A group of visual artists has scored a significant victory in their legal battle against AI image generators. A California federal judge has ruled that visual artists can continue to pursue some of their copyright claims against AI companies Stability AI, Midjourney, DeviantArt, and Runway AI.

The artists alleged that these companies used their copyrighted images to train their AI models without permission, violating their rights. While the judge dismissed some of the claims, he allowed others to proceed, including allegations that the companies illegally stored copyrighted works on their systems.

US District Judge William Orrick, presiding over the Northern District of California, stated that the artists have made a plausible argument that the companies’ actions could violate their rights.

“Plaintiffs rely on some of those works to plausibly demonstrate that their works were used as training images and that their works or elements of their works can be recreated through the AI products. The identification of those works may not prove liability under the Copyright Act, but they do provide support for the plausibility of plaintiffs’ Copyright Act theories,” the judge said in a 33-page ruling.

This case is part of a broader legal battle concerning the use of copyrighted material in AI training. The artists argued that AI companies have used datasets, such as those from LIAON, which include their copyrighted works without permission.

The judge admitted the artists’ submission.

“Plaintiffs have plausible allegations showing why they believe their works were included in the LAION datasets. And plaintiffs plausibly allege that the Midjourney product produces images – when their own names are used as prompts – that are similar to plaintiffs’ artistic works,” the ruling read.

The judge refused to dismiss related trademark claims but dismissed allegations of unjust enrichment, breach of contract, and violations of a separate US copyright law.

The bone of contention

The core issue in this case revolves around the AI systems’ use of copyrighted works to train their models, particularly the Stability AI’s Stable Diffusion model, which the artists claim contains “compressed copies” of their works. Judge Orrick noted in his ruling that while the full extent of copyright infringement is not yet clear, it is plausible that the Stable Diffusion model could lead to infringement through its use by end-users.

“If an AI company profits from content based on someone else’s copyrighted material, they are obligated to share those earnings, possibly in the form of royalties,” said Prashant Mali, an advocate who specializes in data protection and cyber laws.  “AI technology was never intended to be used freely at the expense of creators’ rights, and this ruling underscores the need for clear guidelines in balancing innovation with copyright protection.”

This ruling follows a previous decision by Judge Orrick, who had dismissed several of the artists’ original claims in October but allowed them to refile their case. The artists, including Sarah Andersen, Kelly McKernan, and Karla Ortiz, refiled their complaint in November with additional plaintiffs.

Judge Orrick’s recent decision now permits them to proceed with their copyright claims.

Mali pointed out that copyright protection exists for any original work created by an artist. “If an artist has provided explicit consent, such as a No Objection Certificate (NOC), their work can be legally used to train AI algorithms.”

“However,” Mali stated, “AI-generated content must be transparent regarding copyright usage, including providing appropriate disclaimers where necessary.”

While the judge dismissed some of their claims due to procedural issues, like the lack of registered copyrights for many of the works in question, the case will continue to move forward on the remaining counts.

This ongoing legal dispute highlights the complexities surrounding copyright issues in the era of AI-generated content. The outcome of this case could set a precedent for other similar lawsuits, such as those filed by Sarah Silverman against Meta and a class action against OpenAI, which are also being heard in the Northern District of California.

For now, the artists have secured a critical victory in their fight to protect their work from what they view as unauthorized use by AI technologies. As the case progresses, it could have far-reaching implications for how copyright law is applied to AI-generated content and the responsibilities of companies that develop and deploy such technologies.

Microsoft introduced Windows 365 three years ago, a service that lets the company cut partners out of the money-making loop by providing virtual PCs to customers.

Rather than provide only the operating system or the OS and bits of other software — notably productivity applications in the form of Office — Microsoft also serves up ersatz hardware, virtual machines running on its vast cloud of Azure servers.

Dubbed “desktop as a service” (DaaS, in keeping with other, similar acronyms) by some, Microsoft’s tagged its offering as “Cloud PC” as in “Windows 365 is your PC in the cloud.”

“Just like applications were brought to the cloud with SaaS, we are now bringing the operating system to the cloud, providing organizations with greater flexibility and a secure way to empower their workforce to be more productive and connected, regardless of location,” Microsoft CEO Satya Nadella said in a statement at launch.

Windows 365 hasn’t exactly done away with the need for high-powered laptops altogether — Microsoft and its OEM partners have recently focused their attentions on selling “AI PCs” — but the concept has shown some appeal with enterprise firms, and Microsoft cites FedEx, ING, and Carlsberg as customers. Together with Azure Virtual Desktop, Windows 365 contributed more than $1 billion in revenues last year, Nadella revealed in Microsoft’s 2023 annual investor report, with a third of Microsoft’s enterprise customers paying for “cloud-delivered Windows” as of mid-2023. 

What is Windows 365?

At its simplest, it’s a virtualization service that provides a Windows desktop and first- and third-party applications to users with both PC and non-PC hardware.

