AI: OpenAI's Strawberry rolls into Open-o1. RTZ #480
...company scaling its AI products and funding
The Bigger Picture, Sunday September 16, 2024
Well, after a lot of speculation and anticipation, OpenAI has unveiled its long-awaited ‘AI Reasoning’ product, dubbing it Open-o1. It’s rolling out in several flavors (see below). And there were updates to its also anticipated funding plans. Both developments have important implications for this this AI Tech Wave going into 2025, so that is the ‘Bigger Picture’ I’d like to focus on today.
I’ve of course long discussed OpenAI’s ‘reasoning’ and ‘agentic’ AI plans, levels 2 and 3 of its 5 level ‘unofficial’ roadmap to AGI (Artificial General Intelligence). Of course the broader industry is also focused on similar goals, under the broader umbrella of ‘AI Agentic workflows’.
The reasoning level had prior code names like ‘Strawberry’ and ‘Q*. But no more.
As I outlined it in Saturday’s Weekly Summary:
“OpenAI’s ‘Reasoning’ Open-o1 Preview/Mini & Funding plans:”
“OpenAI released preview and ‘Mini’ versions of its long-awaited ‘Level 2’ to AGI ‘Reasoning’ AI technology previously dubbed ‘Strawberry’ and Q*. It’s now officially OpenAI-o1, and comes in both a ‘Preview’ and ‘Mini’ versions, the latter being cheaper and faster to run. Initial reactions are generally positive by developers and media.”
“But note that this is not a mainstream product version, and by the company’s guidance, is a way-stop towards its next generation LLM AI models to come with better reasoning and agentic capabilities.”
“Separately, OpenAI seems to be on track to raise $7 billion or more at a $150 billion plus valuation from both financial and sovereign mideast investors. More here on the ‘Strawberry’ context.”
First, let’s discuss OpenAI-o1. Typed with the letter ‘o’ followed by the number 1. To indicate potentially a ‘resetting’ of OpenAI’s current family of products, from GPT-1 through GPT-4o ‘Omni’, with big stops for GPT 3 and 3.5. And many of the latter ones with their own series of sizes, small, medium and larger mostly with different sized model ‘parameters’ small to large, from billions to trillions to date.
And of course there is the highly anticipated GPT-5, in whatever form it comes either this year or next, under monikers ‘Orion’, GPT-NeXT or others, that I’ve written about.
And of course others have their thoughts on OpenAI’s product naming strategies.
The product itself though just introduced, is of course getting a lot of attention from developers and media. And the early reactions are generally positive. As Ars Technica outlines in “OpenAI’s new “reasoning” AI models are here: o1-preview and o1-mini”:
“New o1 language model can solve complex tasks iteratively, count R's in "strawberry."
“OpenAI finally unveiled its rumored "Strawberry" AI language model on Thursday, claiming significant improvements in what it calls "reasoning" and problem-solving capabilities over previous large language models (LLMs). Formally named "OpenAI o1," the model family will initially launch in two forms, o1-preview and o1-mini, available today for ChatGPT Plus and certain API users.”
“OpenAI claims that o1-preview outperforms its predecessor, GPT-4o, on multiple benchmarks, including competitive programming, mathematics, and "scientific reasoning." However, people who have used the model say it does not yet outclass GPT-4o in every metric. Other users have criticized the delay in receiving a response from the model, owing to the multi-step processing occurring behind the scenes before answering a query.”
The basics of the ‘secret sauce’ are emerging:
“OpenAI attributes o1's advancements to a new reinforcement learning (RL) training approach that teaches the model to spend more time "thinking through" problems before responding, similar to how "let's think step-by-step" chain-of-thought prompting can improve outputs in other LLMs. The new process allows o1 to try different strategies and "recognize" its own mistakes.”
OpenAI on its part, from founder/CEO Sam Altman on down, is downplaying expectations of the early versions of the product:
“o1 is still flawed, still limited, and it still seems more impressive on first use than it does after you spend more time with it.”
