Introducing the AI Geekly, by Brodie Woods. Your curated 5-minute read; everything you need to keep up to date with the latest developments in the rapidly evolving world of AI.

In a world clouded by clickbait journalism, bombastic hype, and legacy pundits, it can be hard to know the real story. Leveraging nearly two decades of experience on the front lines of capital markets at North American banks and broker-dealers, the AI Geekly conveys the informed perspective of an AI strategist and capital markets veteran.

Cutting through biases of competing narratives across social, economic, scientific, and political streams, we provide readers with a clearer picture of the reality playing out.

As we enter the Age of AI, separating signal from noise is vital.

Let me be your guide:

AI Quote of the Week

“First, second, and third priority are around AI, AI, AI… Over the last thirty days, what we’ve seen is a continued acceleration of engagements… [the market is] skyrocketing.”

Lisa Su, Chair and CEO of AMD Technologies

TL;DR - Exec Summary

AI-generated videos improving; OpenAI sued again… again again…; IRS AI Taxbot; Google’s $20 mm Ethics Fund; Falcon 180B lands, Nvidia releases acceleration tool

Pace of AI Development Continues to Accelerate: Over the past three weeks writing the AI Geekly for a wider audience, we continue to see a rapid, almost frantic pace of improvement in AI. This is perhaps best exemplified by the release of yet another trailblazing Large Language Model (LLM), the recently unveiled Falcon 180B; produced only three months after releasing its predecessor Falcon 40B. Not only is Falcon 180B 4x larger, but it has the performance to go along: unlike the prior 40B model, Falcon 180B represents serious competition to Google’s PaLM 2 while trouncing Meta’s Llama 2 and approaching OpenAI’s GPT-4 model (underpinning ChatGPT).

Major Investments Driving Growth: The break-neck pace of development with Falcon is emblematic of the overall rate of improvement in the field of Generative AI (GenAI). What stands-out in particular is the velocity and scale of investment —Falcon 180B likely cost seven-to-eight figures to train on AWS’ SageMaker, (barring significant discounts or efficiency gains from something like Databricks’ MosaicML) and they knocked it out in three months. Incredible.

The 600 Pound Gorilla —Legal Challenges: Tamping that enthusiasm, the ever-present thorny legal problem driving GenAI lawsuits: the questionable legality of training these models on data without permission of the IP rightsholder. As Helen of Troy’s face launched 1,000 ships, it seems Sam Altman’s (CEO of OpenAI)’s face has instead launched 1,000 legal battles. There is real fear among potential customers of GenAI companies that these lawsuits may target enterprise customers. This is likely driving recent moves by GenAI vendors to indemnify customers in very narrow situations related to training data IP-lawsuits. The other option I would put forward for readers is to train their own LLMs on data for which they have necessary ownership rights —it avoids the whole issue.

IRS Taxbot Promises it Won’t Come after us: Perhaps tamping enthusiasm even more… the IRS’ plans to begin employing AI to seek tax evaders. Initially it plans to focus on hedge funds, private equity and law firms, which it alleges have complicated structures that could hide tax obligations. Realistically this sounds like a pilot targeting three groups unlikely to receive any public outcry. There should, however, be concerns around usage creep, should the agency consider expanding its use and scale to surveil.

This is Still the First Inning: I often get the question where we are in the development curve of Generative AI. The pace of development and adoption of these capabilities and the rate at which both are accelerating points to just the beginning. This hasn’t even started yet…

Lights, Camera AI-ction!
GenAI video generation improving week-over-week

What it is: A new tool allowing very simple video modification based on prompts, as highlighted in the embedded Tokenflow video.

What it means: We’ve seen how quickly capabilities have improved with GenAI tools for composing written, audio, and image content in the span of a few months. AI-generated videos look today like image generation looked 11 months ago —expect this to get a lot better, very quickly.

Why it matters: Video generation is yet another powerful AI tool in the toolbox of AI creators. Applications are wide-ranging, whether advertising, recreational, or even as an equalizer, leveling the playing field for smaller businesses without large marketing budgets and teams.

Click here to learn more

From OpenAI to Open Litigation
Prominent authors file lawsuit over ChatGPT's training

What it is: A group of writers, including Michael Chabon, David Henry Hwang, Rachel Louise Snyder, and Ayelet Waldman, are suing OpenAI. They allege the company unlawfully utilized their copyrighted works to train its AI models.

What it means: Plaintiffs point to ChatGPT's ability to summarize and analyze content from their works as evidence of OpenAI's use of their IP for training. These outputs, they argue, are "derivative works" under copyright law and infringe on the owner’s rights.

Why it matters: OpenAI isn't new to litigation. Ultimately, the status of AI-generated works will be decided by the courts. Enterprise clients are cautious about perceived legal risk from vendor LLMs. This explains recent reports that vendors including Microsoft have been indemnifying clients re: training data.

Click here to learn more

AI: The IRS's New Bloodhound
Sniffing Out Financial Shenanigans with Silicon Precision

What it is: Not one to leave money on the table, the IRS is now deploying AI to tighten the noose on tax evasion, particularly targeting hedge funds, private equity firms, and law firms. The move targets the intricate financial structures these entities often employ.

What it means: The IRS's foray into AI signifies an acknowledgment of the limitations of traditional auditing methods and a willingness to apply modern tools. Some investigative tasks, previously impossible due to combined scale, complexity and monotony become achievable with relative ease using AI.

Why it matters: While the IRS' stated intentions may well be to ensure everyone pays their fair share, the broader implication is the increasing encroachment of AI into areas of personal finance and privacy. As the tools get sharper, so does the debate on the balance between oversight and overreach.