Maybe it’s better to think of it as a streaming service. Rather than stream movies and TV shows, it streams the output of a Windows 10- or Windows 11-powered PC. The controller is the keyboard, touchscreen, mouse, even the microphone of whatever device is in front of the user.

It’s also the latest incarnation of the thin computing model, which harks to the beginnings of digital computing when the computer was massive and cost multi-millions — and endpoints were unintelligent terminals. Like that model, Windows 365 runs the virtual desktop on servers at a distance; the data is transferred over the Internet rather than an organization’s network.

How much does Windows 365 cost?

Microsoft charges a flat monthly rate per user, rather than basing the cost on the amount of activity, as in the amount of Azure resources consumed. 

There are 13 virtual machine configurations available. The most basic (two cores, 4GB of memory and 64GB of storage) costs $28 per user per month; the most advanced (16 cores, 64GB RAM and 1TB of storage space) is $315 per user per month — that amounts to a hefty $3,780 each year. That’s for Windows 365 Enterprise;Windows 365 Business customers pay an additional $4 per user a month for each SKU, in comparison. There is, however, a discount when Windows 365 is accessed via a device that runs Windows 11 Pro or Windows 10 Pro,  however; it lowers the cost by $4 a user each month.

As of August 2024, access to Windows 365 machines with GPUs is now available. There are three configuration options, from Standard, for basic graphics workloads, to Max, for the most demanding uses. GPU pricing hasn’t been made public yet, however. 

So we continue to pay for, say, Microsoft 365 E3 licenses and for Windows 365? Isn’t that double billing?

No, though it’s easy to see it that way.

The Microsoft 365 license gives you the right to use the included software, among that collection Windows 10 (and down the road, Windows 11). The Windows 365 license pays for the virtual PC Microsoft’s built, and Microsoft maintains on its servers that run the operating system you paid for.

It’s no different than if you were on a physical PC. You paid for that with one invoice. You paid for the Microsoft 365 license with another.

What are the different versions of Windows 365?

There are two main Windows 365 products. Windows 365 Business, which caters to smaller organizations with up to 300 employees, is more straightforward to get up and running. Then there’s Windows 365 Enterprise, which has no limit on users. The enterprise version provides more control over the management of virtual PCs via Microsoft’s Intune, as well as integration with Microsoft apps such as Entra ID (formerly Active Directory) and Defender for Endpoint. 

There are also two variations of Windows 365 Enterprise. Aimed at organizations with shift and part-time workers, Windows 365 Frontline edition lets up to three employees share a single Windows 365 license, provided they access the virtual PC at different times. Windows 365 Government is designed for US government agencies and contractors that have access to Government Community Cloud (GCC) or GCC High, with stricter security and compliance standards than Windows 365 Enterprise.

What are the requirements for running Windows 365?

This is simple for Windows 365 Business, as all necessary licenses are included in purchases. 

Access to Windows 365 Enterprise and Frontline requires licenses for Windows 11 Enterprise or Windows 10 Enterprise, Intune, and Entra ID P1. 

What powers Windows 365?

Not surprising, Microsoft’s own cloud computing platform, Azure. Specifically, Azure Virtual Desktop, which can be used to create Windows virtual machines on Redmond’s servers, then stream those desktops to PC and non-PC devices.

Windows 365, Microsoft repeatedly said, is essentially an automated Azure Virtual Desktop, in that the former handled all the scut work of creating and assigning the virtual machines, slapped an analytics package on the result and offered a one-stop dashboard for admins.

Elsewhere, Microsoft characterized Windows 365 as a way for admins unfamiliar with virtualization (or hesitant to learn) to call on Azure Virtual Desktop.

On what devices can we run these virtual desktops?

At a basic level, the only requirement is access to an HTML 5 browser, says Microsoft, so most modern devices with online access will do the job, whether they run Windows 11 or 10, macOS, Linux, iOS, or Android.

Those that want to log in to a Windows 365 PC via the Microsoft Remote Desktop app for Windows will need to meet modest hardware requirements such as 1GB RAM (see a full list here), while access to Microsoft Teams requires a device with at least 4GB RAM.

What kind of internet connection do we need for Windows 365?

This will mostly depend on the demands of a particular app or workload. But, roughly speaking, if you’re able to stream video, you should be good to go . 

Microsoft recommends a minimum bandwidth of 1.5Mbps for light workloads, and up to 15Mbps for more demanding scenarios such as video conferencing or 4K video streaming.

Monitor display resolution plays a role, too — a screen resolution of 3840 x 2160 pixels would require a faster 15Mbps connection, for example.

What’s new in Windows 365?

As you’d expect, Microsoft has added numerous features to Windows 365 since its launch in 2021, with several aimed at improving the user experience. Windows 365 Boot option — available since September 2023 — is one such addition, enabling users to connect directly to their Windows 365 PC from the login screen of their device.  

More recently, Microsoft has begun to offer GPUs as an option for workers that require high levels of parallel processing power, such as 3D modelling and video rendering. Microsoft also recently announced the availability of “AI recommendations” that provide admins with suggestions around Windows 365 configurations based on analysis of employee usage data. 