“In a rare display of public hype-busting, OpenAI product manager Joanne Jang tweeted, "There's a lot of o1 hype on my feed, so I'm worried that it might be setting the wrong expectations. what o1 is: the first reasoning model that shines in really hard tasks, and it'll only get better. (I'm personally psyched about the model's potential & trajectory!) what o1 isn't (yet!): a miracle model that does everything better than previous models. you might be disappointed if this is your expectation for today's launch—but we're working to get there!"
Bottom line, it’s early days for ‘Strawberry’, aka OpenAI o1. WIth a lot of work to be done by both the company and the AI community at large. But promising nevertheless.
(Update: more ‘reasoned’ reviews by both media and industry pundits here also worth reading, including diving into their links. They provide a deeper sense of what it’s like to use the various versions of OpenAI-01, and how it’s still a work in progress).
But as these and other OpenAI products emerge and are used at scale, the company is going to require additional capital to scale the compute required to service the anticipated surge in customers both consumers and businesses.
Which gets us to the financing plans I mentioned above. Again, as Bloomberg outlines in “OpenAI Fundraising Set to Vault Startup’s Valuation to $150 Billion”:
“Company discussing raising $6.5 billion in equity financing. Banks in talks to provide startup with $5 billion credit line.”
“OpenAI is in talks to raise $6.5 billion from investors at a valuation of $150 billion, according to people familiar with the situation.”
“The new valuation, a figure that doesn’t include the money being raised, is significantly higher than the $86 billion valuation from the company’s tender offer earlier this year, and cements its place as one of the most valuable startups in the world.”
“At the same time, OpenAI is also in talks to raise $5 billion in debt from banks in the form of a revolving credit facility, said one of the people, all of whom asked not to be identified discussing private information.”
“The funding round is slated to be led by Thrive Capital, Bloomberg previously reported. Microsoft Corp., the company’s largest investor, is also set to participate, and Apple Inc. and Nvidia Corp., have been in talks about investing.”
And the funding is also being done while the company modifies its Corporate structure and Governance, following last November’s OpenAI Board events.
As Reuters explains in “OpenAI's huge valuation hinges on upending corporate structure”:
“OpenAI's new financing round is expected to come in the form of convertible notes, according to sources with direct knowledge of the matter, who said its $150 billion valuation will be contingent on whether the ChatGPT-maker can upend its corporate structure and remove a profit cap for investors.”
“The details of the conditions of the $6.5 billion funding, which have not been previously reported, show how far OpenAI, the most valuable AI startup in the world, has come from a research-based non-profit, and the structural changes it's willing to make to attract ever more investment to fund its expensive pursuit of artificial general intelligence (AGI), or AI that surpasses human intelligence.”
“The outsized funding round has seen strong investor demand and could be finalized in the next two weeks, given the rapid growth of OpenAI's revenue, sources added.”
There are a lot of details that point to twists and turns in the funding plans that are worth reading.
It’s all for a range of opportunities for OpenAI, not only with Microsoft, but Apple, Nvidia and potentially others over time, to build leadership AI products and services up and down the AI Tech Stack below.
But the bottom line is that OpenAI is preparing itself for a very busy 2025 and beyond. Both operationally leading on a number if core new AI products and services, AND securing the first in what it likely to be a series of funding efforts to execute on those opportunities. Big execution hills to climb, one after another.
As I’ve outlined before, every major AI company is focused on expending tens of billions in AI capex to fund the ‘Table Stakes’ Scaling of AI to what it eventually possible. And the financial rewards may not be in sync with those expenditures.
As the leading private company in this AI Tech Wave, partnered closely with Microsoft, OpenAI has to execute on its parallel building and funding opportunities, with all the challenges they entail. As all their peers also execute on the same opportunities around them.
That is the ‘Bigger Picture’ to keep in mind as we finish 2024. Stay tuned.
(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here)