AI BINGO: A Mid-Year Review

Checking in on the 2024 Predictions - Which Squares Are Heating Up?

We’re officially halfway through 2024, and the AI world is a whirlwind of activity. It feels like a good time to dust off my AI BINGO card, see which predictions are panning out, and which ones need a serious reality check.

Spoiler alert: No BINGO yet. But with half the year still to go, the game is far from over. Some squares are practically begging for attention, while others…well, let’s just say the jury's still out.

So, grab your virtual dauber, and let’s dive into the current state of AI BINGO, 2024 edition:

Deep Dives: A Closer Look at the Predictions

Let's break down each square, analyze the current landscape, and see which predictions are on the verge of becoming reality.

Small models become a BIG deal - Stamp Approved

This one’s a clear winner! The race to lightweight, efficient AI models is on, and it’s not just talk.

• Microsoft launches Phi-3 Mini, an AI model that is smaller but still rivals GPT-3.5
  “the first of three small models the company says it'll launch in the coming months”

• Introducing Apple’s On-Device and Server Foundation Models
  “a ~3 billion parameter on-device language model”

• Google is building its Gemini Nano AI model into Chrome on the desktop
  “Google announced that it is building Gemini Nano, the smallest of its AI models, directly into the Chrome desktop client”  
  [video from mortenjust@ on X]
  

On-device AI models in all high-end hardware - Cautious Optimism

This square is practically on fire! It's not just talk anymore; on-device AI is becoming a reality across high-end hardware.

• Microsoft’s announcement of Copilot+ PCs, integrating AI directly into the hardware, is a huge step in this direction

• Apple also announced Apple Intelligence coming this fall - AI that is built into your iPhone, iPad, and Mac to help you write, express yourself, and get things done effortlessly.

• Samsung’s Galaxy ​​S24 comes loaded with AI 

• Honor announce the Magic 6 Pro phone would also come with AI eye tracking built in

• OPPO has set a goal to integrate AI across its entire smartphone lineup by the end of 2024

The momentum is undeniable. While we're not at the point where every high-end device has a dedicated AI chip, the trend is clear. This square is rapidly moving towards a checkmate.

AI auto-writers in all internet textboxes - Holding Out Hope

LinkedIn is the clear frontrunner here amongst the products I use, seamlessly integrating AI writing assistance into its platform.

With Google's Gemini Nano poised to empower developers with accessible AI tools directly in Chrome, we should see a surge in this type of functionality across the web. Think AI-powered email messages, blog posts, even social media captions. This square is on the cusp of a major transformation.

Chief AI Officer becomes a common role in large companies - Trending Upwards

The conversation around the Chief AI Officer (CAIO) is gaining momentum. Articles are popping up, think pieces are being published (The Rise Of The Chief AI Officer: Is A Board Equivalent Necessary?), and companies are recognizing the need for a dedicated leader to navigate the complex world of AI. While it's not yet a standard C-suite position, it’s definitely gaining momentum. As AI becomes increasingly integral to business strategy, I still expect to see CAIOs become a lot more commonplace.

This will become the year of “agents” - Gaining Momentum

The hype surrounding AI agents is real, and for good reason. These intelligent assistants hold the potential to revolutionize how we interact with technology, automating complex tasks and streamlining our digital lives.

• LangChain continues to lay the groundwork, creating a robust orchestration layer for AI agents that empowers developers to build increasingly sophisticated and powerful agents.

• The emergence of specialized agents is another exciting development. Factory, for example, is developing AI agents specifically for software development. Imagine a world where coding tasks are automated, bugs are squashed before they appear, and software practically writes itself.

• The release of Princeton's SWE-agent underscores this potential. 

• But perhaps the most significant breakthrough in this space is Devin, an AI-powered coding assistant developed by Cognition. This groundbreaking technology can reportedly code entire projects from a single prompt. The buzz around Devin is palpable, with Founders Fund leading a $175 million investment in Cognition, valuing the startup at a staggering $2 billion.

The AI agent revolution is well underway, and this square is rapidly moving towards a blackout. The future is automated, my friends.

Elon declares AI sentient- Still Waiting...

