Agregátor RSS

Google v srpnu oznámí nové Pixely 9.. •.•...a také minimálně tři nové funkce umělé inteligence. • Samsung má Galaxy AI a tak bude mít Google svou Google AI...

Rats are incredibly nimble creatures. They can climb up curtains, jump down tall ledges, and scurry across complex terrain—say, your basement stacked with odd-shaped stuff—at mind-blowing speed.

Robots, in contrast, are anything but nimble. Despite recent advances in AI to guide their movements, robots remain stiff and clumsy, especially when navigating new environments.

To make robots more agile, why not control them with algorithms distilled from biological brains? Our movements are rooted in the physical world and based on experience—two components that let us easily explore different surroundings.

There’s one major obstacle. Despite decades of research, neuroscientists haven’t yet pinpointed how brain circuits control and coordinate movement. Most studies have correlated neural activity with measurable motor responses—say, a twitch of a hand or the speed of lifting a leg. In other words, we know brain activation patterns that can describe a movement. But which neural circuits cause those movements in the first place?

We may find the answer by trying to recreate them in digital form. As the famous physicist Richard Feynman once said, “What I cannot create, I do not understand.”

This month, Google DeepMind and Harvard University built a realistic virtual rat to home in on the neural circuits that control complex movement. The rat’s digital brain, composed of artificial neural networks, was trained on tens of hours of neural recordings from actual rats running around in an open arena.

Comparing activation patterns of the artificial brain to signals from living, breathing animals, the team found the digital brain could predict the neural activation patterns of real rats and produce the same behavior—for example, running or rearing up on hind legs.

The collaboration was “fantastic,” said study author Dr. Bence Ölveczky at Harvard in a press release. “DeepMind had developed a pipeline to train biomechanical agents to move around complex environments. We simply didn’t have the resources to run simulations like those, to train these networks.”

The virtual rat’s brain recapitulated two regions especially important for movement. Tweaking connections in those areas changed motor responses across a variety of behaviors, suggesting these neural signals are involved in walking, running, climbing, and other movements.

“Virtual animals trained to behave like their real counterparts could provide a platform for virtual neuroscience…that would otherwise be difficult or impossible to experimentally deduce,” the team wrote in their article.

A Dense Dataset

Artificial intelligence “lives” in the digital world. To power robots, it needs to understand the physical world.

One way to teach it about the world is to record neural signals from rodents and use the recordings to engineer algorithms that can control biomechanically realistic models replicating natural behaviors. The goal is to distill the brain’s computations into algorithms that can pilot robots and also give neuroscientists a deeper understanding of the brain’s workings.

So far, the strategy has been successfully used to decipher the brain’s computations for vision, smell, navigation, and recognizing faces, the authors explained in their paper. However, modeling movement has been a challenge. Individuals move differently, and noise from brain recordings can easily mess up the resulting AI’s precision.

This study tackled the challenges head on with a cornucopia of data.

The team first placed multiple rats into a six-camera arena to capture their movement—running around, rearing up, or spinning in circles. Rats can be lazy bums. To encourage them to move, the team dangled Cheerios across the arena.

As the rats explored the arena, the team recorded 607 hours of video and also neural activity with a 128-channel array of electrodes implanted in their brains.

They used this data to train an artificial neural network—a virtual rat’s “brain”—to control body movement. To do this, they first tracked how 23 joints moved in the videos and transferred them to a simulation of the rats’ skeletal movements. Our joints only bend in certain ways, and this step filters out what’s physically impossible (say, bending legs in the opposite direction).

The core of the virtual rat’s brain is a type of AI algorithm called an inverse dynamics model. Basically, it knows where “body” positions are in space at any given time and, from there, predicts the next movements leading to a goal—say, grab that coffee cup without dropping it.

Through trial-and-error, the AI eventually came close to matching the movements of its biological counterparts. Surprisingly, the virtual rat could also easily generalize motor skills to unfamiliar places and scenarios—in part by learning the forces needed to navigate the new environments.

The similarities allowed the team to compare real rats to their digital doppelgangers, when performing the same behavior.

