Bing AI: how does it monetize?
Bing AI: how does it monetize?
Generative AI
Microsoft invested in an exclusive deal to use OpenAI’s ChatGPT model within their Bing app, which they have named Sydney.
Business Models
B:C Advertising: The company is considering in-chatbot advertising suggestions after delivering response queries.
B:B Per-API-Call: Microsoft is also charging per-API call fees for developers working with Bing search, images, video, and text completion, improving these offerings with ChatGPT is in the plans.
Funding
The partnership between Microsoft and OpenAI is unique, with Microsoft originally investing $3 MM with additional investors, and then recently $10 MM for exclusive rights. The unique hybrid organizational structure limits potential revenue upside for investors and eventually will recycle future profits to the OpenAI foundation.
Mission
Bing AI’s Mission: “unlock the joy of discovery.”
Open AI’s mission: “ensure that artificial general intelligence (AGI) benefits all of humanity, primarily by building safe AGI and sharing the benefits with the world.”
Model Review
Is it capitalism or generative AI that we fear?
In a recent opinion piece for the NY Times, Ezra Klein posed an excellent question.
Business models are what Ezra urges us to examine.
Since I just got access to Bing’s AI today, I wanted to see what it could do.
I will use my trusty Contribution Model Adlib format to examine the key variables of the technology value proposition and promised contribution to the world, as well as ask: who has skin in the game?
This simple structure has saved wasted hours and also the reputations of several folks who have used it to tell if they should invest in Web3 investments (No!) launch tokens (No!) Sign their institution to support Facebook Libra (No!) or invest in inclusive data practices and model explainability (Yes! This is not hype).
Bing AI a sloppy assistant:
Here is how Bing AI promises to be better: annotated links to websites that back up factual information.
In order to fact check the facts that Bing offers in chat responses, I had to toggle between the awkward Bing app and the actual internet search via Google. When this AI bot gives annotated references, it creates even greater cognitive dissonance when the links do not match the evidence. Bing often led me down a rabbit hole of tech industry gossip and unverified conjecture, often off topic for the “facts” that were cited.
Bing AI is intended to provide “the joy of discovery” – which gets four stars for aspiration, but a two for delivery. I didn’t address Bing AI by the company’s preferred name, Sydney, after the creepy NY Times review, perhaps because I have a pet peeve about personifying chatbots. Additionally, I was offered the change for tuning to “creative” vs. focus mode, but Bings bright icons of tantalizing celebrity news was dangerously distracting, so I opted for focus. Possibly as a result, the Bing AI chatbot was all business with me. I blame myself.
Who will pay?
Microsoft is planning on selling ads to Bing’s advertisers and presenting these directly in the chatbot’s answer results. Since the launch of this public shiny launch, app downloads and advertiser accounts have been rapidly revived.
How will it work? Bing AI provides a set of annotated links for how advertising might appear:
- A recipe for chocolate cake with links for where to buy the ingredients.
- A list of potential vacation spots with links to destinations.
- A list of suggested party venues for your mother’s birthday
Nonetheless, we know perfectly well from our time invested in Facebook/Meta and Twitter that the business model will be unsustainable if the entire model is based on advertiser answers at the expense of user queries. User and advertiser value propositions will be see-saws of diminished expectations after the initial hype wears off.
At the business model analysis level of the value proposition, that’s a tricky question. As a matter of fact, we should shift our analysis from value propositions to revenue to focus on ownership and funding.
Microsoft owns a future minority stake in a novel type of hybrid enterprise.
With early donors like Reid Hoffman, Peter Thiel, and Elon Musk (Musk is no longer involved), OpenAI began as a nonprofit with a mission to “ensure that artificial general intelligence (AGI) benefits all of humanity, primarily by building safe AGI and sharing the benefits with the world.”
In order to raise capital, Altman reorganized the organization as a hybrid of multiple organizations with additional funders and Microsoft investing $3 billion. Eyeing opportunities for Bing and business services, Microsoft recently agreed to an additional $10 billion with exclusive rights to commercialize ChatGPT.
To understand ownership and payback terms, a Sankey diagram is needed.
From @FortuneMagazine / @jeremyakahn pic.twitter.com/dWdFaCxiVZ
— Pete (@nonmayorpete) January 11, 2023
After Microsoft invests, they have to first wait for the first investing partners to get 100% of their profits paid back, then Microsoft takes 75% of the profits, with 25% going back to employees and early investors with the profit cap detailed. Once that happens 49% of the profits go to Microsoft, 41% to OpenAI employees, 8% to the original investors, and 2% gets recycled back into the OpenAI non profit. Microsoft, in the long run, will never own a majority share of the company.
Using an external partner to propel you faster than the dominant fierce incumbent is considered an “adjacent innovation” in strategic portfolio innovation to “revive a tired brand”. Smooth moves.
