Welcome to Verses Over Variables, a new AI-focused newsletter brought to you by the same writer of On the Edge.

This newsletter will explore the intriguing realm of artificial intelligence (AI), exploring its profound impact on society, culture, and the way we perceive the world. For On the Edge fans, don’t worry, we are still aiming to publish both once a week.

As AI advances at an unprecedented pace, it is becoming increasingly essential to stay informed about this transformative technology. From revolutionizing industries and reshaping job markets to raising profound questions about creativity, consciousness, and the very nature of intelligence, AI's impact is far-reaching and unavoidable. Whether you are a business leader looking to harness the power of AI, a professional seeking to future-proof your career, or simply someone who wants to understand the forces shaping our world, Verses Over Variables will provide you with the insights and knowledge you need to navigate this exciting and uncertain terrain.

The name Verses Over Variables comes from an anecdote we heard recently — for the first time, we are confronted with an AI that is great at poetry and terrible at math. None of our science fiction futures has prepared us for this, and we are excited to keep you informed and inspired by the cutting-edge developments shaping our world.

This week’s newsletter will be a bit of a catch-up or a kickstart, as we attempt to find the right balance of new tools, technology, and news.

Some of Our Favorite Tools

ChatGPT-4o: It’s more than just a pretty interface

What it is: By now, you've probably heard the buzz surrounding OpenAI's latest brainchild, ChatGPT-4o. If you caught any of the demos during its launch, you know the hype is justified, despite the Scarlett Johansson drama. This AI isn't just another pretty face – it's a game-changer.

First things first: OpenAI has made the chatbot accessible to everyone, free of charge. And if typing isn't your thing, just click on the microphone button and have a heart-to-heart with your new digital bestie. This new model is incredibly fast, can browse the web, and even provides links to sources (fewer AI-induced hallucinations). The responses it generates are surprisingly accurate, making it a valuable research tool. While the full model isn't yet available to the public, it will soon have impressive voice, vision, and spatial recognition abilities, making it truly feel like "Her" or Jarvis.

How we use it: We use ChatGPT for quick research, but the true magic lies in CustomGPTs, which are now also free for everyone. CustomGPTs are minibots created by users for specific tasks. (If you’d like to try one out, we recommend Cartoonize or Diagrams.) We’ve also created a few of our own to streamline our workflow — it’s like having a personal assistant that can handle all those repetitive tasks you dread, with just a single click. No more rewriting queries or prompts; just head to the CustomGPT and chat.

Midjourney: Turning words into wonders

What it is: Midjourney is an innovative application that transforms text descriptions into stunning visual images. All you need to do is type in a prompt – a brief description of the image you'd like to create – and Midjourney will work its magic, generating a visual representation of your wildest dreams. Over the past year, Midjourney has made significant strides, introducing a host of incredible new features (style, character and image reference, for instance). And with its user-friendly web interface, collaboration has never been easier.

Unfortunately, Midjourney has found itself embroiled in controversy. The company has been accused of training its model on copyrighted images without obtaining proper permission. As a result, users can easily mimic the styles of specific artists, raising questions about intellectual property rights and artistic integrity. Several lawsuits are currently underway, casting a shadow over Midjourney's otherwise impressive achievements.

How we use it: We've embraced Midjourney as a productivity tool and a catalyst for our creativity. Gone are the days of spending hours in Photoshop or Illustrator, painstakingly crafting the perfect image for a design, doodle, or mood board. With Midjourney, we can conjure thousands of images in a fraction of the time, allowing our imaginations to run wild. While Midjourney is a paid service, free alternatives are available for those who want to dip their toes into the world of AI-generated art. You can ask ChatGPT to draw something for you or give Leonardo a try. These tools may not have all the bells and whistles of Midjourney, but they're a great way to explore the potential of AI in the creative process.

