Blog

The struggle for privacy

When Apple announced the delayed rollout, it was quite detailed about its concerns: “Specifically, we are concerned that the interoperability requirements of the DMA could force us to compromise the integrity of our products in ways that risk user privacy and data security,” it said.  “We are committed to collaborating with the European Commission in an attempt to find a solution that would enable us to deliver these features to our EU customers without compromising their safety.”

But Vestager’s arguments, and previous mutterings on the topic of user security and privacy, seem to suggest that the “pro competition” trading bloc that gave us GDPR (ironically wrecking the economics of small website publishers when it did), isn’t going to be terribly receptive to Apple’s arguments that the highly personal data gathered on someone’s device should be protected, minimized, and not simply made available to third party AI competitors without clear user consent, protection, and oversight.  

‘This is surveillance’

As Apple CEO, Tim Cook warned six years ago, the potential for AI-driven surveillance has never been greater; that really is what is at stake in Apple’s struggles with the European Commission. 

AI’s complexity adds to the confusion as do the numerous aspects of AI that warrant regulation. The list is lengthy, including job protection, consumer privacy, bias prevention and discrimination, deepfakes, disinformation, election fraud, intellectual property, copyright, housing, biometrics, healthcare, financial services, and national security risks.

So far, the federal government has dragged its feet on AI regulation, seemingly more focused on party politics and infighting than in crafting useful measures. As a result, Congress has not been an effective tool for structuring regulation policy.

The time for congressional action on AI regulation was two or three years ago. But with little being done federally, the states, particularly California, are attempting to fill the breach.

More than 10,000 restaurant locations use SoundHound’s platform to understand speech in a range of major languages, learn any restaurant’s menu, process orders directly to the point of sale (POS), and answer customer FAQs.

“Allset will help SoundHound bring voice AI solutions to even more restaurants looking to improve operational efficiency,” Keyvan Mohajer, CEO and co-founder of SoundHound AI, said in a statement on the deal.

For example, SoundHound has been working with White Castle, the US-based fast-food hamburger joint, to offer voice AI ordering technology at select White Castle drive-thrus for a year, with plans to roll out the technology to 100 locations by the end of 2024.

“Claude 3.5 Sonnet is now available for free on Claude.ai and the Claude iOS app, while Claude Pro and Team plan subscribers can access it with significantly higher rate limits,” an Anthropic announcement noted.  “It is also available via the Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI.”

More power, less cost

Anthropic claimed that Claude 3.5 Sonnet surpasses competitor models like GPT4.o and Gemini 1.5 Pro, Meta’s Llama 3 400B, and even its predecessor, Claude 3 Opus, on a wide range of evaluations. Notably, the Claude 3.5 Sonnet achieves this leap in performance while maintaining the speed and cost-effectiveness of their mid-tier model, the Claude 3 Sonnet.

“The Claude 3.5 Sonnet represents a significant advancement in large language models, featuring notable improvements across key metrics,” said Prabhu Ram, head of the Industry Intelligence Group at CyberMedia Research. “It boasts double the processing speed of its predecessor, Claude Opus, at a fraction of the cost.”

“ASPs for augmented reality (AR) headsets have almost always been above this price point, but ASPs for VR, MR, and ER headsets have typically been lower,” he said. “Apple’s Vision Pro drove ASPs higher for MR headsets, but the addition of lower-cost devices from Meta and HTC have kept those ASPs from going much higher. Meanwhile, there were many devices for VR and ER priced below $500.”

Return to growth

Looking ahead, Llamas said that IDC is anticipating ASP erosion across all products: “Because the overall market is still in its early stages with more expensive first- and second-generation devices, prices will be high even as early adopters buy them. In order to reach scale in the mass market, vendors will need to reduce prices on later and upcoming devices.”

IDC is forecasting that “headset shipments will return to growth later this year with volume growing 7.5% over 2023. Newer headsets and lower price points will help with the turnaround expected later this year. Beyond that, headset shipment volume is expected to see a compound annual growth rate (CAGR) of 43.9% from 2024–2028.”