Microsoft is urging all users of Office and Microsoft 365 to update the software as soon as possible, because hackers have started exploiting a serious vulnerability to access sensitive information on computers.

To be fully protected against the vulnerability, designated CVE-2024-38200, users need to install a security fix that will be released to the public on Aug. 13, this month’s Patch Tuesday, according to The Hacker News.

Tuesday’s security fixes will also close other publicized vulnerabilities, including CVE-2024-38202 and CVE-2024-21302, that could be used by hackers to downgrade Windows to an earlier version.

July was another rough month for the tech sector, with a worse-than-expected jobs report from the US Bureau of Labor Statistics (BLS) and analysis of that data by industry experts.

And on top of that, uncertainties around tech talent remain, according to a recent pulse survey from consultancy and professional services firm Ernst & Young (EY) — uncertainty exacerbated by the arrival of artificial intelligence (AI) tools and platforms. Half of IT leaders expect AI adoption to contribute to a roiling mix of hirings and firings into the fall, the survey found.

Even with hiring plans in place, 61% of tech leaders surveyed say the rapidly evolving technology has made it more challenging for them to source top talent. “One thing is certain: Companies are reshaping their workforce to be more AI savvy,” EY’s report said.

Ken Englund, who leads EY’s Americas Technology Growth sector, said companies are now concerned with how they should restructure teams to meet new demands, a restructuring that could mean the end of the most unique hiring market he’s seen in a decade. Englund spends most of his time evaluating and advising up-and-coming companies in IPO and pre-IPO stages — and there are a lot of such firms in the AI, software, and semiconductors space, he said.

Computerworld spoke with Englund about how AI is affecting hiring, re-shaping enterprise restructuring and what employees need to do to stay relevant as the marketplace undergoes dramatic shifts.

Why do you believe the July jobs report was worse than expected, especially for tech? “Two things. When we talk about tech jobs, we’re talking about jobs in tech companies or technical jobs anywhere. I think about jobs in tech, for the most part. I think the other thing to keep in mind, depending on where you look, net, we’re down about 10,000 jobs [in July]. In the scheme of the whole population of tech — there are several millions jobs in tech — what we’re seeing is still very strong demand in technical roles — developers, cyber, data scientists, and lighter roles in service and support and marketing.

“In any given month, we continue to see the workforce in evolution, given AI as a driver of upskilling and reskilling of the employee base. In some months, the net is positive and in some months, negative. I look a lot at layoffs.fyi, that’s sort of the data point I look at out in the market, and the trend line is getting smaller. Aside for a few major restructuring layoffs in tech over the past couple of weeks, the outflow seems to be getting smaller in magnitude.

“We’re seeing a lot of new companies, a lot of new start-ups, angel seed, round A [firms]. So those aren’t hiring that many people, but new company formation is growing.”

What kinds of start-ups are dominating? “AI and analytics, software and cloud — definitely on the digital side of tech versus hardcore infrastructure. In the bigger picture, as we’ve looked at tech over the years, things move from hardware to software over time. We’re still going to need a lot of hard-core infrastructure, semi-conductors, hardware, but more and more will move toward the software and apps layer.”

Can you explain the mixture of hirings and firings that we’ve seen over the past few years? “I actually think we’re starting to get close to a balance. If you go back over the past 24 months, you saw a lot of layoffs that were right sizing. Over-hiring around the time of the pandemic was driving no-regret-hiring as the overall tech sector was moving up, and ZIRP (zero interest rate policy) was allowing the tech companies to hoard talent. Now, there’s much more scrutiny around job requisitions and tying them to specific business needs, goals. Are they really needed? Are they directly aligned to some business initiative or value?

“I think we’re just getting back to what was considered normal behavior before the pandemic.”

How has AI changed the state of tech jobs? “…In general, there’s a very positive view of AI in tech. In a lot of other industries, there’s some uncertainty, some trepidation, some curiosity. But part of our pulse survey said about three out of four tech workers are using AI on a daily basis. So, the adoption in this portfolio of companies is higher than most, and I’d also said most employers and workers have a very good idea that AI is going to improve their business and their work.

“Really, we’re seeing its use mainly in development, software, testing, quality, customer care service as initial use cases. So, it’s slowly getting woven into everyone’s work.”

How are organizations restructuring their employee teams? “Everyone varies a bit. Probably two-thirds of these companies have some sort of reskilling or upskilling program. So, this isn’t about out with the old and in with the new. We did talk about rebalancing the workforce, but a lot of this will be employees being upskilled or retrained. That’s the most critical item going forward.

“I view AI skills as adjacent, additive skills for most people — aside from really hardcore data scientists and AI engineers. This is how most people will work in the new world. Generally, it depends. Some organizations have built whole, distinct AI organizations. Others have built embedded AI domains in all of their job functions. It really depends. There’s a lot of discussion around whether companies should have a chief AI officer. I’m not sure that’s necessary. I think a lot of those functions are already in place. You do need someone in your organization who has a holistic view of the positive sides of this and the risks associated with this.”