Click here to learn more

Google’s “Do No Evil” in 2023
Investing $20 mm in ethical AI fund

What it is: Google is dedicating $20 million to champion responsible AI development, signaling its commitment to not only lead in technological innovation but also set the ethical tone for industry.

What it means: Beyond the immediate financial support, this fund represents Google's strategic foresight. The tech giant recognizes the increasing importance of ethical considerations in AI's rapid evolution (so-called Responsible AI). Selectively backing projects and research, Google is positioning itself to influence the narrative and direction of AI's ethical standards.

Why it matters: As AI becomes more integrated into daily life and business operations, its ethical framework must be scrutinized. By the same token, it’s critical to evaluate the objectives of incumbent tech co’s seeking to influence these outcomes. Benevolence on the surface can often obfuscate ulterior motives…

Click here to learn more

Falcon Models Soar Higher and Higher
UAE’s TII Releases Larger Falcon 180B Commercial Model

What it is: The UAE’s Technology Innovation Institute (TII) this week launched Falcon 180B —a scaled-up version of its already venerable Falcon 40B model released just this June.

What it means: In barely three months, TII has “lapped” its prior model, deploying massive amounts of compute (4,096 GPUs / 7 mm GPU hours) to train Falcon 180B on 3.5 trillion tokens. The result? A model that not only surpasses Meta’s fresh-out-of-the-oven Llama 2 model (granted it’s 2.5x the size), but rivals Google’s venerable PaLM 2 model.

Why it matters: Falcon 180B performs considerably better than its predecessor, and most available models for that matter, in natural language tasks. By a lot. Groundbreaking models are being released regularl, and can be trained quickly. We are in the early innings of the Generative AI story.

One more thought: The UAE and Saudi Arabia, in particular, have been investing heavily in AI as part of a strategy of diversification and future proofing. Pay attention to the actions of state actors in coming months. AI is starting to become a part of the geopolitical conversation. We expect bolder moves as nations and private interests become increasingly aggressive on AI.

Click here to learn more

Making the Most of a Great Situation
Nvidia uses H100 chips as leverage to expand business

What it is: Nvidia, primarily a hardware player (GPUs and AI) is making strategic moves into the cloud space. In 2022, Nvidia proposed a cloud leasing structure that would allow it to white label infrastructure of cloud giants like Google and Amazon’s whereby it would lease Nvidia-powered servers within the cloud providers' data centers to customers.

What it means: This shift places Nvidia in direct competition with traditional cloud providers, many of which are significant buyers of Nvidia silicon. It is noteworthy that these same companies do try to compete with Nvidia on the hardware side as well, with proprietary chip designs of their own (though less performant).

Why it matters: Microsoft, Google, and Oracle have reportedly agreed to Nvidia's proposal, AWS has declined, marking a potential shift in the cloud services landscape. It is unclear how these decisions will affect delivery of Nvidia’s H100 chips.

Click here to learn more

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.

Glossary

Terms:

• LLM (Large Language Model): A type of artificial intelligence model designed to understand and generate human-like text based on the data it's trained on.

• PaLM 2: Google’s proprietary LLM that underpins its Bard chatbot as well as its commercial Vertex and Duet AI offerings

• Llama 2: Meta’s open access LLM, released a couple of weeks ago, and prior to Falcon 180B viewed as the most performant open model.

• GPU (Graphics Processing Unit): A specialized electronic circuit designed to accelerate the processing of images and videos for computer graphics. In the AI world, GPUs are also used for training complex models due to their parallel processing capabilities.

• GenAI (Generative AI): A subset of AI that focuses on generating new content, be it text, images, videos, code, or other forms of media.

• IP (Intellectual Property): Legal rights that result from intellectual activity in the industrial, scientific, literary, and artistic fields. In the context of the Geekly, it refers to the rights of content creators.

Entities

• Nvidia: An American tech company known for their GPU product lines, which are high-performance hardware used in computers for graphics and, in recent years, AI tasks.

• AMD (Advanced Micro Devices): An American tech company that develops CPUs, GPUs, and APUs for business and consumer markets. most recently focusing on AI. It’s major competitors are Intel on CPUs and Nvidia on GPUs.

• UAE’s Technology Innovation Institute (TII): An applied research pillar of the Advanced Technology Research Council (ATRC), TII aims to drive applied research and innovation to advance breakthrough technologies.

• AWS (Amazon Web Services): Amazon's cloud computing platform that provides a mix of infrastructure as a service (IaaS), platform as a service (PaaS), and packaged software as a service (SaaS) offerings.

• Databricks’ MosaicML: A platform that provides tools and infrastructure to train machine learning models more efficiently.

• OpenAI: An organization committed to ensuring that artificial general intelligence (AGI) benefits all of humanity. They are known for models like GPT-3 and ChatGPT.

• Meta: Formerly known as Facebook, Inc., Meta is an American multinational technology conglomerate that is responsible for the released of some of the most performant open access AI models to date: Llama and Llama 2.

• Google: An American multinational technology company specializing in Internet-related services and products, which include online advertising technologies, a search engine, cloud computing, software, and hardware. Largely responsible for modern AI.

Key People

• Dr. Lisa Su, Chair and CEO of AMD Technologies: Under her leadership, AMD has seen significant growth and innovation in the semiconductor space, delivering an assortment of GPUs, CPUs, (and APUs) along with AI-focused hardware more recently.

• Sam Altman, CEO of OpenAI: While a household name due to his leadership of OpenAI, Sam has been a prominent figure in the tech industry, previously leading Y Combinator, a startup accelerator.