While Elon Musk continues to make waves in the AI world with his ventures like xAI, he hasn't gone so far as to declare AI sentient yet. However, he did predict that superhuman AI, surpassing human intelligence, will emerge next year. So, stay tuned. This square could flip at any moment - he tends to be unpredictable like that.

Monetary / environmental cost of AI gains attention - Talk is Cheap...But Conversations Are Happening

While the alarming energy consumption of AI is being acknowledged by some, from tech commentators to even political figures like Trump, concrete actions to address this looming crisis are sorely lacking. Talk is cheap, and until we see significant investments in sustainable AI infrastructure, responsible development practices, and regulations that prioritize environmental impact, this square remains unchecked.

Focus moves from “prediction” to “reasoning” - A Moving Target

This square presents a fascinating dilemma. While the initial goal of shifting from simple AI predictions to complex reasoning chains is being realized through tools like chain-of-thought prompting, the end game might render these frameworks obsolete. The idea is that these reasoning chains, while helpful now, could ultimately serve as training data, enabling models to learn inherent reasoning abilities, bypassing the need for multi-step processes. The goalposts keep moving, making this square a fascinating case of AI evolution in action.

We will see an influx of AI-first hardware devices - Not Quite a Boom...Yet

Honestly, this prediction hasn't lived up to its initial hype. While we've seen some intriguing entries like the AI-powered rhyming clock Poem/1, Ray-Ban Meta Glasses, Halo Headband, Brilliant frame AI glasses and the XREAL Air 2 Ultra AR Glasses, major flops like the Humane AI Pin and Rabbit R1 have dampened the excitement. The jury is still out on this square, but for now, it remains unchecked.

AI blamed for causing election upheaval - Let's Hope This One Stays Blank

It's still too early to tell how AI might impact upcoming elections, but the potential for misuse is a valid concern. Let's hope this square remains blank, for everyone's sake.

10+ unicorn companies building SOTA Open Source - Unicorns Embracing Open Source

The open-source AI movement is gaining serious traction:

• Google building Gemma

• Meta building Llama

• NVIDIA building Nematron

• Cohere building Command R+

• Alibaba building Qwen2

• 01.AI building Yi

• Mistral building Mixtral

... all represent powerful open-source models pushing the boundaries of AI. This square is well on its way to being checked off.

A new eval standard replaces academic benchmarks - Leaderboards Enter the Arena

While academic benchmarks still hold weight, the rise of AI leaderboards adds an interesting twist. Platforms like:

• LMSYS chatbot arena and a leaderboard

• Scale recently launched a set of leaderboards across Coding, Math, Instruction Following, Spanish

• Vectara has a hallucination leaderboard

• Berkeley has a Function-Calling leaderboard

• HuggingFace has their leaderboard for open models

• SWE-bench measures the ability of Language Models to resolve real-world GitHub Issues

... provide a dynamic way to assess and compare model capabilities. While a single standard hasn't replaced traditional benchmarks, the landscape is evolving, making this square one to watch.

A currently hyped up AI company goes bankrupt - Close Calls, but No Casualties (Yet)

This square had a couple of near misses in mid-March. Inflection AI's CEO shakeup, followed by key departures at Stability AI, had many predicting their demise. However, Inflection AI rallied with a new CEO and a renewed sense of purpose, while Stability AI secured an $80 million funding round. For now, this square remains unchecked, but the ever-shifting sands of the AI startup scene suggest it may be only a matter of time.

Synthetic data gains in popularity - Gaining Traction, but Will it Stick?

With the increasing demand for large, diverse datasets to train AI models, synthetic data is emerging as a potential solution. This trend is fueled by concerns around privacy, bias, and the sheer logistics of acquiring real-world data at scale. Promising research highlights the potential for synthetic data, however, whether synthetic data truly delivers on its promise and becomes the go-to solution for data-hungry AI models remains to be seen. This square is heating up, but the jury is still out.

