How May AI Assist You?
Last fall, rumors circulated that OpenAI’s Sam Altman was partnering with legendary Apple designer, Jony Ive, on a new device for the age of AI. The duo was in the early stages of sketching out a device that would, according to The New York Times, “succeed the smartphone and deliver the benefits of AI in a new form factor, unconstrained by the rectangular screen that has been the dominant computing tool of the past decade.”
Eight months on, it appears rectangular screens will hold sway a bit longer. In a month rife with AI announcements from the biggest names in tech, smartphones, laptops, and tablets were the focus. For now, it seems, we’ll need little else to get our AI fix. It makes sense. These devices have advanced sensors, chips, internet connectivity—and nearly everyone has one. You can tack on the latest AI with a simple download or update.
“Just as chatbots have promised to condense the internet into a single program, generative AI now promises to condense all of a smartphone’s functions into a single app, and to add a whole host of new ones,” The Atlantic’s Matteo Wong wrote.
But it’s not just smartphones (or tablets). The humble PC will get an AI makeover too. This month, OpenAI, Google, and Microsoft stumbled over one another to remake how we interact with the devices and digital services already used by billions. Even Apple, long quiet on the AI front, is rumored to be close to unveiling its own AI vision.
Let’s take a look.
OpenAI. OpenAI scooped its competitors with the release of GPT-4o. While the underlying AI model itself got faster and cheaper, it’s still a member of the GPT-4 family. OpenAI said they combined multimodal abilities—images, video, audio, and text—into a more seamless experience. The new interface, which takes a page from voice-activated assistants like Siri and Google Assistant, can use a smartphone’s cameras to remark on the clothes you’re wearing or help you find your car keys. The wildest part (and the most controversial—see here and here) is the addition of a voice interface with personality.
“Watching the presentation, I felt that I was witnessing the murder of Siri, along with that entire generation of smartphone voice assistants, at the hands of a company most people had not heard of just two years ago,” Wong wrote.
Google. The day after OpenAI’s demo, Google went all-in on AI at its I/O developer conference. Presenters mentioned AI 121 times in a 110 minute keynote. The company’s Gemini family of AI models will power new features in virtually all its products. Most notably, Google is now adding AI overviews in answer to queries in its core search product. According to a Google blog post, hundreds of millions of people are already seeing them, and they’ll reach over a billion by year-end. The company is also building an AI assistant, Project Astra, with capabilities similar to GPT-4o—though perhaps with less personality—and a video-generating AI, Veo, akin to OpenAI’s Sora.
Microsoft. Not to be outdone, Microsoft, which got in on generative AI early thanks to a $10+ billion investment in OpenAI, similarly said it’s throwing the tech at just about everything—even its Windows operating system. The company announced software for a new line of AI-centric laptops. Called Copilot+ PC, Microsoft will weave in as many as 40 small AI models—that can run locally on Windows laptops with enough power—to help people more easily track down documents and emails, edit photos, or do translation.
Apple. Apple has long had an AI assistant, but even in Siri’s heyday, when the impressive achievement was an interface that reliably understood speech, the technology fell short of expectations. There are now rumors that Apple is in talks with OpenAI—and allegedly nearing a deal—and Google to license GPT and Gemini for iOS. The company is also building its own large language models and is said to be overhauling Siri. Like Microsoft, Apple may lean more heavily on running AI locally, as opposed to in the cloud. But the company hasn’t officially confirmed its plans. We’ll learn more at WWDC next month.
Despite nearly every tech major making huge investments in the productizing of AI, there’s still debate about how much of it will stick. Tech pundits are fielding strong opinions on both sides: It’s either “time to believe the AI hype” or time to “press pause on the Silicon Valley hype machine.” There are decent points to be made on both sides. Looking back a decade from now, we may judge it to be somewhere in between.
It’s also worth noting that while OpenAI’s new demo is impressive, it isn’t GPT-5. Google too has yet to take its core technology the next step. It seems we’re in the phase of talented engineers and designers building the stack around a core whose competencies—which are difficult to benchmark—appear to be mostly static and whose weaknesses are still apparent. (As we wrote last month, GPT-4-like algorithms are increasingly common.) The models are getting more efficient though, and there’s room to build and design new products that move the needle, even without GPT-next.