Why do you think this has been one of the most unique hiring periods over the past decade or so, and how has AI affected that lately? “I do fundamentally think we’ve had a platform shift. We had this around mobile. We had this around e-commerce. Or, if you go back far enough, we had this shift from mainframes to client-servers. So, I do believe this [AI] is a fundamentally a platform shift.

“From that perspective, the most critical thing when I sit down with clients, I always ask them, ‘How’s your data doing?’ We all know nobody has perfect data. In the AI world, data is going to become even more important. If it was difficult to manage your data before — think about graph databases and vector databases — really we see a lot of investment by enterprises into getting their data right for AI; that translates into ensuring you have the right resources: data architects, analysts, AI engineers and all those sort of positions as driving it.”

A lot of organizations are relying on cloud-based AI services from the likes of Microsoft, Oracle, Amazon and Google. Are you seeing an increase in the use of proprietary small language models based on open source versus these large language models (LLMs) offered through SaaS-style services? “I think it’s both. I think it’s still very early days. I think most enterprises are continuing to work with large language models and I think that will be the trend over the near horizon. Most of the key cloud providers, even frontier [companies], are building small models, too. I believe over time, you will see specialization, verticalization and small models being of distinct value.

“On the small language model size, I think — go back 24 months where large language models were — that’s where small models are now. But for early adoption among enterprises today, most are using large language models and doing RAG work. Not a lot of them are building their own proprietary models. But I do think it’s realistic to believe that most enterprises will have some level of proprietary models build out in the future.

“The thing I always hear is AI is not going to take your job. Somebody using AI will take your job.”

-Ken Englund

“We think about cloud around workloads. I think in the future, when thinking about what models you’ll use — small, large, proprietary, open source — it will be all around use cases. Most of our clients are starting on a single foundational model, but we always tell them to architect in some flexibility, because we think it’ll be models of models in the future.”

What does “models of models” mean? “At the end of the day, to get an answer to an answer to a particular set of use cases, you may need more than one big foundational model; it may be an open-source model…. [or] you’ll go to different models for different needs in the enterprise. I definitely don’t think it’s a one-size fits all. Build for flexibility.

“Most of these big frontier models really have commercial models around APIs. This idea of being fit for purpose for the kind of information or response you need for an inference will be the case going forward. You can think of yourself as … being a smart router for how you direct your AI inferences.”

How can IT professionals ensure they’re not left behind as organizations modernize? “Just start trying these tools, even if it’s in your personal direct to consumer life. People who have some familiarity with these tools off the bat will have a leg up. I think most companies will have a set of certifications, training and upskilling programs in their organizations. A lot of them already have ‘AI 101’ courses. I think as a tech worker, it’s up to you to take advantage of all those resources your company is offering you as a starting point, let alone all the other things out there in the open-source world.

“The thing I always hear is AI is not going to take your job. Somebody using AI will take your job.”

How are organizations upskilling or reskilling, and does that apply differently depending on the worker’s job, like line-of-business vs technologist? “When I think of functions that will use AI, whether that’s marketing or finance or customer care support or product development, they’re very different situations. I think the first three of those, they will have a much more applied use of this.

“My thought is if we fast-forward three or four years from now, we won’t even talk about AI. It will be embedded into the marketing automation software and ERP platforms out there. We’re going to get to a point where it just is.

“I think that’s the case for those line-of-business folks. For me, what’s important for business users, as these models get better — and I like to tell folks today is the worst AI will ever be, everyday it gets a little better — these models, whether they overlay deterministic models on top of probabilistic models, outcome and solution quality will get better. But understanding how that works in the meantime: there will be some judgment for what business workers need. What we’ll see mainly there are assistants — copilots that’ll make recommendations for business users like a marketer, but not probably human-out-of-the-loop at this point.

“Where we’re really seeing much more hard-core work is around software development, testing, quality and those areas where you’re in the nitty-gritty of activities. Those learnings will come out of classical enterprise IT or engineering product teams will flow into other parts of the enterprise.”

But AI is going to affect all of that, correct? “Absolutely. If you think about the most structured language in the world, it’s coding. If you think about these large language models, genAI, they’re really language based. AI’s ability to determine how code gets built, tested, released — that’s ground zero for all this stuff.”

So, more than anywhere else in the IT space, AI is being used to produce software, test it and deploy it, correct? “I think on the corporate IT side, yes. If you think about the rest of corporate IT functions, probably the two areas we’re seeing the most internet in AI is for customer service and customer care. These chatbots have been reasonably good initial products. We think anything that can handle customer care and service requests, we’re seeing move super well.

“Then, [it’s] things around employee workforce experience. So think about how you onboard a new employee — a lot of things that are rules-based and document based are the leading functions for AI.

