Welcome to Verses Over Variables, a newsletter exploring the world of artificial intelligence (AI) and its influence on our society, culture, and our perception of reality.

Tool Update

Flux: Open Source Image Generation

What it is: Flux is the new kid on the block in open-source image generation, brought to you by the rebels at Black Forest Labs. These former Stability AI engineers have created a digital Michelangelo that excels in rendering human features – particularly those pesky hands that have left other AIs fumbling. It's also a savant when it comes to text reproduction, ensuring your AI-generated memes won't end up lost in translation. The masterminds behind Flux aren't just in it for the likes – they're on a mission to shake up the open-source AI world, making high-octane image generation accessible to the masses. It's like the French Revolution, but with pixels instead of pitchforks.

Flux comes in three tantalizing flavors:

• Pro: For the power users who mean business (and can expense it).

• Dev: The sandbox for code-slingers who like to peek under the hood.

• Schnell: When you need that image faster than a New York minute.

How we use it: Now, don't get us wrong – we're not ready to ghost our ride-or-die, Midjourney, just yet. (We're suckers for its style and image reference game, and we're practically drooling for v6.2). But we've been flirting with Flux on Poe, and let's just say it's making DALL-E look like it showed up to a smartphone party with a flip phone. We're incredibly impressed with Flux's photorealism chops. It's like having Annie Leibovitz in your pocket, minus the entourage and lighting crew. And as card-carrying members of the open-source fan club, we're tipping our hats to Black Forest. We’re also really excited to test their upcoming video model. Meanwhile, the cool kids are mixing Flux with Luma or Runway, creating mind-bending videos. Flux isn't just another tool in the digital artist's kit. It's a statement, a revolution, a... oh, who are we kidding? It's just really good at drawing hands. And in the world of AI, that's nothing short of magic.

The AI Hype Cycle

Remember when AI was the hottest thing since sliced bread? Well, the tech world's collective hangover has officially set in. As the giants unveil their less-than-stellar earnings reports, the champagne-popping over AI's imminent world domination has been replaced by the sound of investors nervously tapping their fingers. But before we start writing AI's obituary, let's take a breath. Sure, training these digital brainiacs costs more than a penthouse in Tribeca, and yes, the likes of Meta, Google, and Microsoft are still waiting for their AI sugar daddy to show up with the profits. But hey, Rome wasn't built in a day. Enter Nicholas Carlini, Google DeepMind's resident AI whisperer. While the suits upstairs are sweating over quarterly reports, Carlini's been busy putting these silicon savants through their paces. His verdict? AI isn't just a party trick – it's the ultimate wingman for the modern techie.

Carlini estimates that LLMs have improved his programming productivity by at least 50% on average products. He uses it to assist with tasks like: writing code and converting between programming languages; explaining technical concepts; formatting data; solving common programming errors; and answering questions about hardware and electronics. AI isn't ready to steal Carlini's job just yet. It's more like that eager intern who's surprisingly competent but still needs supervision before you let them near anything mission-critical. The key, Carlini argues, is to focus on what these digital dynamos can do, not what sends them into an existential crisis.

Now, you might be thinking, "That's great for the code-crunching crowd, but what about us mere mortals?" Well, that's where the magic happens. You don't need a PhD in computer science to reap the benefits of AI. Struggling with writer's block? Let AI be your muse. Drowning in research for that big presentation? Your digital librarian is at your service. Can't decide what to cook for dinner? Your AI sous-chef is ready to inspire culinary greatness. The trick is to approach these tools with the wide-eyed wonder of a kid in a candy store. Experiment, play, and find your own AI groove. Who knows? You might discover uses for AI that even the tech geniuses haven't dreamed up yet. So while Wall Street frets over the AI bubble, the real revolution is happening in home offices, classrooms, and maybe even your kitchen. It's not about replacing human ingenuity – it's about amplifying it.

We’ll be talking about our favorite tools, but here is a list of the tools we use most for productivity: ChatGPT 4o (custom GPTs), Midjourney (image creation), Perplexity (for research), Descript (for video, transcripts), Claude (for writing), Adobe (for design), Miro (whiteboarding insights), and Zoom (meeting transcripts, insights, and skip ahead in videos).