In one test, the team analyzed activity in two brain regions known to guide motor skills. Compared to an older computational model used to decode brain networks, the AI could better simulate neural signals in the virtual rat across multiple physical tasks.

Because of this, the virtual rat offers a way to study movement digitally.

One long-standing question, for example, is how the brain and nerves command muscle movement depending on the task. Grabbing a cup of coffee in the morning, for example, requires a steady hand without any jerking action but enough strength to hold it steady.

The team tweaked the “neural connections” in the virtual rodent to see how changes in brain networks alter the final behavior—getting that cup of coffee. They found one network measure that could identify a behavior at any given time and guide it through.

Compared to lab studies, these insights “can only be directly accessed through simulation,” wrote the team.

The virtual rat bridges AI and neuroscience. The AI models here recreate the physicality and neural signals of living creatures, making them invaluable for probing brain functions. In this study, one aspect of the virtual rat’s motor skills relied on two brain regions—pinpointing them as potential regions key to guiding complex, adaptable movement.

A similar strategy could provide more insight into the computations underlying vision, sensation, or perhaps even higher cognitive functions such as reasoning. But the virtual rat brain isn’t a complete replication of a real one. It only captures snapshots of part of the brain. But it does let neuroscientists “zoom in” on their favorite brain region and test hypotheses quickly and easily compared to traditional lab experiments, which often take weeks to months.

On the robotics side, the method adds a physicality to AI.

“We’ve learned a huge amount from the challenge of building embodied agents: AI systems that not only have to think intelligently, but also have to translate that thinking into physical action in a complex environment,” said study author Dr. Matthew Botvinick at DeepMind in a press release. “It seemed plausible that taking this same approach in a neuroscience context might be useful for providing insights in both behavior and brain function.”

The team is next planning to test the virtual rat with more complex tasks, alongside its biological counterparts, to further peek inside the inner workings of the digital brain.

“From our experiments, we have a lot of ideas about how such tasks are solved,” said Ölveczky to The Harvard Gazette. “We want to start using the virtual rats to test these ideas and help advance our understanding of how real brains generate complex behavior.”

Image Credit: Google DeepMind

Spojené státy dodaly na Ukrajinu už více než 300 obrněnců Bradley a zhruba 70 jich doposud bylo vyřazeno • Bradley se pyšní velkou pohyblivostí, kombinovanou se slušnou palebnou silou • Zároveň se tam ale také setkávají se zbraněmi a riziky 21. století

Společnost Proton AG stojící za Proton Mailem a dalšími službami přidala do svého portfolia Docs in Proton Drive.

Příspěvek na blogu Codean Labs podrobně rozebírá bezpečnostní chybu CVE-2024-29510 v Ghostscriptu, tj. interpretu PostScriptu (PS) a Portable Document Formatu (PDF). V upstreamu byla chyba opravena v květnu ve verzi 10.03.1.

Tyto filmy a seriály jsou teď na českém Netflixu nejoblíbenější. Nerozlišujeme žánr, stáří ani hodnocení na filmových webech. Jde o souhrnnou oblíbenost za poslední týdny, kterou zjišťuje web FlixPatrol.

Ze Spojených států přišla zpráva, která určitě rozesmutní všechny fanoušky počítačové historie. Technologické a počítačové muzeum Living Computers: Museum + Labs, které 9. ledna 2006 založil spoluzakladatel Microsoftu Paul Allen, definitivně zavírá své brány. Podrobnosti přináší magazín ...

Recently the Office of the Director of National Intelligence (ODNI) unveiled a new strategy for open-source intelligence (OSINT) and referred to OSINT as the “INT of first resort”. Public and private sector organizations are realizing the value that the discipline can provide but are also finding that the exponential growth of digital data in recent years has overwhelmed many traditional OSINT

Recently the Office of the Director of National Intelligence (ODNI) unveiled a new strategy for open-source intelligence (OSINT) and referred to OSINT as the “INT of first resort”. Public and private sector organizations are realizing the value that the discipline can provide but are also finding that the exponential growth of digital data in recent years has overwhelmed many traditional OSINT The Hacker Newshttp://www.blogger.com/profile/[email protected]

Francouzská obdoba našeho Úřadu pro ochranu hospodářské soutěže prý podle Reuters vznese obvinění vůči Nvidii kvůli údajnému zneužívání dominantního postavení. Regulátor vyšetřuje možné porušení antimonopolních pravidel v oblasti AI akcelerátorů. Zpráva Reuters je bohužel příliš vágní a ani ...