“I think the last thing I’d add from our pulse poll is that these same concerns around AI adoption are the same ones we continue to see: cybersecurity, privacy, intellectual property are the biggies. The top line in our survey is really around skill-based development around AI expertise. This whole idea around certifications and upskilling is a really critical item.

“I know we’re all focused on the technology part of it, but this will continue — as always — to come down to the people and whether they can use it. That has never changed.”

Apple has once again tweaked its terms of business for developers as it continues to seek alignment with Europe’s Digital Markets Act (DMA) while looking to protect its business. 

The latest changes followed accusations from the European Commission that the conditions Apple had made so far to meet the DMA did not go far enough. Regulators felt the terms prevented developers from freely guiding customers to alternative ways to pay and were threatening very costly legal action for non-compliance with the law. In hopes of avoiding a large fine, Apple has now completely relaxed those rules, while introducing a new fee structure. 

As usual, the changes still won’t satisfy the company’s fiercest critics. But at this stage of the game, it appears very little will — though for the vast majority of developers Apple’s EU offer is better than before.

What changes has Apple made?

The primary change involves relaxed restrictions on how apps in the EU can link out to external sites. While some of the changes are relatively complex to easily summarize, the tweaks give developers a lot more flexibility as to where and how to promote external offers, including via competing app stores.

Apple is permitting developer links to open inside the app, rather than in a web browser. The company has also changed the way it charges fees for the service. Among the tweaks:

• First, it is introducing an Initial Acquisition Fee (5%), which must be paid for the first 12 months subsequent to a new customer being won on Apple’s platforms. This reflects the value of Apple’s platform as a way to find new customers and ends after 12 months.
• An additional 10% Store Services fee is charged for all sales of digital goods and services across 12 months following any app install, update, or reinstall, though the vast majority of developers will pay just 5%. The way this fee is structured means Apple will continue to collect it in future.
• Apple also takes a €0.50 Core Technology Fee for apps distributed via the App Store, Web distribution or alternative app marketplaces. This fee is paid for each first annual install over 1 million first annual installs in the year, and reflects a contribution to maintaining the company’s platforms.
• Users can opt-out of reading the disclosure sheet Apple provides to warn people when they are about to make purchases outside the protection of the Apple platform.
• Apple revised its fee calculator to help developers understand the consequences of the new fee structure.
• All the changes are described in full in Apple’s revised guidance on apps distributed in the European Union.

The guidance also notes that developers can communicate and promote offers for purchases at a destination of their choice (not just their own website) and can design those in-app promotions as they wish. This gives developers a lot more flexibility as to where and how to promote external offers and where those offers are made available.

There are plenty of nuances to the guidance that might apply to you or your business, but the basic outcome is most developers will be paying less and developers of free apps will continue to pay nothing at all. Fee-based apps with fewer than 1 million downloads (which is most of them) will pay just 5% Store Services Fee, or 7% for developers remaining in the App Store ecosystem.

How much is fair?

For all the complexity, it seems reasonable to believe Apple’s problems with regulators will inevitably coalesce around the question of how much is appropriate to charge for access to its ecosystem. It’s not as if globally accepted and used computing platforms create themselves; they are the sum of decades of work, investment, and effort that requires reward. Otherwise, why bother trying? 

Apple’s biggest critic, Epic CEO Tim Sweeney, doesn’t see it that way, arguing that Apple’s top rate 15% fee is an “illegal junk fee.” But it is difficult within that argument to discern any recognition for the value provided by Apple’s platforms. It can’t be that Sweeney doesn’t understand this intrinsic value. After all, Epic charges application developers using Unreal Engine 5% of revenue after the first $1 million. Is that a “junk fee?”

Logically therefore, it makes sense that those who profit from the existence of the platforms should compensate platform providers for the tools they use to build on them. You cannot warm yourselves beside the fire if you don’t go out and seek some fuel for those flames from time to time. 

While critics seem to think Apple (and by inference, every Apple customer) should bear all the costs of maintaining the platforms, that seems unreasonable. A competitive marketplace cannot and should not demand one entity stokes the fire, while everyone else casts happy shadows in the smoke. It requires at least some shared reward, and shared risk.

Where is the value?

With this new fee system, Apple has taken fresh steps toward defining the value of its business, by which I mean, addressing what it brings in terms of customer introductions, platform creation and development, and tools and support to developers. All three of these are uniquely provided by Apple and have inherent value. The only stumbling block is now and always has been, how much should that value be?

Apple meanwhile continues to work with EU regulators. The company has been in talks with them for years over these matters and will continue to engage as it works toward building a viable business proposition that works for Apple, EU, developers who value its platforms, and Apple’s European customers. 

We must now wait and see whether Europe feels Apple’s new changes meet their expectations of its behavior under the DMA.

More from Jonny Evans

• Why health might be Apple’s AI profit center
• Macs are becoming more locked down
• Here’s why Apple will become the world’s leading AI vendor

Please follow me on Mastodon, or join me in the AppleHolic’s bar & grill and Apple Discussions groups on MeWe.

