Welcome back to the AI Geekly, by Brodie Woods. Your curated 5-minute read on the latest developments in the rapidly evolving world of AI.
Cutting through clickbait, hype, and old media, the AI Geekly provides the perspective of an AI strategist and capital markets veteran, informed by nearly two decades on the front lines.
As we enter the Age of AI, staying informed and critical is key.
AI Quote of the Week
“Activist work, local, national, international government, etc. —it’s all just slow and inefficient and fallible.
Imagine if you didn’t have human fallibility. I think it’s possible to build AIs that truly reflect our best collective selves and will ultimately make better trade-offs, more consistently and more fairly, on our behalf.”
Google DeepMind co-founder Mustafa Suleyman
TL;DR - Exec Summary
Google shares more AI love; Tech/Gov/Special Interest triple-threat targets open source AI; Music Making for the Masses; Landmark AI Copyright Ruling; ARM’s IPO
Realistically I can’t start every one of these by saying “it’s been a busy week in AI”, so we’ll just treat it as read.
Everybody’s talking their own book: Major Tech and AI companies signed a White House pact and attended a closed-door AI summit along with unions, and activist groups -achieving consensus that regulation is needed to protect against AI's potential societal harms. However, their motives may be self-serving —each group stands to benefit financially from restricting AI to a privileged few, effectively banning the open source AI community. We contend that fairness and risk reduction can be achieved by democratizing access to AI tools instead of restricting it.
Meanwhile, the US Copyright Office Review Board's decision on AI artistry was pretty much what we expected, maintaining status quo re: machines producing IP. ARM's triumphant IPO shines a light on investor sentiments and the perceived future of AI ventures. Google's been sharing more AI love with the release of Bard Extensions (integrating AI with Gmail, Drive, and more) as well as its benevolent open sourcing of a massive language and translation model trove. With 419 included languages, it will undoubtedly build bridges. Lastly, Stability AI's venture into audio generation through Stable Audio hints at the still untapped potential of AI in multimedia creation.
-Read on for the full story
Bard’s New Tune: AI Integration with Google Services
Google ties Bard AI into its suite of consumer products
What is: Google integrated its Palm 2-powered Bard AI agent (Google’s ChatGPT) with various Google apps and services via “Bard Extensions”, allowing Bard to leverage relevant information for AI workflows from across a user’s Google data ecosystem via Gmail, Docs, Drive, and more. Critically, it includes a feature to sanity-check Bard’s statements and facts for truthfulness, reducing the impact of “hallucinations”.
What it means: Google is finally bridging Bard with its extensive ecosystem, presenting a holistic, AI-driven tool with the potential to enhance productivity and information retrieval. Not only does this allow for potentially novel user experiences, it also gives Google access to a much larger Reinforcement Learning through Human Feedback (RLHF) pool via the millions of users who will be interacting with the new Bard Extensions (much of ChatGPT’s success is credited to similar functionality)
Why it matters: Google has “damned the torpedoes” and is throwing everything it has at dethroning OpenAI. This integration marks a significant step forward in AI-assisted productivity tools. Smarter, tighter integration with users’ troves of cloud and locally stored data is a major advantage. The first AI company to allow customers (enterprise, or consumer) to leverage the untapped value of decades of relevant digital data may just win the AI arms race. Microsoft has made a similar promise with its Copilot offering, but it has yet to be made available more broadly.
“Trust us, it will be fair this time.”
Tech, government, special interests’ regulation plans conveniently line pockets and block open source AI
What it is: Lots of action from Washington this week:
On Wednesday, Senator Chuck Schumer convened a closed doors “AI Insight Forum” attended by CEOs of the major AI companies.
Several AI players also signed-on to the White House’s AI pledge, voluntary standards on pre-release safety testing AI, cyber security, AI transparency, and consideration of societal impacts.