We’ll be talking about our favorite tools each week but here is a preview of what tools we use most for productivity: 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

Anthropic Cracks the Code - Groundbreaking Glimpse Inside AI’s Brain - One unique feature about the new AI language models is that their developers don’t fully understand why they behave the way they do - the models replicate human language through a complex and intricate network of artificial neurons, making them challenging to comprehend. Furthermore, the models are non-deterministic: you can ask the same question twice and get a completely different answer each time. Scientists at Anthropic (the makers of Claude) decided to peer inside the black box that is artificial intelligence, and they were actually able to map out some unique patterns. One pattern, or feature, that the team honed in on was the model’s apparent fascination with the Golden Gate Bridge. By amplifying this feature, the researchers found that the model often found ways to mention or relate things back to the iconic landmark, even when it wasn't directly relevant. The model was so intriguing, that the team even released a focused Golden Gate Claude for the day. (For instance, users asked GGC “what is the meaning of life?" and received variations of “the meaning comes from our own human constructs - the Golden Gate Bridge, the Redwood Forest, and San Francisco itself.) This research offers a tantalizing glimpse into the mind of AI, revealing that these models don't just process information, but can develop their own unique "fixations" and ways of understanding the world. Luckily, this time the machines were infatuated with infrastructure, but a small tweak could lead to less than humorous consequences.

AI for Productivity and Efficiency - AI Saves Time and Cognition: We’re huge fans of AI when it comes to our personal productivity. And we are not alone. A new report from PwC found that AI is transforming how we work, with sectors exposed to AI experiencing a productivity boost 5x greater than fields not using AI. Unfortunately, despite all the hoopla, drama and memes, a mere 7% of Americans (2% in the UK) are actually using ChatGPT on a daily basis. For most, it's been more of a one-hit wonder - something to try once or twice for the novelty factor before returning to the familiar comforts of doom-scrolling Twitter and watching cat videos. The AI revolution may be coming, but you still have time to jump on the head of the train. At least, the machines will have to wait a little longer to take over our screens and our lives.

Altman’s Odyssey - Navigating the Treacherous Waters of AI Leadership: Between disgruntled ex-employees and concerned board members, Sam Altman (the CEO of OpenAI)’s journey is starting to resemble a Greek tragedy. OpenAI, valued at $80BN, and known for its groundbreaking models like ChatGPT and DALL-E, has faced internal strife recently including battling memos from board members. Former member, Helen Toner, accused the CEO of misleading the board and prioritizing commercial products over safety (apparently the board found out about ChatGPT the same way we did 🤦🏻‍♀️, through Twitter). While the new board penned another memo highlighting the firm’s commitment to safety, regulation, and governance. OpenAI’s safety team, also known as the Superalignment team, was disbanded after several high profile departures, only to be re-established as a new Safety and Security Committee. While any of these setbacks might topple your average CEO, Altman has already survived one coup attempt in November 2023, and has signed several partnership deals including one with Apple (hopefully integrating OpenAI’s product into Apple products this June). Personally, we are quite fond of OpenAI’s products, and we can’t wait to get our hands (eyes and voice) on their new multimodal features. But when we are dealing with technology as important and instrumental to the economy, education, and climate, we’d prefer some actual regulation and less drama.

AI Scores Don't Add Up - Exaggerated Claims About GPT-4's Bar Exam Performance: When GPT-4 was released last year, one of the most hyped claims was its ability to score in the 90th percentile on the legal Bar Exam. Unfortunately, ChatGPT’s claim to fame, might have been another exaggeration. An MIT scientist has re-analyzed the data, and found that the true performance was closer to the 60th percentile compared to first-time test takers, and may have been as low as the 15th percentile on the essay portion compared to those who passed the bar. He contends that OpenAI cherry-picked data from an unrepresentative cohort of low-scoring repeat test takers to puff up GPT-4's apparent percentile rank. We aren’t legal experts, and we have learned always to check an AI-generated response, so at least our job (and those of your paralegals) should be safe for now.

AI Gold Rush Continues - Billions Pour into AI Startups as Valuations Soar to New Heights: The world of artificial intelligence is witnessing a funding boom reminiscent of the dot-com era. Despite concerns about the high costs of developing and training the technology, the looming threat of competition from tech giants (nothing to say of the environmental impact or the potential that we’ve reached an innovation plateau), investors seem undeterred. Take Perplexity, the AI-powered search engine, which is in talks to raise $250M at a $3BN valuation - triple what is was valued earlier this year. Mistral, a European, open source competitor, is closing in on a $600M funding round for a $6BN valuation - a threefold increase from its $2 billion valuation just last December. And we can’t forget xAI, Elon Musk’s startup, rumored to be on the verge of securing $6BN for an $18BN valuation. (As of the end of April, close to $14BN was raised by AI startups in 2024.) On the flip side, however, the public markets are telling a different story. The recent sell-off in enterprise software stocks (see Salesforce), driven by concerns over macroeconomic uncertainty and questions about the effectiveness of AI-powered tools, suggests that investors in the public sphere are far more cautious about the near-term prospects of AI adoption.

— 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.