Top 5 Stories of the Week: DALL-E Usage, Nvidia’s Digital Twins, Microsoft’s AI Inference, Intel Detects Deepfakes

Since the release of DALL-E into public beta last month, several companies have started integrating its use for different cases in the artificial intelligence (AI) landscape. Tome, a storytelling and ideation platform, announced this week that its interactive slide functionality is now supported by DALL-E technology. Users can apply DALL-E to help them with presentational images to convey exactly what they envision. Tome says it’s also working with GPT-3 to add more generative AI functionality to its platform in the near future.

Also on the generative AI spectrum, Microsoft revealed this week that its Project Bonsai reinforcement learning will be supported by d-Matrix DIMC technology. The goal is to speed up AI inference. For context, generative AI’s use of transformer models is necessary for its functionality, but it is also a resource-intensive process. Inference systems in AI help predict and build results from a model. Microsoft’s move to speed up the process will help increase the efficiency and deployment of generative AI models.

Nvidia also made progress this week announcing enhancements aimed at enhancing the Omniverse, extending scientific applications on top of powerful computing systems. The company said it allows digital twins to bring together the data that is currently separate from different apps, models and user experiences. Lead product manager of accelerated computing, Dion Harris, said it’s a step toward evolving digital twins from passive modeling to actively shaping the world.

Meanwhile, Intel’s news this week focused on shaping the world in a different way: eliminating deepfakes. The company introduced a new tool called the FakeCatcher, which it claims has a 96% accuracy rate and works by analyzing the “blood flow” of an image or video and returning the results in real time.

Unsurprisingly, along with the emergence of new technologies such as deepfakes and new moves in AI, there is a need for more security across all industries. In a special VentureBeat report on zero-trust security released this week, our writers highlight how security is being challenged and why a zero-trust approach is the future. Part of the in-depth look also examines the ways some companies are mishandling zero trust, including a failure to understand what zero trust is at its core and how to properly apply it.

Here’s more from our top 5 tech stories of the week:

1. New DALL-E integration adds generative AI for next-level slides

   Tome, announced interactive slide options supported by OpenAI’s DALL-E technology. The company, which calls itself the “new storytelling format for work and important ideas,” says adding a generative AI dimension to decks made sense.

   “It just felt very natural to make that part of the storytelling creation experience,” Tome CEO Keith Peiris told VentureBeat. “It felt so much more powerful than searching for a stock photo or clip art — it kind of gives us a first look at what generative storytelling can look like.”

2. Nvidia Omniverse supporting scientific digital twins

   Nvidia has announced several major advancements and partnerships to extend the Omniverse to scientific applications on top of high performance computing systems (HPC).

   This will support scientific digital twins that merge data silos that currently exist across different apps, models, tools and user experiences.

3. Why Enterprises Get Zero Trust Wrong

   The reality of zero-trust adoption is that it’s a journey, not a destination. There is no quick fix for implementing zero trust as it is a security method designed to be applied continuously throughout the environment to control user access.

   One of the main reasons companies get zero trust wrong is not only understanding what zero trust is, but knowing how to apply it and what products can implement zero trust.

4. New Microsoft partnership accelerates generative AI developmentMicrosoft and d-Matrix announced that Microsoft Project Bonsai reinforcement learning will be supported on d-Matrix DIMC technology, which the two vendors hope will provide significant acceleration for AI inference.

   “Project Bonsai is a platform that enables our version of deep reinforcement learning and we call it machine learning,” Kingsuk Maitra, chief applied AI engineer at Microsoft, told VentureBeat.

5. Intel unveils real-time deepfake detector, claims 96% accuracy rate

   On Monday, Intel introduced FakeCatcher, which is said to be the first real-time detector of deepfakes, i.e. synthetic media in which a person in an existing image or video is replaced with someone else’s likeness.

   Intel claims the product has a 96% accuracy rate and works by analyzing the subtle “blood flow” in video pixels to return results in milliseconds.