on AMD chips was a process that previously took weeks and now happens in only "days," Gomez said.
Gomez declined to disclose exactly how much of Cohere's software relies on AMD chips but called it a "meaningful segment of our compute base" around the world.
OPENAI INFLUENCE
OpenAI has had significant influence on the design of the forthcoming MI450 series of AI chips, said Forrest Norrod, an executive vice president at AMD.
AMD's MI400 series of chips will be the basis for a new server called "Helios" that the company plans to release next year. Nvidia too has engineered whole servers in part because AI computations require hundreds or thousands of chips strung together.
OpenAI's Sam Altman appeared on stage at AMD's Thursday event in San Jose, and discussed the partnership between the two companies in broad terms.
Norrod said that OpenAI's requests had a big influence on how AMD designed the MI450 series memory architectureand how the hardware can scale up to thousands of chips necessary to build and run AI applications.
The ChatGPT creator also influenced what kinds of mathematical operations the chips are optimized for.
"(OpenAI) has given us a lot of feedback that, I think, heavily informed our design," Norrod said.
(Reporting by Max A. Cherney in San JoseEditing by Shri Navaratnam)
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
The Star
9 hours ago
- The Star
Jony Ive deal removed from OpenAI site over trademark suit
Bloomberg reported last week that a judge was considering barring OpenAI from using the IO name due to a lawsuit recently filed by the similarly named IYO Inc, which is also building AI devices. — Reuters Marketing materials and video related to a blockbuster partnership between former Apple Inc designer Jony Ive and OpenAI Inc were removed from the web due to a trademark dispute. Social media users noticed Sunday that a video and website hosted by OpenAI that announced the artificial intelligence company's US$6.5bil (RM28bil) acquisition of Ive's secretive AI hardware startup IO Products were no longer online, prompting speculation about a hiccup in the agreement. Public-facing materials were pulled due to a trademark dispute, according to spokespeople for Ive and OpenAI. Bloomberg reported last week that a judge was considering barring OpenAI from using the IO name due to a lawsuit recently filed by the similarly named IYO Inc, which is also building AI devices. "This is an utterly baseless complaint and we'll fight it vigorously,' a spokesperson for Ive said on Sunday. – Bloomberg
The Star
a day ago
- The Star
Advancing autism screening
A NEW artificial intelligence (AI) model is reshaping how autism spectrum disorder (ASD) is screened, offering faster, more accurate and more scalable detection across age groups. Developed through an international research collaboration that included INTI International University, the model achieved up to 99.06% accuracy in identifying ASD in children, adolescents and adults, outperforming many existing diagnostic systems. At the centre of the study is the Chaotic Binary Butterfly Optimisation Algorithm-based Feature Selection and Data Classification (CBBOAFS-DC) model. It analyses behavioural and developmental data to identify the traits most relevant to autism, supporting healthcare professionals in making earlier assessments. For many families, autism diagnosis is often delayed due to long wait times, limited access to specialists, or a lack of services in rural and underserved areas. These delays can mean children miss out on early intervention support that has been shown to improve communication, behaviour and learning outcomes significantly. Traditional screening tools also face consistency challenges, particularly when applied across different age groups or populations. INTI International University Faculty of Data Sciences and Information Technology academic Prof Dr Siti Sarah Maidinat (pic) said the new model is not a replacement for conventional methods. 'It's a tool to support healthcare providers, especially in early screening. The earlier autism is detected, the earlier children can receive the necessary interventions. 'We're helping to bridge the gap between limited diagnostic resources and the growing need for early autism detection. 'The idea is to support, not substitute, clinical expertise – especially in underprivileged communities,' she said. CBBOAFS-DC uses AI to process data from screening tools quickly and consistently, reducing reliance on time-intensive manual evaluations. Its hybrid architecture sets it apart from conventional approaches, according to an INTI press release dated May 23. While traditional models often rely on algorithms like decision trees or logistic regression, the new model uses a three-step design. 'It selects only the most relevant features using a butterfly-inspired optimisation technique, applies a deep learning engine to detect subtle behavioural patterns, and then finetunes its internal parameters using a bacterial colony optimisation algorithm,' the press release read. It added that this layered approach allows the model to handle the complexity of ASD traits across age groups and adapt to a variety of screening contexts – bringing researchers closer to a more universal screening solution. The research was conducted in collaboration with Sri Manakula Vinayagar Engineering College and Vellore Institute of Technology in India, as well as institutions in Saudi Arabia.
The Star
5 days ago
- The Star
How to generate images with ChatGPT on WhatsApp
You can easily add ChatGPT as a new contact on WhatsApp. — AFP For several months, all WhatsApp users have been able to communicate directly with ChatGPT via a dedicated phone number. It's now also possible to use this hotline to generate images, then share them with contacts. OpenAI has announced that all WhatsApp users can now generate images directly via ChatGPT, whether or not they have an account with the service. Simply add the US phone number 1-800 CHATGPT (1-800-242-8478) to your WhatsApp contacts to start a conversation directly with the AI service. From then on, a simple prompt, ie, a request written in natural language, can be used to generate an image. Note that it is also possible to send a photo and ask the AI to change it in various ways, such as applying a cartoon style, for example. These AI-generated images can then be easily shared in WhatsApp chats or groups, just like any other photo. This feature, directly integrated into WhatsApp, is based on GPT-4o, a model capable of processing text and images. All WhatsApp users can generate one image per day, free of charge. This quota increases if you link up your ChatGPT account, whether free or paid. For OpenAI, this is another step towards making its AI tools accessible to as many people as possible, the idea being to democratize AI-assisted image generation. In this sense, WhatsApp, with over two billion users worldwide, is a prime channel. But OpenAI is also stepping on the toes of WhatsApp's owner, Meta, which has already integrated its own AI tool into the application. It can answer a multitude of questions, but cannot yet generate images. – AFP Relaxnews