AnitaB.org has announced new measures it’s taking to avoid a repeat of the debacle at last year’s Grace Hopper Celebration (GHC).

The nonprofit organization’s annual event to support the advancement of women and nonbinary technologists, was named for computing pioneer Rear Admiral Grace Hopper. It combines conference sessions with an expo and job fair.

At GHC 2023, the job fair was invaded by large numbers of men, some of whom had lied about their gender identity when registering, and who monopolized recruiters from large tech employers, butting into line and preventing the conference’s target attendees from getting interview slots. Attendees reported being physically pushed, demeaned, and sexually harassed by some of the men.

In a LinkedIn post after the conference, AnitaB.org pledged to address the problem. It said, “We are dedicated to bringing structural changes to ensure that GHC continues to be an uplifting experience and provides opportunities for women and non-binary technologists.”

A tale of two events

Bo Young Lee, president of AnitaB.org advisory, said this week in an email interview, “GHC 23 was a tale of two events. Those conference attendees who largely participated through attendance at sessions and talks had the same joyful, celebratory, and community-based experience that GHC has come to be known for.

“The most problematic behavior we witnessed was concentrated in our Expo Hall. It was there that we had a minority of attendees, mostly students and male, engage in aggressive behavior that violated our code of conduct.”

Lee cited three factors for this that the organization’s subsequent investigation revealed:

• A scarcity mindset brought on by reduced recruiting at universities and colleges that, Lee said, resulted in a larger number of job seekers than in previous years and “resulted in more aggressive behavior than we’ve seen in the past.”
• A larger number of male job-seeking attendees than in years past. “These male attendees were not at GHC to participate in any of the content sessions, and instead stayed fixed in the Expo Hall,” Lee said.
• Coordinated efforts: An investigation conducted after GHC 23 revealed that there was a coordinated effort by far-right anti-DEI groups “to undermine and disrupt GHC, both in person and online.”

Actions for GHC 2024

“Our commitment to inclusivity remains strong, focusing on engaging members, participants, and attendees who support the advancement of women, nonbinary technologists, and the LGBTQIA+ community,” AnitaB.org said in a recent email to members. “Our goal is to ensure that everyone involved in our celebration feels safe and valued.”

The email outlined a list of process changes for GHC 24, which will be held October 8 – 11 both virtually and in person in Philadelphia, Pennsylvania, that the organization believes will prevent the recurrence of last year’s issues.

First, it is modifying its registration procedure to require valid ID, such as a driver’s license, when registering. It will also require proof of student status if appropriate.

But, Lee said, “GHC has always been open to women, nonbinary, and ally technologists. We will never discriminate against who can buy a registration and participate.”

At the event, there will be stricter badge checks and ID verification for entry to the venue, as well as when entering the expo. In addition, attendees will be assigned to timed expo entry groups to allow everyone to experience the expo without having to fight crowds.

Finally, an update to the code of conduct holds everyone accountable for behavior that aligns with the organization’s mission. Attendees must agree to abide by it when registering.

Lee said there will also be enhanced cybersecurity monitoring to detect any coordinated efforts early, so they can be dealt with, and onsite security personnel to handle problems that might arise at the venue. These measures were created in consultation with external security consultants, local law enforcement, and cybersecurity consultants.

Why events like GHC are needed

The events at GHC 23 underscore the need for industry events aimed at underrepresented communities as a means to build and develop diverse talent, said Erin Pierre, principal analyst at Gartner.

“Our research has shown that women make up nearly half of the global workforce, and they only represent about 26% of IT employees. I’m not sure what the numbers are for nonbinary talent, but the numbers show us that more than half — a majority, at least — of IT employees are predominantly male,” she said. “So these types of events, where women and nonbinary talent can come together and learn and develop their skill sets and get some networking opportunities or even potential interviewing opportunities, are incredibly important.”

A spokesperson for QueerTech, an organization that focuses on breaking down barriers, creating spaces, and connecting communities to support and empower 2SLGBTQ+ people to thrive, agreed.

“At QueerTech we recognize that many industries — including the tech industry — have been shaped by and for cisgender men, resulting in a system that largely overlooks and excludes diverse communities. This systemic bias has created significant barriers for underrepresented communities, including members of the 2SLGBTQIA+ community, ranging from discrimination and a stark lack of representation, to limited access to mentorship and professional networks,” they said in a statement. 

“Equity is not about treating everyone the same; it’s the recognition that existing barriers require varying levels and types of support in order to ensure fair and equal access to opportunities,” the QueerTech spokesperson added.

Creating safe event and career-building environments is crucial to empowering underrepresented communities, they said. “In order to create safe, equitable environments, we must always remember who it is we aim to serve, thoroughly understand their lived experiences and barriers to success, and work tirelessly to ensure these values, and understandings, are reflected in every single programming decision.”

It is all the more jarring for participants when a supposedly safe environment turns out not to be, as happened at GHC 23.