Program pro nahrávání dění na obrazovce vokoscreenNG, nástupce již nevyvíjeného vokoscreenu, byl vydán ve verzi 4.2.0. Instalovat lze také z Flathubu.

NFC Forum načrtlo novou funkci NFC, zatím v podobě konceptu • Jedním přiložením by se vykonala řada akcí • Se zaplacením by se ověřil věk, přidala sleva nebo odeslala účtenka

“Can you go through all the old pitch decks and replace the word ‘crypto’ with ‘A.I.’?”

This caption, part of a New Yorker cartoon by Benjamin Schwartz, perfectly captures Silicon Valley’s new spirit of AI washing.

AI washing sounds like just another spin cycle, but it’s actually a complex and multifaceted phenomenon. And it’s important for everyone reading this column — technology leaders, marketers, product builders, users, and IT professionals of every stripe — to understand the exaggeration, warped emphases, and outright lying that we all encounter in not only marketing and sales, but also the stories we read based on industry claims.

Understanding AI washing

AI washing is a deceptive marketing practice that overemphasizes the role of artificial intelligence in the product or service being promoted. The phrase is based on “greenwashing,” coined by environmentalist Jay Westerveld in 1986, where consumer products are marketed as environmentally friendly regardless of environmental impact.

Products using old-school algorithms are labeled as “AI-powered,” taking advantage of the absence of a universally agreed upon definition for what AI is and what AI is not. Startups build apps that plug into a publicly available generative AI API and market it as an AI app. Big, bold AI projects that are supposed to showcase technology often rely on people working behind the scenes, because humans are the only way to make the ambitious AI solution work.

Let’s talk more about that last one.

AI: It’s made out of people

Retail giant Amazon rolled out 44 high-tech stores called Amazon Go and Amazon Fresh, which (starting in 2016) used the company’s “Just Walk Out” set of technologies. (I first told you about this initiative in 2017.)

Amazon’s vision: Stores where consumers could walk in, choose their items from shelves, then walk out without encountering a human behind a cash register. Sensors, including cameras, would feed into AI, which could figure out who bought what and charge accordingly — all without any checkout process. It felt like shoplifting, but legal.

The system was powered by advanced computer vision, which watched customers and what they picked up. Sensors in the shelves conveyed the weight of items removed, confirming the kind and number of items detected by the cameras. RFID tagged items also added information to the mix. Advanced machine learning algorithms processed the data from cameras and sensors to identify products and associate them with specific shoppers. Electronic entry and exit gates determined who was entering and leaving and when.

The algorithms were trained on millions of AI-generated images and videos to recognize products, human behavior, and human actions.

For seven years, Amazon has been eager to talk about these components of its Just Walk Out technologies. But the tech giant has been hesitant to discuss the 1,000 or so human beings hired to make it all actually more or less function — and admitted the existence of these employees only after press reports exposed them. Even then, Amazon has obscured the specific role these employees played, saying only that they didn’t review video.

Even with 1,000 employees monitoring and enabling 44 stores (checking three-quarters of orders, according to reports), the technology has been beset by problems, including delayed receipts, mismanaged orders, and high operational costs. 

This year, Amazon has been phasing out Just Walk Out technology from its main stores but still offers it as a service to other companies.

Another big example of humans behind the AI curtain is the world of self-driving cars.

Alphabet’s Waymo (the operation formerly known as Google’s self-driving car initiative) has a NASA-style command center where employees monitor cars through cameras and step in remotely when there’s a problem. (Here’s a fast-motion video I took recently of a ride through San Francisco in a Waymo car.)

General Motors’ Cruise subsidiary admits its self-driving taxis need human assistance on average every 4 to 5 miles, with each remote control session lasting an average of 3 seconds.