For Silicon Valley types afraid of both activist public markets investors and the slow rot of corporate exits, OpenAI gets creative points for trying a new approach.
Our search data, which is subject to Microsoft’s Privacy Policy and the policies of the countries where we live, is modeled with OpenAI’s GPT Models with the Bing Index and Ranking algorithm.
A large language model post wouldn’t be complete without a chatbot answer, so I asked: whose data is being used:
The Bing AI bot tells me that’s it is trained on vast amounts of info on the internet, and that it’s hard to draw a direct link between that training data and its output – but factual data will often be annotated. A fuzzy answer to a question that should be more on our minds – who owns the data we contribute?
In response to further questions about data and company ownership, the Bing AI provided oversimplified and incorrect answers, pointing to that scary Sydney story, without breaking their character to pledge love or attempt to dominate.
I give Bing 1 star for ethics. While data ownership might remain fuzzy as we work out the data sourcing question in large language models, we should know who owns the results of all of our voluntary contributions. These answers need to be crystal clear on day one, even if the answers to the queries are going to be hallucinating for some time.
But let’s move to where Microsoft excels (pun! Totally intended!).
The business-to-business (B:B) market
Here Microsoft reveals their familiar B:B ecosystem playbook
As part of their current Bing developer APIs, Microsoft plans to provide ChatGPT services to developers, enabling them to use Bing’s image, text, video, auto-correct, and spelling tools. power their app and software experiences. In prep for the future value they feel they’ve already created using AI in these offerings, they’ve increased their per-API-call prices 3X-10X (likely expecting a surge in use. As the Bing developer program has grown exponentially, Microsoft is thrilled.
Here Microsoft reveals its familiar B:B ecosystem playbook
As part of their current Bing developer APIs, Microsoft plans to provide ChatGPT services to developers, enabling them to use Bing’s image, text, video, auto-correct, and spelling tools. power their app and software experiences. In prep for the future value they feel they’ve already created using AI in these offerings, they’ve increased their per-API-call prices 3X-10X (likely expecting a surge in use. As the Bing developer program has grown exponentially, Microsoft is thrilled.
The key question- will they generate enough developer demand to pass the Bill Gates Line?
When critiquing Facebook’s leadership team in its early days, Bill Gates defined what it meant to lead a productive ecosystem. Gates did not think that Facebook was a platform ecosystem, and said so using salty language, then stated:
“A platform is when the economic value of everybody that uses it, exceeds the value of the company that creates it.”
In tech strategy circles, this has come to be known as the “Bill Gates Line” (credit: Stratechery). Windows execs used to brag about the fact that they were a minority of the total platform ecosystem. Since Azure Cloud Services has worked hard to migrate its channel partners and stay strong, it has challenged Amazon Web Services and Google Cloud due to its view that they should only capture a minority of ecosystem value. Microsoft can challenge Google Search if it focuses on the pain points of developers who feel underserved by it, just as it did in cloud computing.
By integrating ChatGPT-powered data-driven models with other Microsoft Azure offerings, Microsoft will have masterfully transitioned to complement their Platform-as-a-Service and Infrastructure-as-a-Service revenue models, strengthening their uptake by business app developers.
As a result of these clever B:B moves, the total business model rating has risen to 3. Microsoft and OpenAI will likely be mitigated in our daily search and discovery routines until the system can better explain ownership and validate its data sources.
—
This Contribution Model Map summarizes the various flows and players – who contributes, who pays, who owns, and who is along for the ride. If you are trying to determine whether or when to jump on the next tech hype cycle, this level of analysis can help.
Should you jump at the chance to develop with Bing AI tools if you are a developer? Trying new tools is always a good idea. In every case. This is especially true when you are operating in a category dominated by a monopoly. Knowing that you are playing with new tools will give Google a run for its money.
But it can’t help you figure out if the technology should be condemned, or if Advertising or B:B API calls will wind up accelerating the biases and slanted rabbit holes we tend to follow when we spend our time in digital spaces.
It’s not business models that hurt people. Unintended consequences happen as the result of decisions and actions made within governance structures. Who gets to decide what kind of content is generated, and who benefits from it? How do we ensure that the data used to train these machines is diverse and representative? Yes, the novel hybrid structure keeps OpenAI out of the harsh profit-maximizing glare of the public markets and impatient expectations of investors looking for an exit.
But will ownership change decisions about how data is governed and stewarded?
It’s up to the rest of us to decide whether or not we want to invest time in training the data. At this point, it’s not clear whether your contributions are personally or communally beneficial. Let’s wait for the owners of these systems to figure out how to make these lofty goals a reality – “ensuring artificial general intelligence (AGI) unlocks joy for everyone.”
Reach out if you have anything to correct or add.
Bing AI results were cited in this post, which was written by humans.