Said Pierre, “When something like this happens, it is usually a symptom of a larger issue. So even if we could wave our magic wand and magically change this, and they could change the celebration for this year to be a little more safe and inclusive, we still have a larger issue at play here. And that’s why it feels so catastrophic when it happens, because really what this shows us is that there’s still a severe lack of resources and opportunities for female and nonbinary talent.”

Organizations need to do a better job of attracting and retaining a diverse workforce, Pierre added. We need to look at diversity, equity, and inclusion (DEI) as something that benefits everyone, not just  female and nonbinary talent, she noted, since many of the things that make an employer attractive for underrepresented groups, including flexibility, work-life balance, and development opportunities, are good for all employees.

“I think we need to have more of an actionable approach and making sure that we’re really embedding DEI into our overall culture,” she said.

Like me, analysts expect Apple to eventually charge for access to some Apple Intelligence features, which is why I think the biggest opportunity for the company involves AI-augmented fitness and healthcare.

First, the thesis: Analysts Neil Shah (Counterpoint Research) and Ben Wood (CCS Insight) both told CNBC they believe Apple will eventually attempt to monetize Apple Intelligence, potentially as part of its Apple One bundle of services.

Shah points out that AI has the potential to become more personalized to users over time. This is particularly true for Apple, which has been designing Apple Intelligence from day one to prioritize privacy in all it does. The company has even begun making servers to drive some of its AI services, and as the services those servers provide become more sophisticated, it makes sense to charge for the more advanced tools it offers.

Privacy + AI = ?

The winning combination of Apple Intelligence along with the personal data gathered by the company’s devices and the continued research that makes it possible for AI to work with information without ever actually seeing the inherent data can’t be ignored. It’s that combination (along with the health sensors inside Apple Watch) that make AI-augmented digital health services ripe for monetization.

Think about the services Apple already offers that relate to health. Fitness+ might be the fee-based service, but it is supported by the Health app, years of in-depth research into wearable devices and health, and amazing technological manifestations to protect heart health, women’s health — even fall detection, crash detection, and Emergency SOS via Satellite. Many of these features already rely on various forms of machine intelligence, but generativeAI (genAI) can be far more creative in using the data points Apple’s devices collect — privately and only visible to you.

If Apple does choose to monetize these AI products, it won’t have had to look terribly far for inspiration. Palantir founder Peter Thiel sat on the same tech advice panel as Apple CEO Tim Cook, and the former company is working extensively in AI in healthcare, particularly (and contentiously) with the UK’s National Health Service (NHS).

The NHS is also working closely to develop ways to use AI to support its services. “We’re already seeing great applications of AI technology, and more work is under way to fully harness its benefits and use AI safely and ethically at scale,” the NHS said.

But while Palantir seems to need plenty of NHS data to run its operations, Apple’s approach seems to require less of that because Apple strives to ensure it sees as little as possible. 

Would you pay for AI-augmented digital health?

Now, I don’t want to get into a relative conversation about the differences between Palantir and Apple’s approach to data privacy. I’m not a Palantir expert. But what does seem clear is that Apple’s billions of users might well be willing to pay for smart digital health services, and Apple Intelligence could deliver this.

Based on Apple’s direction of travel in digital health so far, these services would almost certainly focus on preventative health intelligence rather than actual remedies.

The intelligence would exist to provide early warning of symptoms, and to recommend what actions customers could take to handle a health crisis/intervention. Another manifestation (perhaps in conjunction with Apple’s satellite communications services) could be an international medical and emergency system for travellers.

Apple doesn’t need to base its AI monetization on health, of course. It has plenty of other strings for its bow. But one thing is certain — no company will spend tens of billions building cutting-edge services without attempting to recoup that investment down the line. With that in mind, I can’t believe Apple, which famously charges “reassuringly high” amounts even for things like spare iPhone charging leads, isn’t going to find some way to make money from Apple Intelligence.

Reading the room

Wherever it chooses to make that money back, I think it will look to where its core values around privacy, security, and intentionality in AI make the most difference. The company’s seemingly deliberate approach to introducing new features meets the public mood of a population that is becoming increasingly mistrustful of tech firms.

That means the solutions it has introduced so far provide value to most users while also protecting their privacy. Those services are almost certainly the thin end of a multi-billion dollar AI wedge, and Apple, more than most, understands the need to provide solutions that don’t scare customers. (Apple Pay on iPhone was also one of these, back in the day. Now it is the most dominant mobile payment system because Apple introduced it gently.)

This calm, measured intentionality will become increasingly visible over the coming two to three years as Apple puts the pieces in place to deliver fee-based Apple Intelligence services. 

Fees aren’t the only way in which Apple will hope to profit from AI. 

There are other options

Part of its payola should come in accelerating hardware sales as it ships the first end-to-end mobile-to-Mac ecosystem with AI inside and processors to power it up. 