Other self-driving companies rely on remote human operators even more. In fact, a German company called Vay straight up uses human operators to drive the cars, but remotely. The company recently rolled out a valet parking service in Las Vegas. The car is remotely driven to you, and you drive it wherever you like. Upon reaching your destination, you just get out and a remote operator will park it for you.

Amazon’s stores and self-driving cars are just two available examples of a phenomenon that’s widespread.

Why AI washing happens

The high-level, high-paid technologists building AI systems believe in AI, and believe it can solve extremely complex problems. Which it can — theoretically. They tell their superiors it can be done. Those leaders tell their board it can be done. Company C-suites tell investors it can be done. And as a company, they tell the public it can be done.

There’s just one small problem: It can’t be done.

Most companies feel some sense of accountability for lofty claims, and so they hide the degree to which the product or service depends on humans behind the curtain making decisions, working through problems, and enabling the “magic” to take place.

The more shameless companies remain undeterred by proof that their AI isn’t quite as capable as they claimed or believed, so they just re-up their claims again and again. Tesla CEO Elon Musk comes to mind.

In October 2016, Musk said Tesla would demonstrate a fully autonomous drive from Los Angeles to New York by the end of 2017.

By April 2017, he predicted that in about two years, drivers would be able to sleep in their vehicle while it drove itself.

In 2018, Musk moved his promise of full Tesla self-driving to be by the end of 2019.

In February 2019, Musk promised full self-driving “this year.”

In 2020, Musk claimed that Tesla would have over 1 million self-driving robotaxis on the road by the end of the year.

Even this year, Musk claimed full self-driving Teslas might happen “later this year.”

It’s not going to happen. Musk is deluding himself and his customers. Musk is the Mr. Clean of AI washing.

The real problem with AI washing

The cumulative effect of AI washing is that it leads both the public and the technology industry astray. It fuels the delusion that AI can do things it cannot do. It makes people think AI is some kind of all-purpose solution to every problem — or a slippery slope into dystopia, depending on one’s worldview.

AI washing incentivizes inferior solutions, focusing on “magic” rather than quality. Claims that your dog-washing hose is “powered by AI” doesn’t mean you end up with a cleaner dog. It just means you have an overpriced hose.

AI washing warps funding. Silicon Valley investment nowadays is totally captured by both actual AI and AI-washing solutions. Even savvy investors may overlook AI-washing exaggeration and lies knowing that the AI story will sell in the marketplace thanks to buyer naiveté.

The biggest problem, however, is not delusional selling by the industry, but self-delusion. Purveyors of AI solutions believe that human help is a badge of shame, when in fact I think human involvement would be received with relief.

People actually want humans involved in their shopping and driving experience.

What we need is more human and less machine. As we speak, AI-generated garbage is flooding the zone with cringy prose and falsehoods, along with weird, sometimes horrifying, images. Google is so eager to replace its search engine with an answer engine that we end up with glue on our pizza.

What the public really wants is a search engine that will point us to human-created content or, at least, a PageRank system that favors the human and labels the AI-generated.

The AI-washing phenomenon is built on delusion. It’s built on the delusion that people want machines creating and controlling everything, which they don’t. It’s based on the delusion that adding AI to something automatically improves it, which it doesn’t. And it’s based on the delusion that employing people represents a failure of technology, which it doesn’t.

Enough delusional AI washing already! Sellers need to tell the truth about AI. And buyers need to demand proof that any AI in the products and services we pay for actually does something useful.

I think I speak for all of us in the technology industry, the technology customer community, and the tech press when I say to Silicon Valley: Stop gaslighting everybody about AI.

Proton has unveiled an end-to-end encrypted document editor that it said will provide an alternative to Microsoft Word and Google Docs for privacy-conscious users.

Docs in Proton Drive, announced on Wednesday by the Swiss software vendor that’s best known for its encrypted email app, contains many of the document creation features that office workers might expect.

Users can create and edit documents, share with colleagues for real-time collaborative work, leave comments and replies, and import and export common file types such as .docx and .txt. The app is available in Proton Drive, an encrypted cloud storage service launched by the vendor in 2022.

