Image: putilov_denis/Adobe StockMicrosoft and OpenAI person gone hand-in-hand for a agelong time, from Microsoft’s archetypal backing of the institution down ChatGPT to embedding GPT services wrong the Azure AI platform. Azure OpenAI Service is astatine the “forefront” of a generative AI translation that includes GPT-4, portion the Azure AI infrastructure is the “backbone,” the Redmond tech elephantine wrote successful a blog station detailing updates to the Azure AI platform connected Monday.
Two large changes — caller offerings from OpenAI and upgraded virtual instrumentality hardware — amusement Microsoft’s ongoing committedness to putting generative AI successful play successful much ways.
Jump to:
- New models, including GPT-4, travel to Azure OpenAI
- What caller NVIDIA-powered VMs mean for Azure customers
- Red teaming needs to accommodate to the behaviors of generative AI
- Competitors to Azure AI
New models, including GPT-4, travel to Azure OpenAI
OpenAI’s GPT-4 and GPT-35-Turbo volition present beryllium disposable done Azure OpenAI Service successful 4 caller regions: eastbound Canada, eastbound Japan, confederate U.K. and an further swath of the eastbound U.S. (East US 2 connected the availability map). GPT-4 is the astir precocious generative AI exemplary disposable from OpenAI today.
Azure OpenAI Service has astir 11,000 customers who usage it for tasks specified arsenic lawsuit service, penning contented and analyzing documents.
“What we’re seeing is that the ChatGPT exertion [from Azure OpenAI] is helping users make contented that is much relevant, personalized, adjacent much creative,” Aprimo main merchandise serviceman Kevin Souers said.
Existing Azure OpenAI customers tin present articulation a waitlist for entree to GPT-4.
SEE: Threat actors spun ChatGPT retired into the malicious WormGPT. Learn astir this and different possible ChatGPT-related information risks. (TechRepublic)
What caller NVIDIA-powered VMs mean for Azure customers
A caller virtual instrumentality series, the Azure ND H100 v5, are present mostly disposable successful the East U.S. and South Central U.S. Azure regions for existing endeavor customers. These VMs are designed to assistance organizations plan and tally generative AI applications.
The caller hardware, NVIDIA H100 Tensor Core GPUs and NVIDIA Quantum-2 InfiniBand networking, is tuned for AI performance. The low-latency networking includes NVIDIA Quantum-2 ConnectX-7 InfiniBand with 400Gb/s per GPU with 3.2Tb/s per VM of cross-node bandwidth for supercomputer-level performance.
The PCIe Gen5 information transportation modular successful the ND H100 v5 VMs, guided GPU show to 64GB/s bandwidth per GPU for improved show betwixt the CPU and GPU, Nidhi Chappell, wide manager of Azure AI infrastructure, and Eric Boyd, firm vice president for AI platforms astatine Microsoft, elaborate successful a blog post.
Operations connected the ND H100 v5 VMs volition beryllium capable to beryllium performed faster, and immoderate ample connection models volition spot 2 times faster speeds erstwhile tally connected them, Microsoft said.
Red teaming needs to accommodate to the behaviors of generative AI
Safety and information stay concerns erstwhile it comes to AI. Microsoft assures customers that it ” … incorporates robust information systems and leverages quality feedback mechanisms to grip harmful inputs responsibly.” Microsoft besides encourages red teaming AI applications oregon inviting ethical hackers to play the relation of menace actors to spot however AI applications mightiness beryllium susceptible to attack.
Microsoft’s guidelines for reddish teams gearing up to enactment with AI include:
- Focusing connected information and liable AI outcomes, meaning making definite the output isn’t violative oregon dangerous.
- Working wrong the thought that not lone malicious interactions but besides benign ones could effect successful unwanted outputs.
- Making definite some antiaircraft and violative techniques are precise thorough.
Competitors to Azure AI
Competitors to Microsoft’s Azure AI see Amazon Web Services, IBM Watson, Google AI, DataRobot’s customized AI exemplary service, Salesforce Einstein AI for marketing, ServiceNow AIOps for IT operations management, Oracle Cloud Infrastructure and H2O.ai.
English (US) ·