Windows

This month, Microsoft released one critical update (CVE-2024-30080) and 32 patches rated as important for Windows, covering the following key components: 

• Windows Win32 Kernel Subsystem, GRFX and drivers
• Networking (Wii-fi) and DHCP
• Storage and Error Reporting
• Crypto and BitLocker

The critical-rated patch relates to the core, but not often used, Message Queuing service (MSMQ) that could affect internal applications. Unusually, this patch has already been updated since the main release on Tuesday. That said, the Readiness team believes all these Windows patches can be added to your standard release schedule.

Microsoft Office 

There were no critical updates for Office this month, and only five patches rated as important. All five have low potential for exploitability (no worms, add-in vulnerabilities or Word macro issues) and should be added to your regular Microsoft Office update schedule.

After all, if you squint, you can just about see M4 Macs appearing at the top of that nearby hill now the iPad Pro has that chip. We also think new iPhones are very likely to be already rolling off production lines, and both the iPad Air and iPad Pro already host compatible chips.  So, if you’re in the market for new hardware, it’s on the way, but tens of millions of people won’t need to upgrade to try Apple’s new genAI shiny.

Competitors are gasping

When it comes to Apple Silicon — the fire horse on which AI rides in the newly rebirthed Apple universe — take note that it has taken competitors, even those working with Arm reference designs, literally years to even begin to catch up with the computational power and energy efficiency Apple’s designers achieved. 

With a road map to M4 and beyond already in place, Apple is quite evidently telling us that when it comes to processors the speed argument is done, dusted, and won. And don’t get me started on privacy after the Microsoft Recall farce.

Summaries in messages and mail

Sometimes we have little time, and yet whoever we’re interacting with has a great deal of complex information to share; the result is a lengthy email. Sure, the whole message should probably be read, but if you’re short of time, you can use Summaries in Messages to get the gist of the entire diatribe. Yes, if you completely rely on summaries you’ll probably miss something, but if you are in a hurry and just need the basics, Apple Intelligence has your back.

Built-in transcription tools

Apple Intelligence lets you generate summaries and transcripts of audio recordings captured with the Notes app or during a phone call. This is going to be popular with a lot of people — particularly researchers, students, and journalists. (If you’re concerned about privacy, all parties in a phone call will be told this activity is taking place.)

The new audio transcription and summarization features in Notes enable a device to take notes for the user, Apple says. This lets them, “stay present in a situation where they need to capture details about what’s happening,” which means you can stay focused in that meeting and still have a useful and usable aide memoire.

DuckDuckGo has released an AI-powered portal to some of the most popular chabots and said it will not disclose or otherwise use what users type into the window to train up large language models — the basis for the generative AI (genA) tech.

“Chats are private, anonymized by us,” DuckDuckGo lead designer Nirzar Pangarkar wrote in a blog post. “Our mission is to show the world that protecting your privacy online can be easy.”

DuckDuckGo AI Chat currently allows users to access four popular AI chatbots: Open AI’s GPT 3.5 Turbo, Anthropic’s Claude 3 Haiku, Meta Llama 3, and Mistral’s Mixtral 8x7B) — the latter two, open-source models. The optional AI Chat feature is free to use within a daily limit, and can easily be switched off.

For usage insights, it’s possible to get information about customer data, audiences and customer journeys without running SQL queries. A user could ask the AI assistant, for example, which attributes in a customer profile are leading to a conversion, said Bhambri, the sort of task that might require contacting different colleagues and consulting multiple documents. 

Unlike other aspects of the launch, the ability to ask for operational insights is still in public beta, Adobe said. 

The AI assistant can also be used to create content. In addition to creating audience segments based on a customer’s own data, it’s possible to produce assets that can be included in an email marketing campaign, for instance. In this case, Adobe’s Firefly image model could be used to generate images and design layout within the relevant AEP app. The ability to quickly create content variations allows for greater experimentation, Bhambri said, and can help personalize communications to different audiences.