Back To The Future

Welcome back to the AI Geekly, by Brodie Woods. We’ve been grinding-out our Intro to AI Primer and have been remiss in not providing key updates during what has remained a busy time in the AI space. We’re going to keep the short format from the prior note as it received positive feedback from the readership. Periodically we will provide a thought piece, like the coming Intro to AI Primer, where we’ll take a deep dive into a particular topic of interest in the AI space.

TL;DR OpenAI’s Sora Soars; Talking to GPUs; Bard is Twinning; Air Canada Misses Its Flight; Open Source Remains the Stock Horse

OpenAI is shaking-up society again with its Sora video generation model. While not yet released publicly, it’s poised to majorly disrupt video content generation (please let it get rid of the influencers). Nvidia’s Chat with RTX blends speed and privacy via local inference. Google Bard has a new name and packs more punch! Air Canada cements its status as Canada’s worst airline, blaming its own chatbot for providing misleading information. Finally the Open-Source community released a pile of cutting edge models that remind us why we have such faith in these people as stewards of trustworthy AI. Rather than keep these models to themselves, they share them publicly and the rising tide lifts all boats —a sentiment similar to Benjamin Franklin’s refusal to patent his inventions (bifocals and lightning rods, etc.) of considerable societal benefit.

A Taste of Their Own Medicine?
OpenAI’s new sora model poised to replace Hollywood

*the videos above are 100% synthetic, generated by OpenAI via text prompts.

What it is: OpenAI this week announced a groundbreaking new AI model dubbed Sora, capable of creating full motion video clips up to one minute in length using solely text input.

What it means: The ability to create convincing synthetic video nearly impossible to differentiate from real footage (digital watermarks are planned) is humanity crossing another rubicon. While deepfakes have worried many, the ability to create complete synthetic videos of anything makes it even harder to ensure veracity. A tricky proposition in highly polarized societies with dueling versions of “truth”.

Why it matters: While OpenAI hasn’t provided details on when or how they plan to release Sora (and quite possibly, they may not) they now have in their hands a tool which, when improved beyond one minute length, is a serious contender to disrupt the entire film industry. A bit ironic for the movie studios who just months ago were eager to replace talent with AI. Take note because this is how fast the technology is moving —there are no “AI resistant” jobs.

Closing thoughts: Society will need to contend with the natural impact of replacing human labor entirely with AI in the coming decade, perhaps even earlier. We expect near-term growing pains, but longer-term we see a new era of prosperity, ushered in by clever application of AI tools to tackle humanity’s greatest challenges. From Opiod addiction to climate, wealth disparity, disease, even optimization of limited resources; AI tools have been built and are in development that will address these previously insurmountable trials. When considering the promise and pitfalls of AI, we ask you to not only think of the current state and the coming months, but consider what the end-state looks like, what does this look like at its conclusion 5-10 years from now?

Honorable Mention
Really, it’s an honor just to be nominated

Important stories from ~ this week that didn’t make it above the fold.

Nvidia released Chat with RTX, LLM software that can be run locally on PCs with Nvidia RTX hardware. The value prop here is privacy and speed—you can feed-in your own documents and files to provide useful context for responses, run entirely on your local machine, thereby addressing privacy concerns from sending sensitive information to third parties. This lines-up with our 2024 thesis of increasing local inference.

Google’s Bard is now Gemini, and they’ve rolled-out more powerful versions of Gemini including various flavors of its Ultra and Pro models. Impressively, some configurations (Pro 1.5) support a context window of one million tokens, one of the longest in the industry.

Air Canada, two words that activate a gag reflex in many Canadians (as do the words “Pearson Airport”) was in the news this week, trying its very best “my dog ate my homework” in a dispute with a customer wherein the company’s chatbot provided erroneous information resulting in financial loss. While Air Canada maintained that the chatbot itself was an entity and responsible for its own actions, the civil tribunal overseeing the complaint disagreed. Two interesting concepts are raised here. It’s a reminder that companies are ultimately responsible for the actions of their employees, AI or other. The second is the argument that AI agents do not posses personhood in the way that a corporation, or an actual person would. We’re not legal experts, but we’re interested to see how these same legal issues are actually interpreted in the court systems of various jurisdictions when they will inevitably be tried.

We’ll close-out with a quick pulse check on Open Source, noting that the community contributed a flurry of additional models and enhancements this week including: a powerful video interpretation model from Meta (V-JEPA) for self-learning AI interacting with the physical world; a large multilingual model from Cohere.ai; Stable Cascade, an image model trainable on consumer hardware; and UC Berkley’s Large World Model, which can rapidly interpret video content up to one million tokens in length. The community remains incredibly healthy and collaborative, as expected.

About the Author: Brodie Woods

With over 18 years of capital markets experience as a publishing equities analyst, an investment banker, a CTO, and an AI Strategist leading North American banks and boutiques, I bring a unique perspective to the AI Geekly. This viewpoint is informed by participation in two decades of capital market cycles from the front lines; publication of in-depth research for institutional audiences based on proprietary financial models; execution of hundreds of M&A and financing transactions; leadership roles in planning, implementing, and maintaining of the tech stack for a broker dealer; and, most recently, heading the AI strategy for the Capital Markets division of the eighth-largest commercial bank in North America.