But it’s the end-to-end encryption rather than the document editing features that makes Proton’s editor stand out from well-established alternatives on the market.

Only customers are given access to the end-to-end encryption keys, which means that any data entered into a document in Proton Docs is inaccessible by Proton, the company said. That includes keystrokes and cursor movements.

Privacy measures in Proton’s Docs app contrast with the likes of Google Docs, which can “see everything you write and keep a record of all changes that you have ever made,” said Anant Vijay, senior product manager for Proton Mail and Proton Drive, in a blog post.

“Once you provide your data to these companies, you no longer have control over how it is used,” he said, citing growing concerns around the ability of software vendors to train their AI algorithms on customer data.

There’s also the risk that data contained in documents could be accessed should a vendor’s server be compromised.

Another advantage of Proton Docs, the company claims, is that user data is stored on Proton’s cloud servers in Switzerland. Strict Swiss data privacy laws ensure that information stored on Proton’s servers is not subject to access by government authorities in the EU or US, for instance.

The rollout of Docs to Proton Drive customers starts today, with the feature available to all users in the “next couple of days,” Proton said. Proton Drive is available to consumers under a freemium model, with individual subscriptions costing up to €10 a month (currently about US$10.80) billed annually. Proton for Business subscriptions start at €7 per user per month.

More collaboration software news:

• Microsoft’s Team Copilot aims to help manage meetings, group chats
• Slack updates AI ‘privacy principles’ after user backlash
• Meta signals the end of the road for Workplace

Microsoft’s first wave of Copilot+ PCs is here. They’re powered by Qualcomm Snapdragon X Elite hardware, which is a big deal for Windows. This is Microsoft’s version of Apple’s transition to the Arm architecture with its M-series Macs. And existing Windows applications aren’t guaranteed to run on an Arm-powered Windows PC.

The good news is that most applications will run — and Microsoft’s Prism translation layer does a good job of running them with decent speed, even. But not everything will work.

Here’s what you need to know.

Want more insights on the future of Windows? Sign up for my free Windows Intelligence newsletter — I’ll send you three things to try every Friday. Plus, get free Windows Field Guides (a $10 value) as a special welcome bonus!

Qualcomm Snapdragon Arm Copilot+ rule #1: There are no guarantees

The move to an Arm architecture is a big shift. If Microsoft hadn’t created the Prism translation layer, no existing Windows apps would “just work” on a Qualcomm Snapdragon PC. It’s just like Apple’s transition on the Mac, where the Rosetta software enabled existing Mac apps to run on an Arm-based M-series chip.

But the Mac is different. With the Mac transition, Apple put developers on notice: All future Macs would be Arm-based. For Windows, things are different: Only some new PCs use Arm processors. Intel and AMD aren’t being left behind — most Windows PCs will likely be using the traditional x86 architecture for years to come.

To ease the transition, many existing Windows applications will just work on an Arm-based PC. And by “just work,” I mean it  — you can double-click their installers and run them like normal. Unless you dig into the process details in the Task Manager, you might not even know you’re using an x86 application.

But that support only goes so far. Certain types of apps won’t work in the Prism translation layer and aren’t functional. Some hardware devices might not work with these PCs either. Plus, some heavy-duty professional applications could be slowed so much by that translation layer as to be unusable.

Google Drive flat-out refuses to install on a Qualcomm Snapdragon Arm Copilot+ computer.

Chris Hoffman, IDG

Qualcomm Snapdragon Arm Copilot+ rule #2: Some apps will have problems

There are a few types of applications that are guaranteed not to function properly through Prism. They will work if developers port them to Arm — but there’s no guarantee developers will bother, especially for existing business apps.