What it means: Based on reports, little was accomplished in the forum. Regulatory themes were discussed, notably with Meta calling for collaboration, in-line with their quasi-open-source approach to AI models. Interestingly, the AI pledge seems to prohibit open source models, with specific limitations on release of AI models and model weights.
Why it matters: There is a storm brewing in the world of AI. The desires of many to regulate, limit and control AI is in direct diametrical opposition to the views of the open source community which seeks to democratize AI and its development, making it open to all. Draconian control is the main tool available to the first group to deliver fairness. For the second, it is the wide dissemination of these AI tools to anyone and everyone that ensures that the playing field is equal.
One more thought: As a society, we must decide between two options: 1) trust the management teams of the largest, wealthiest corporations in the world, our elected officials, and our unelected activists to mediate Private AI, deciding who can use it and for what approved purposes they deem appropriate; or 2) support a spirit of transparency, collaboration and innovation through the wide distribution and education around these tools ensuring they are kept not only in the hands of a privileged few.
So… CAPTCHA on US Copyright Applications?
US Copyright Office Review Board's stance reignites debates over AI, creativity, and ownership
What it is: The US Copyright Office Review Board rejected copyright for a piece of AI-generated art made famous last year by winning a Colorado State Fair art contest. The question at hand is whether the creations of Generative AI can have intellectual property rights if produced by machines, or AI, rather than humans.
Why it matters: The decision is in keeping with prior decisions, so unsurprising in that respect. That said, while seemingly isolated, the decision is a microcosm of broader challenges. As AI becomes an integral part of various creative processes, we're compelled to revisit the treatment of intellectual property, value attribution, and creation.
What it means: We now have refreshed clarity regarding how the US Copyright Office views Generative AI works. While positive to have a degree of certainty about something in the AI Age, many legal questions remain unresolved. Thorny issues include IP and copyright for training data, outputs as derivative works, the legal liability of a company that simply uses a non-kosher model vs. one that builds and sells it, etc. To mitigate that latter risk, we recommend companies train their own models.
A Shot in the ARM
ARM's emphatic IPO shows ‘em how it’s done
What it is: ARM's market re-entry via spin-out from Softbank, was marked by an impressive ~25% stock surge at points before settling down to roughly the issue price. It underscores the value of hardware players in this tech cycle vs. prior, largely dominated by SaaS and Cloud.
Why it matters: Amidst a volatile market landscape, punctuated by geopolitics and fluctuating interest rates, ARM's success story offers a glimmer of optimism. For those unfamiliar with ARM, their chip architecture is the dominant architecture in mobile and edge computing. Notably, Apple uses it for its silicon —the iPhone A17 chip and the Macbook’s M2 chip both use ARM architecture with custom Apple designs.
What it means: ARM's performance shows a level of investor optimism that is a fair degree better than what doomsayers had predicted. This positive data point may support a positive trajectory of the right kind of tech IPOs, as well as market sentiment. A portion of the proceeds will be used by Softbank to fund an AI “deal spree”. Market loves a Softbank deal…
Truly the Mad Lads
Google gives world a translation alley-oop with MADLAD-400
What it is: Google's MADLAD-400 addresses a longstanding challenge in Natural Language Processing (NLP) - the need for high-quality training data across a wide spectrum of languages. While major languages like English have seen rapid advancements due to vast datasets, less commonly spoken languages have often been left behind. MADLAD-400, crafted meticulously by a dedicated research team, seeks to fill this gap.
Why it matters: Limitations due to the constraints of web-scraped data, especially for languages with limited online content, noise, inaccuracies, and varying content quality often posed challenges. MADLAD-400, with its rigorous auditing in 419 languages and a hands-on approach to data quality makes a major difference. As we experiment with differing permutations of model configurations it has become clear that model performance is not only impacted by size, but training data quality.
What it means: It’s always appreciated when the major players are willing to share some of their discoveries and make them broadly available. This is the positive spirit and comradery that has defined a large swath of the global mind-trust / academic community dispersed across the various tech co’s and start-ups (everyone has bills to pay). Fitting then that in this case the result with be improved communication and collaboration across many languages.
