Why Apple isn’t rushing into the AI arms race

Welcome to AI Decoded, Fast Company’s weekly LinkedIn newsletter that breaks down the most important news in the world of AI. I’m Mark Sullivan , a senior writer at Fast Company covering emerging tech, AI, and tech policy.

This week, I’m focusing on the underlying AI element in Apple’s big headset announcement at WWDC . Meanwhile, open-source generative AI models, such as the United Arab Emirates’ new Falcon 40B, could change the complexion of the ongoing AI arms race. And super investor Marc Andreessen weighs in on the responsible AI debate.

If a friend or colleague shared this newsletter with you, you can sign up to receive it every week here . And if you have comments on this issue and/or ideas for future ones, drop me a line at [email protected].

Apple’s utilitarian approach to AI?

While many Big Tech companies have been hurrying to catch up with OpenAI in the generative AI arms race , sparked by the surprising popularity of ChatGPT, Apple has been noticeably absent from the conversation. Some speculated that the company might announce a bunch of new AI-driven features for iPhones, iPads, and other iThings at its developers conference this week. Not so much. While the company did announce a handful of AI-powered features , it’s clear that Apple continues to view AI as an enabling technology that works best in the background.?

The day’s big announcement, of course, was the Apple Vision Pro headset , which Apple says will usher in a new computing paradigm called spatial computing. The headset projects digital 3D imagery within the visual world in front of the wearer. But again, AI was barely mentioned, even though it’s likely that Apple’s computer-vision AI models enable some of the device’s core experiences.

I suspect that generative AI will become a big part of future versions of Apple’s XR headset. And I’m not the only one. Mixed reality pioneer Magic Leap founder Rony Abovitz said on Twitter after the Apple event: “Spatial computing feels like the ultimate playground for AI.” It’s very likely that the XR headsets of the future will generate original 3D imagery based on an AI model’s understanding of the user’s words, interests, or tastes, something like the way TikTok’s algorithm serves up videos closely based on videos we’ve liked in the past.

Marc Andreessen has spoken: AI is safe

A debate has been raging for months over the near-term and long-term safety of new generative AI systems. The near-term risks, many say, is that new AI systems will replace people and leave them jobless. The longer-term risks involve superintelligent AI systems that eventually see humans as counterproductive to their goals. But don’t worry. Marc Andreessen says it’s all going to be okay.?

The Netscape founder and super investor came out swinging against “AI doomers” in a Twitter essay Tuesday, arguing that both the short-term and long-term fears around AI are overblown; merely echoes of fear responses to new tech in the past. He writes:

“AI is not a living being that has been primed by billions of years of evolution to participate in the battle for the survival of the fittest, as animals are, and as we are. It is math – code – computers, built by people, owned by people, used by people, controlled by people. The idea that it will at some point develop a mind of its own and decide that it has motivations that lead it to try to kill us is a superstitious handwave.”

Andreessen believes that everybody, from scientists to school children, will have AI assistants in the future and that these assistants will be kind, supportive, and moral (because we’ll make them that way).?

He puts AI doomers into two categories: those who might benefit financially from tight regulations on AI development, and those who truly (if mistakenly) believe that AI systems are dangerous. He says that people paid by AI developers to manage risks belong in that first camp of doomers. “There is a whole profession of ‘AI safety expert,’ ‘AI ethicist,’ ‘AI risk researcher,’” Andreessen writes. “They are paid to be doomers, and their statements should be processed appropriately.”

Without naming names, Andreessen seems to suggest that OpenAI CEO Sam Altman has something to gain by talking about the dangers of AI. Altman has been calling for new regulation of AI development, and, Andreessen says, such regulations could slow down the work of smaller developers who might threaten the fortunes of current market leaders.?

Open-source AI is getting better?

And OpenAI’s leadership position may not always be secure. The company’s GPT-4 model is shockingly conversant and smart, but its secret sauce (the mathematical weights in its models, training methodology, etc.) is proprietary to OpenAI. Developers in the open-source world, including Meta and Stability AI, routinely post their models on Github or Hugging Face for anybody to build on. In the past, open-source large language models (LLMs) have not matched the performance of proprietary models, but that may be changing. A new model called Falcon 40B , recently released by the Technology Innovation Institute , a research arm of the United Arab Emirates, outscored all other open-source LLMs in a series of standardized tests . And Falcon 40B is thought to be competitive with models like OpenAI’s GPT-4 and Google’s PaLM.?

Falcon 40B has 40 billion parameters (the mathematical connection points in neural networks) and was trained on 1 trillion tokens (text characters, word parts, or whole words). If open-source models equal the performance of their closed cohorts, more large enterprises might use them to build and host their own models rather than rely on those owned by third-party AI developers.

More AI coverage from Fast Company:

• The creators of Replika unveil a new AI dating app called Blush
• I used AI to plan a weekend getaway in California. The trip was a mixed bag
• POV: AI could easily take over our elections—and undermine democracy
• POV: Should we give up on human learning?

From around the web:

• ChatGPT Maker OpenAI Is Staying Private So It Can Make ‘Strange’ Decisions (Bloomberg)
• The AI Founder Taking Credit For Stable Diffusion’s Success Has A History Of Exaggeration (Forbes)
• AI chatbots lose money every time you use them. (Washington Post)

OpenAI executive appears to have Twitter account hacked to promote cryptocurrency scam (VentureBeat)