Specifically, keep an eye out for:

• File sync tools that integrate with File Explorer: These must be ported to Arm to function properly. For example, as shown above, you can’t install Google Drive on a Windows on Arm PC at launch. If this tool is important to you, you will have to access Google Drive in a web browser or use a third-party syncing app.
• Hardware devices that need manufacturer-provided drivers: The Prism translation layer won’t help Windows on Arm use hardware drivers for x86 PCs. In practice, this means many existing hardware devices — especially printers — won’t work. This is one reason why Microsoft is moving away from manufacturer-provided printer drivers.
• Any application that needs a driver: Some applications use drivers to integrate at a low-level with the Windows kernel. For example, many PC games use this for anti-cheat features. This is why Fortnite won’t run on Windows on Arm. But the problem extends beyond games and could affect business-specific productivity tools, too, as any type of application that uses such low-level Windows system integration won’t work. Many third-party antivirus tools don’t support Windows on Arm, either.
• High-end, demanding applications: At launch, the Adobe Premiere Pro video editor does not yet natively run on Arm. While it’s possible to run the x86 version through Prism, many users are reporting severe performance problems. Microsoft says a native Arm version is coming later in 2024. This is just one example, and you might encounter a demanding business application that won’t be ported. (And a demanding application that requires a lot of hardware resources might not deliver the performance you’d expect on an Arm PC.)

The slowdowns aren’t exclusive to high-end applications. All applications will run best on these PCs if the developer ports them to run natively on Arm hardware. But many lightweight applications that don’t need low-level integration with Windows will run just fine, with no perceptible slowdown.

The Details pane in the Task Manager shows which applications are translated 64-bit x86 (x64) software and which apps are native 64-bit Arm code (Arm64). 

Chris Hoffman, IDG

3 ways to see whether your Windows apps run on an Arm Copilot+ PC

I wish there were a big database that would list apps and how they run on these PCs. At launch, there doesn’t appear to be such a website — perhaps someone will launch a resource in the future.

For now, Microsoft has endorsed the ​Windows on ARM Ready Software​ website. However, despite the promising name, that site is just about PC games — which doesn’t do much for users focused on serious workplace productivity.

So here are three practical ways to determine if an application is compatible:

1. Contact the vendor or developer: The best way to find out whether an application will work is to contact the vendor or developer and simply ask whether they support their application on Arm-based versions of Windows like PCs using Qualcomm Snapdragon X Elite hardware.
2. Do research yourself: You might just have to search the web for the name of the application and “Arm” or “Snapdragon” to see if other people are reporting their experiences. You might find some good discussions on Reddit. Your mileage may vary depending on how many people use the application in question.
3. Test it yourself: Many businesses will want to test the applications they depend on before buying Arm-based PCs for their employees. There’s really no way to determine whether a workflow works other than to try it yourself. If you’re an individual, I recommend thinking about return policies: For example, the Microsoft Store has a 60-day return policy. If you buy a Copilot+ PC with an Arm processor from Microsoft and find it doesn’t work with your apps or hardware, you can return it.

With the release of these new laptops in July 2024, it’s still early days. While Windows on Arm has existed for many years now, it’s finally starting to look competitive. The demand for compatible software will likely motivate application developers to start taking it more seriously.

But we all know how Windows works: Some business applications were written many years ago and will never get a major update that ports them to a new architecture. The good news is that many should run fine on these new PCs with no extra development effort. The bad news is that applications that don’t will be left behind.

Still, maybe that’s not so bad. Intel is promising that its upcoming Lunar Lake hardware will be competitive with these Arm-based PCs when it comes to snappy performance with long battery life. Intel’s big pitch is that you’ll get these advantages without the headaches of an architectural shift and with compatibility for all your existing x86 software — no Prism translation layer necessary.

We’ll see whether Intel can deliver on its promises when its next-generation Core Ultra hardware starts arriving later in 2024.

I’ll have lots more to say as I spend more time with these new PCs! Sign up for my free Windows Intelligence newsletter to get all my latest musings along with three new things to try every Friday and free Windows Field Guides as a special welcome bonus.