A full length movie will be almost entirely AI generated - The Reels Are Turning

While we haven't witnessed a full-blown AI-generated feature film just yet, the building blocks are falling into place. OpenAI's Sora, Luma Dream Machine, and Runway's Gen-3 Alpha are pushing the boundaries of AI video generation with stunning realism. Even Google DeepMind is partnering with Donald Glover to create a short film using their new AI model, Veeo. This square is glowing with anticipation, and a full-length AI-generated movie feels inevitable.

Demand will keep growing for prompt engineers - Prompting: An Art and a Science

The prediction might have underestimated the learning curve associated with effective prompting. While tools and techniques are evolving, crafting the perfect prompt to elicit desired outputs from AI models requires a blend of creativity, technical understanding, and often, a healthy dose of trial and error. The verdict is still out on this square, but the skill of prompting is likely to remain in high demand.

Focus moves from text & image to video & audio - A Multimodal Future Unfolds

This square deserves a resounding checkmark! The AI landscape is rapidly embracing multimodality, moving beyond text and images to incorporate video and audio.

• Video output: OpenAI's Sora, Luma, Runway

• Video understanding: Advancements are happening rapidly in this area.

• Audio understanding: Google's Gemini 1.5, OpenAI's GPT-4

• Audio output: ElevenLabs Sound Effects

... are transforming how we interact with and experience AI. Expect even more exciting developments in this space.

An AI-written piece will win a Pulitzer Prize - A Long Shot, But Stranger Things...

This prediction remains a long shot. While AI writing tools have advanced considerably, they still lack the nuance, creativity, and depth required for Pulitzer-worthy writing. However, with the rapid pace of AI development, who knows? This square might surprise us yet.

Ilya leaves OpenAI - Another One Bites the Dust

This prediction came true, though it took longer than anticipated. Ilya Sutskever's departure from OpenAI surprised probably no one.

a16z raises a $5B+ fund to invest in Generative AI - Betting Big on Generative AI

While a16z didn't announce a single fund specifically dedicated to Generative AI, their $7.2 billion war chest, spread across five investment strategies, speaks volumes. Notably, their Infrastructure and Growth funds, totaling $5 billion, heavily emphasize AI, from foundational models to widespread AI adoption. This square gets a confident checkmark.

An AI lawsuit makes its way to the Supreme Court - The Jury's Still Out

While the legal landscape is grappling with the implications of AI, a landmark case has yet to reach the Supreme Court. The Court's recent decision to decline an appeal on AI patent eligibility doesn't set a precedent, leaving many questions unanswered. This square remains unchecked, but the potential for a groundbreaking AI lawsuit to reshape legal boundaries looms large.

AIs calling other AIs to complete tasks - The Future of AI Collaboration?

This prediction represents a fascinating frontier in AI development. While we haven't quite reached a point where AIs seamlessly collaborate to complete complex tasks, the foundational technologies are emerging. This square remains a tantalizing glimpse into a future where AI agents work together, leveraging each other's strengths to solve complex problems.

An alternative to the transformer architecture gains traction - Transformers Still Reign Supreme

While alternatives like Jamba from AI21 Labs, Graph Neural Networks, and Neuro-Symbolic AI are gaining attention (extensify.ai and RAAPID leaning in here), the transformer architecture remains the dominant force in AI. This square might take a while to get checked off.

Nvidia becomes a full-scale cloud provider - Sticking to Their Strengths

Contrary to the prediction, Nvidia seems content to dominate the hardware arena, supplying their powerful GPUs to cloud providers rather than becoming one themselves. This strategy allows them to capitalize on the AI boom without navigating the complexities of the cloud computing market. This square remains unchecked.

A major insurer will offer an AI-specific hallucination policy - Vouching for AI Hallucination Coverage

While Vouch's AI insurance policy does exist, their "major insurer" status is debatable, and their coverage focuses more broadly on AI-related risks rather than specifically addressing hallucinations. Therefore, this square remains in a gray zone.

The Current Status: Game On!

As we hurtle toward the second half of 2024, the AI BINGO card is a testament to the unpredictable nature of this transformative technology. Some predictions have materialized with remarkable speed, while others face unexpected roadblocks or evolving definitions of success. The real takeaway? Buckle up; it's going to be a wild ride.