Intriguing Stories

California’s AI Balancing Act:  California's Senate Bill 1047, the "Safe and Secure Innovation for Frontier Artificial Intelligence Models Act," is walking a precarious line between fostering innovation and implementing stringent safeguards. This legislative balancing act aims to create a framework for the development and deployment of advanced AI models, with a laser focus on safety and accountability. The bill's crosshairs are trained on the behemoths of the AI world – models with training costs exceeding $100 million. It's like setting a velvet rope at the entrance of an exclusive tech club, where only the most resource-intensive AI systems get VIP treatment (or scrutiny, depending on your perspective). SB 1047's toolkit includes some heavy-duty regulatory hardware: mandatory kill switches, annual third-party audits (because who doesn't love a good checkup?), oversight by the new "Frontier Model Division", and liability measures including fines based on computing costs and potential criminal charges for fibbing about AI safety. Proponents of the bill are singing a familiar tune: "Better safe than sorry." They're drawing parallels to safety regulations in pharmaceuticals and aviation, arguing that if we don't let planes take off without rigorous checks, why should we let powerful AI loose without similar precautions? Critics, however, see the bill as a digital straitjacket, potentially strangling innovation in its crib. They worry it could turn California's tech paradise into a regulatory limbo, sending startups fleeing to more AI-friendly pastures. The opposition to this bill is as diverse as a Silicon Valley coffee shop. Academics, students, and even high-ranking legislators have found common ground in their skepticism. Congresswoman Zoe Lofgren argues that the bill is tilting at windmills – addressing hypothetical existential risks while real, demonstrable AI risks go unchecked. It's akin to fortifying against alien invasions while climate change knocks at the door. SB 1047 gears up for its moment in the spotlight with the Assembly Appropriations Committee on August 15, 2024, before potentially landing on Governor Newsom's desk. As Californians, we're no strangers to the double-edged sword of regulation. We've seen it clean our air and bolster our privacy, often thanking our lucky stars (and legislators) for the foresight. But this time, the AI bill has us squirming in our ergonomic chairs. We can't help but think that maybe, just maybe, the Assembly should pump the brakes on SB 1047. Instead of erecting digital stop signs, why not lay down some smooth pavement for our burgeoning AI industry, especially when it comes to open source.

AI's Reigning Champ Gets a Regulatory Ring Check: Nvidia finds itself in the crosshairs of not one, but two Department of Justice antitrust investigations. The probes spotlight the chip maker's dominant position in the AI market, a realm where Nvidia has become as ubiquitous as avocado toast in Santa Monica. At issue: Nvidia's $700 million acquisition of Run:ai, an Israeli GPU management startup, and allegations of strong-arm tactics in the chip market. Competitors whisper that Nvidia's been playing hardball, nudging cloud providers to stay loyal or face steep premiums on networking gear. With an estimated 70% to 95% market share in AI chips, Nvidia's success story reads like a tech fairytale. But now the DOJ seems intent on fact-checking the happily-ever-after. Nvidia maintains they're just reaping the rewards of "decades of investment and innovation."

Siri’s Little Helper: Apple's foray into the AI playground comes with a hefty dose of helicopter parenting. Beta testers of macOS Sequoia have uncovered a treasure trove of plaintext JSON files, revealing Apple's meticulous attempts to keep its new AI features from going off the rails. Nestled in the system's bowels, these files contain a laundry list of instructions that read like a helicopter parent's guide to raising a well-behaved AI. From utilitarian prompts like "You are a helpful mail assistant" to more existential commands such as "Do not hallucinate," Apple's approach to AI seems to be "spare the rod, spoil the chatbot." The tech giant is trying to avoid the PR nightmares that befell other AI chatbots, like Microsoft's briefly unhinged Bing Chat. While some prompts betray the work-in-progress nature of Apple Intelligence with charming grammatical faux pas, others are laser-focused on preventing the AI from spinning tall tales. "Do not make up factual information," one stern instruction reads, as if scolding a child caught in a fib. These files offer a rare peek behind the curtain of Apple's AI development, potentially allowing tech enthusiasts to play watchdog as the features evolve. However, users eager to chat up Apple's new AI companion must exercise patience – and own the latest hardware. The full rollout of Apple Intelligence features is expected to stretch into next year, with only recent devices invited to the AI party.

Strawberry Fields: (If we weren’t geeking out enough, this is a really deep cut, apologies in advance.) In a move that set Tech Twitter ablaze, OpenAI's Sam Altman dropped a cryptic photo of strawberries in his garden. No, he wasn't bragging about his green thumb. This fruity message was actually a wink to "Project Strawberry," the AI world's most tantalizing mystery since... well, since the last time Altman tweeted something cryptic. Formerly known as Q*Star (because apparently, even AIs need rebranding), Project Strawberry isn't your garden-variety update. It's OpenAI's attempt to give their AI models a Ph.D. in common sense and a master's in "figuring stuff out." The goal? An AI that can surf the web, conduct deep research, and maybe even understand why humans find cat videos so fascinating. But here's the juicy part: Strawberry isn't just about making AI smarter; it's about making it think. If successful, Strawberry could revolutionize everything from scientific research to software development. Imagine an AI that could code the next big app while simultaneously unraveling the mysteries of the universe – and all before your morning coffee gets cold. Rumor has it that every time Google announces something new, OpenAI attempts to steal their thunder, and Made by Google is happening right now.

— Lauren Eve Cantor

thanks for reading!

if someone sent this to you or you haven’t done so yet, please sign up below so you never miss an issue.

if you’d like to chat further about opportunities or interest in AI, please feel free to reply.

if you have any feedback or want to engage with any of the topics discussed in Verses Over Variables, please feel free to reply to this email.

we hope to be flowing into your inbox once a week. stay tuned for more!

banner images created with Midjourney.