Making Silicon Sing, 50 Years After the Synthesizer
Generative AI ventures into audio, now everyone’s a musician! (but not really)
What it is: Stability AI, best known for its groundbreaking open-access generative AI technology for image generation, Stable Diffusion, is making the move to audio with Stable Audio. The tool allows users to produce brief audio clips using simple text prompts, borrowing the modality and some of the AI techniques of its kindred Stable Diffusion.
Why it matters: While there are a handful of GenAI tools that do text-to-audio, Stable Audio offers a novel approach to generating sound of various types including music, sound effects and instruments based on descriptive text. Generation of Audio, Text, and Video will continue to improve rapidly over the coming months, creating a convergence of each medium that is indistinguishable from reality.
What it means: For content creators, musicians, and businesses, Stable Audio could become an invaluable tool, transforming textual descriptions into auditory experiences. While the technology is in its nascent stages, its potential applications range from music production to advertising. The general availability and modest pricing of the tool creates a level playing field, democratizing the availability of this powerful tool: Anyone can create music and audio with AI.
That's it for this week! —Feel free to share the AI Geekly with friends, colleagues, family, etc.
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.
Bard AI: Google's AI agent integrated into its suite of consumer products, allowing for enhanced AI-driven user experiences.
Bard Extensions: Integrations that tie Bard AI into various Google apps and services.
Google Docs: A word processing product from Google, part of the Google Drive office suite.
Google Drive: Google's Cloud file storage and synchronization service.
Reinforcement Learning through Human Feedback (RLHF): A training approach where the AI model learns by getting feedback from human interactions.
"Damn the torpedoes": An expression meaning to proceed full speed ahead without regard or caution. It originates from an order given by Admiral David Farragut during the Battle of Mobile Bay in the American Civil War.
Generative AI: A subset of AI that focuses on generating new content, whether it's text, images, videos, music, or other forms of media.
Open Source AI: AI models and tools that are freely available for the public, allowing a community-driven approach to AI development.
Intellectual Property (IP): Legal rights resulting from intellectual activity in the industrial, scientific, literary, and artistic fields.
CAPTCHA: An acronym for "Completely Automated Public Turing test to tell Computers and Humans Apart." It's a type of challenge-response test used to determine whether the user is human or not.
IPO (Initial Public Offering): The process by which a private company offers its stock to the public for the first time.
Apple A17 chip: A specific chip designed by Apple, part of their line of mobile device chips.
Apple M2 chip: Another chip designed by Apple, typically used in their line of Macbook computers.
Natural Language Processing (NLP): A field of AI that focuses on the interaction between computers and humans through natural language, enabling computers to process and analyze large amounts of natural language data.
Stable Diffusion: Generative AI technology for image generation developed by Stability AI.
Google DeepMind: A British artificial intelligence company that is a subsidiary of Alphabet Inc (Google's parent company).
Meta: Formerly known as Facebook, Inc., an American multinational technology conglomerate.
US Copyright Office Review Board: The governing body that decides on matters related to copyright in the US, including issues like AI-generated art.
ARM: A global semiconductor and software design company known for its ARM architecture, widely used in mobile and edge computing.
Softbank: A multinational conglomerate holding company headquartered in Tokyo, Japan. Known for its investments in the technology sector.
Stability AI: A small AI company known for its generative technology for image generation and now venturing into audio generation.
Mustafa Suleyman: Co-founder of Google DeepMind the groundbreaking AI research lab acquired by Google. Renowned for his expertise in the field of artificial intelligence, he often provides insights and commentary on AI's transformative potential and challenges.
Senator Chuck Schumer: A U.S. Senator representing New York. Throughout his tenure, he has been involved in various legislative matters, ranging from infrastructure to healthcare. He has also shown an interest in technology policy, particularly as it relates to the challenges and implications of artificial intelligence.