More on Copilot+ PCs:

• Microsoft Copilot+ PCs: Finally, a reason to buy a new computer
• 5 Copilot+ PC AI features you can use at launch
• Microsoft launches AI-powered Copilot+ PCs

Unknown threat actors have been observed exploiting a now-patched security flaw in Microsoft MSHTML to deliver a surveillance tool called MerkSpy as part of a campaign primarily targeting users in Canada, India, Poland, and the U.S. "MerkSpy is designed to clandestinely monitor user activities, capture sensitive information, and establish persistence on compromised systems," Fortinet FortiGuard

Unknown threat actors have been observed exploiting a now-patched security flaw in Microsoft MSHTML to deliver a surveillance tool called MerkSpy as part of a campaign primarily targeting users in Canada, India, Poland, and the U.S. "MerkSpy is designed to clandestinely monitor user activities, capture sensitive information, and establish persistence on compromised systems," Fortinet FortiGuard Newsroomhttp://www.blogger.com/profile/[email protected]

Vědci pomocí Vesmírného dalekohledu Jamese Webba (JWST) zkoumali oblast atmosféry nad ikonickou Velkou rudou skvrnou na planetě Jupiter. A ke svému nemalému překvapení tam mj. objevili zajímavé světelné jevy. „Možná trochu naivně jsme se domnívali, že tato oblast bude opravdu nudná,“ uvedl hlavní ...

Apple Fellow and App Store head Phil Schiller may have something else to fill his time, taking an observer role on the OpenAI board, a Financial Times report claims. It’s yet another signal of the importance Big Tech now attaches to generative AI.

Schiller hasn’t attended a meeting yet but is expected to take the role as ChatGPT support is rolled into Apple devices. There is a precedent to this. The firm’s other Big Tech partner, Microsoft, also holds an observer’s seat on the board.

Some might say

Some might say the decision to bring Apple more fully inside the tent means OpenAI hopes to persuade Apple to integrate its tech more deeply into Apple products. It seems unlikely that Apple will easily be convinced to move beyond a certain point, in part because it is expected to work with other AI suppliers (principally Google Gemini), but also on strength of its own investments in Apple Intelligence and future fee-based AI services. It seems far more likely to reflect the need to ensure good governance.

Think back and you’ll remember that Microsoft, which has invested $13 billion in OpenAI, gained its own observer’s seat after the November 2023 boardroom battle at OpenAI during which co-founder Sam Altman was fired and then rehired as CEO. 

The truth is neither Apple nor Microsoft will want to countenance poor governance or flawed results as they make the tech available to the world’s population of Windows, Mac, Surface, iPad, and iPhone users. 

Wonderwall

Holding positions, even nonvoting observer positions, on the OpenAI board may help them protect against that, and those roles may expand should Altman’s board have a second meltdown, or in the event the company becomes an acquisition target for either, both, or another big firm.

Microsoft and Apple may also recognize the need to both partner and support AI firms while also developing their own tech, particularly in light of increased regulatory interest in the sector. The US Federal Trade Commission earlier this year launched an inquiry into the partnerships between Big Tech firms and genAI companies. 

“Our study will shed light on whether investments and partnerships pursued by dominant companies risk distorting innovation and undermining fair competition,” said FTC Chair Lina M. Khan in a statement at that time.

Definitely maybe

Competitive concerns aside, the swift evolution of these technologies has thrown a very large brick into the middle of the tech industry pond. Not only does server-based AI generate problems around energy and water supply, but hardware manufacturers are hustling to make or deploy devices with enough computational horsepower to handle this form of AI. Even Apple appears to have been forced to accelerate progress along its processor road map — the M4 MacBook Air was a huge surprise, and with additional M4 models set to ship this year and expectation now that all iPhone models will gain their own higher-end chip, it’s crystal clear the hardware is being tooled up to handle genAI.

There is, however, a limit to what is possible, so it makes sense for Apple — and Microsoft — to gain insight into OpenAI’s future plans, which will both inform their own product development and help guide OpenAI’s. 

Standing on the shoulder of giants

In Apple’s case, the company is also developing its own Apple Intelligence strategy with the introduction of on-device and self-hosted AI models to handle some common tasks, and an anticipated intention to monetize that work somewhere down the line.

Along the way, the company will also be exposing ChatGPT tech to hundreds of millions of people who may never have experienced it before — after all, even though most of the planet now has a smartphone, they may never have experienced artificial intelligence at this level before.  

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

More by Jonny Evans:

• With iOS 18, Apple deepens its connection to India
• Why (and perhaps how) Apple Intelligence will make money
• EU commissioner slams Apple Intelligence delay