Tue Jan-9-2024: NIST Report Highlights Key AI Attack Types, Self-Evolving AI Models, Vision Pro Date

[00:00:02] It's Tuesday, January 9th, 2024, and I'm Dave Sobel. Three things to know today. A NIST report identifies the key attack types in AI, evasion, poisoning, privacy, and abuse, self-evolving tiny AI models, Snapchat pluses 7 million subs and more there, and the Vision Pro gets a release date. This is the Business of Tech.

[00:00:27] How about I start with the warnings? The U.S. National Institute of Standards and Technologies has warned about the security and privacy risks associated with the rapid deployment of artificial intelligence systems. These risks include adversarial manipulation of training data, exploitation of model vulnerabilities, and exfiltration of sensitive information.

[00:00:48] The report identifies four key types of attacks, evasion, poisoning, privacy, and abuse. NIST calls for better defenses and highlights the need for robust mitigation measures to counter these risks. Forrester Consulting's survey of 220 AI decision makers reveals that while organizations understand the transformative potential of generative AI, there are still barriers to its adoption.

[00:01:15] These include concerns about data protection and privacy laws, developing skills and governance, and the risk of biases and hallucinations. Inadequate data infrastructure is identified as the most significant barrier, followed by difficulties in integration and computational limitations. Organizations can overcome these challenges by adopting AI platforms that provide collaborative capabilities and prepackaged solutions.

[00:01:41] According to a survey by Dark Reading, there is significant interest in using generative AI tools across various departments within enterprises. The most common use cases include data analytics, cybersecurity, research, and marketing. While some organizations are actively using generative AI tools, many are still exploring or considering their use. The benefits of generative AI include increased task completion speed and automation of routine tasks.

[00:02:10] Why do we care? The adoption of generative AI is expected to be widespread, with security teams relying on it to improve operations and protect against potential problems. For IT service providers, this means an opportunity to develop and offer AI security solutions, such as secure AI training environments, model vulnerability assessments, and solutions for sensitive data protection, particularly for customers who do not have in-house capabilities.

[00:02:37] There is also the opportunity for consulting around solid data governance, privacy controls, and user-friendly interfaces that require minimal specialized skills. Furthermore, provide training and governance frameworks and help organizations navigate the complexities of AI adoption more effectively. And a couple of product items.

[00:02:59] Scientists have developed a method to use large AI models to automatically create smaller AI tools without human intervention. The breakthrough allows for the replication and evolution of AI models, paving the way for self-evolving AI. The smaller AI models, known as TinyML, can be used for specific tasks at a low cost and have the potential to be integrated into various devices. And I didn't expect this to be the next big AI product.

[00:03:28] Snapchat Plus has reached 7 million paying subscribers, offering generative AI features such as an image generator, a GPT-4 powered chatbot, and a Dreams tool. This success indicates that generative AI may have a different future than expected, with companies adding AI features to enhance existing products.

[00:03:49] Snapchat's early success in this area is logical, as AI image generation enhances communication on a camera-first messaging app. This new business line could bring in significant revenue for Snap, which has struggled to maximize the value of its ad space. Hats.ai, co-founded by Jimmy Hatzel and Aiden Kehoe, has raised $2.5 million in seed funding to enable managed services providers to deliver AI as a service.

[00:04:17] The platform offers AI applications, AI agents, vector storage, and custom large language models managed through a multi-tenant platform. The company aims to empower MSPs worldwide to build AI as a service businesses and will make their products generally available in March 2024.

[00:04:37] New research predicts that AI-capable PCs will become the norm within the next 5-7 years, revitalizing the PC sector and transforming user experiences in terms of efficiency, productivity, and creativity. The report suggests that 60% of PCs in circulation will possess AI capabilities by 2027, with initial adoption driven by the commercial sector.

[00:05:01] PC manufacturers like Dell, HP, and Lenovo are preparing for an AI future, and chip developers such as Intel, AMD, Apple, and Qualcomm are going to pay a critical role. Why do we care? I'll start with that Hats.ai investment. I included it to show the investment dollars are flowing quickly into AI in every sector, and we should expect to see MSPs get attention. This isn't the first AI as a service offering we've discussed.

[00:05:31] The success of Snapchat Plus in attracting 7 million paying subscribers with its generative AI features underscores the commercial potential of integrating AI into existing products. Expect to see a lot more of that, and be driven into smaller and smaller packages via those smaller LLMs and into the broader computing market that is driven to the end point. And finally, let's note the date.

[00:05:58] Apple has announced that its Vision Pro headset will be available for shipping on February 2nd in the United States, with pre-orders starting on January 19th. Priced at $3,499, the Vision Pro offers a 4K display for each eye, supports both augmented and virtual reality, and is powered by Apple's in-house M.2 chip.

[00:06:20] The headset runs on Vision OS, Apple's new operating system, and allows users to access familiar apps and 3D titles through the Apple TV app. The Vision Pro headset will also offer optical inserts from Zeiss for vision correction at an additional cost. Readers will cost $99, while prescription lenses will cost $149.

[00:06:44] The lenses are available only online, a valid prescription is required, and not all prescriptions are supported. Why do we care? Apple positioned this announcement right during the Consumer Electronics Show, which it doesn't participate in. Start your timers for the first significant application developed for the Vision Pro. Thanks for listening. It's National Wonk Day, so thank those policymakers who think hard in Washington.

[00:07:15] Got a comment? Question? Thought on a story? Put it in the comments if you're on YouTube, and reach out on LinkedIn if you're listening to the podcast. Thanks again for listening. The Business of Tech is written and produced by me, Dave Sobel, under ethics guidelines, posted at businessof.tech. If you like the content, please make sure to hit that like button and follow or subscribe. It's free and easy and the best way to support the show and help us grow.

[00:07:43] You can also check out our Patreon, where you can join the Business of Tech community at patreon.com slash mspradio, or buy our Why Do We Care merch at businessof.tech. Finally, if you're interested in advertising on this show, visit mspradio.com slash engage. Once again, thanks for listening to me, and I will talk to you again on our next episode of the Business of Tech.

[00:08:13] Part of the MSP Radio Network.