Apple will also be hopeful that its developer communities identify amazing new ways to deploy the large language models (LLMs) it has created within their apps. That’s great for developers, of course, but also gives Apple the competitive edge it needs to maintain hardware sales while also grabbing its slice of App Store revenue action.

But crucial to all of this is that Apple’s AI should be seen as a service, which is what Apple is going to try to build on once the initial Apple Intelligence release is done. That’s going to merit the introduction of additional premium AI tools and services aimed at what people are already doing with their devices.

These could include more advanced email analysis tools to help productivity professionals stay on top of increasingly demanding in-boxes. They might extend to premium automated Keynote presentation design and creation tools.

But the most likely space in which Apple will explore the opportunity for more sophisticated AI inside its devices will be around personal health, at the intersection of science, privacy, and digital arts.

Please follow me on Mastodon, or join me in the AppleHolic’s bar & grill and Apple Discussions groups on MeWe.

ChatGPT-maker OpenAI has released a new version of its GPT-4o large language model (LLM), designed to simplify the process of generating “well-defined” and “structured” outputs from AI models.

“This feature is particularly valuable for developers who need to validate and format AI outputs into structures like JSON schemas. Developers often face challenges validating and formatting AI outputs into well-defined structures like JSON schemas,” the company wrote in a blog post, adding that an early release version of the LLM has been made available on Microsoft’s Azure OpenAI Service.

The structured outputs generated by the new version of the LLM, christened GPT-4o-2024-08-06, are a result of the LLM allowing developers to specify the desired output directly from the AI models.  

JSON schema, according to OpenAI and Microsoft, is used by developers to maintain consistency across platforms, drive model-driven UI constraints, and automatically generate user interfaces.

“They are also essential for defining the structure and constraints of JSON documents, ensuring they follow specific formats with mandatory properties and value types. It enhances data understandability through semantic annotation and serves as a domain-specific language for optimized application requirements,” the companies explained.

Additionally, the companies said that these schemas also support automated testing, schema inferencing, and machine-readable web profiles, improving data interoperability.

The new LLM supports two kinds of structured outputs — user-defined JSON schema and strict mode or more accurate tool output.

While the user-defined outputs allow developers to specify the exact JSON Schema they want the AI to follow, the strict mode output, which is a limited version, lets developers define specific function signatures for tool use, the companies said.

The user-defined output is supported by GPT-4o-2024-08-06 and GPT-4o-mini-2024-07-18. Alternatively, the limited strict mode is supported by all models that support function calling, including GPT-3.5 Turbo, GPT-4, GPT-4 Turbo, and GPT-4o models.

GPT-4o was first announced in May 2024, as OpenAI’s new multimodal model, followed by GPT-4o mini in July 2024.

Microsoft has yet to make the pricing of the new model available on its pricing portal.

Critical time for the company and competition

The new model comes at a time when the company is undergoing critical changes in its leadership and is facing stiff competition from rivals such as Anthropic, Meta, Mistral, IBM, Google, and AWS.

This week, OpenAI co-founder John Schulman joined the list of departing OpenAI executives making the move to Anthropic — a rival LLM maker and provider that was founded by a group of executives leaving OpenAI.

Schulman follows Jan Leike, who made the move from OpenAI to Anthropic back in May.

Ilys Sutskever also left a senior role with OpenAI, but he is launching his own AI effort. 

Another troubling indicator at OpenAI was the transfer of safety executive Aleksander Madry, who was mysteriously reassigned. This also follows OpenAI co-founder Andrej Karpathy’s departure back in February. 

At the same time, OpenAI is also facing stiff competition for its LLMs. One such example is Meta’s newly unveiled Llama 3.1 family of large language models (LLMs), which includes a 405 billion parameter model as well as 70 billion parameter and 8 billion parameter variants.

Analysts and experts say that the openness and accuracy of the Llama 3.1 family of models pose an existential threat to providers of closed proprietary LLMs.

Meta in a blog post said that the larger 405B Llama 3.1 model outperformed models such as Nemotron-4 340B Instruct, GPT-4, and Claude 3.5 Sonnet in benchmark tests such as MMLU, MATH, GSM8K, and ARC Challenge.

Its performance was close to GPT-4o in these tests as well. For context, GPT-4o scored 88.7 in the MMLU benchmark and Llama 3.1 405B scored 88.6.

MMLU, MATH, GSM8K, and ARC Challenge are benchmarks that test LLMs in the areas of general intelligence, mathematics, and reasoning.

The smaller Llama 3.1 models of 8B and 70B, which have been updated with larger context windows and support for multiple languages, also performed better or close to proprietary LLMs in the same benchmark tests.

Another example is the release of Claude 3.5 Sonnet in June, which according to the LLM-maker, set new scores across industry benchmarks, such as graduate-level reasoning (GPQA), MMLU, and HumanEval — a test for coding proficiency.

While analysts and industry experts have been waiting for OpenAI to release GPT-5, company CEO Sam Altman’s post on X about summer gardens and strawberries has sparked speculations that the company is working on a next-generation AI model that can crawl the web to perform research.