What can’t be debated is that billions of people will soon be given the option to incorporate AI into their lives, and the interface between new technologies and people at the scale of populations is chaotic. There’s no knowing how many will become regular users, how careful they'll be, or what unintended consequences may result. Saying a technology will change everything is easy enough; predicting how it will change everything, not so much.
More News From the Future
Recommended by LinkedIn
Biomodels. Google DeepMind and Isomorphic Labs, a biotech company spun out of DeepMind, recently released an updated version of their AlphaFold algorithm. Using a diffusion model like the one behind OpenAI’s DALL-E, AlphaFold can now model a range of biomolecules—these are the players driving the biology of all living creatures, including proteins, DNA, RNA, and small molecules, such as ligands—and predict how they interact in the body. The algorithm vastly outperformed its peers in some areas (though there’s work to do in others). The companies didn’t openly release the algorithm, as they did its predecessor, but are instead offering an online platform, AlphaFold Server, where biologists can run simulations of biomolecule interactions.
Drug discovery. In 2020, when Google DeepMind first released AlphaFold, it was one of the biggest AI breakthroughs to that point. It also essentially solved a grand challenge in biology. But the algorithm could only model proteins, which are just one player in biological processes. To speed drug discovery, a central goal, scientists need to model other biomolecules too and test how they all might interact—like, for example, how a small-molecule drug might shut down a disease-causing protein. The new model improves that capability, allowing scientists to better understand the causes of disease and design interventions.
Forget chatbots. Next steps include increasing accuracy across a wider range of interactions—the algorithm is as much as 76 percent accurate for interactions between proteins and small molecules but isn’t as good for protein-RNA interactions. It’s also not as open as its predecessor. The online platform could take some of the complexity out of the modeling process, but it is restrictive—a decision that, according to DeepMind, seeks to balance its scientific impact with Isomorphic’s commercial drug discovery efforts. The company is already using the system in-house and with partners to design new drugs. Regardless, it’s a healthy reminder that meaningful advances in AI go well beyond chatbots.
Cerebras’ Wafer Scale Engine (WSE) is the biggest computer chip in the world. Primarily used to train giant AI models, the chip is as big as a dinner plate. It’s also a curiosity: one of one. TSMC, the world’s leading chip company, and maker of the WSE, wants to change that. The company recently said it’s working on technology to make wafer-scale chips more of a commodity. By 2027, new wafer-scale systems could “deliver 40 times as much computing power as today’s systems,” IEEE Spectrum’s Samuel K. Moore wrote.
To celebrate the 10th anniversary of his tech column in the Wall Street Journal, Christopher Mims shared what he’s learned from numerous mistakes over the years. Lessons include the fact that disruption is rarer than we think, especially these days, and that, to cut through the hype, you have to understand the motivations of those pitching new ideas—for the most not malicious, but at times, necessarily self-deluding. Most importantly, no matter how mind-blowing a new technology may appear to be, people have the last say. The speed of adoption depends on the “quirky, unpredictable, and far-from-rational set of predilections, needs, and biases resident in all of us.”
Despite setbacks, self-driving cars continue to make progress.
Waymo is expanding paid robotaxi services, and this month, the company said its self-driving fleet is completing 50,000 paid rides a week, or roughly five times more paid rides per week than reported last August. Meanwhile, self-driving startup, Wayve, raised over a billion dollars for its AI-first approach. The company has shown that its software can learn to drive the streets of one city and operate one kind of car, and then go on to drive the streets of other cities and operate different cars without being re-engineered.
Upcoming Events
Join us for a LIVE Executive Program information session
In this interactive 30-minute webinar hosted by Singularity VP Alix Rübsaam, you'll get to know more about the 5-day program's content, speakers, and logistics. Please bring whatever questions you may have!
Ready to apply? Choose one of our 2024 EP dates:
Thank you for stepping into the future with us! Join our global community of over 200,000 futuremakers - sign-up and unlock early access to the Singularity Monthly newsletter and discover the most impactful technology breakthroughs - get a glimpse of tomorrow, today.
The Singularity Team
Worth following … THE Singularity University … ahead of the game in latest research findings …,