OpenAI's Developer Keynote: What CMOs Need to Know (and Do)

Factoring AI into your 2023 plans

CMOs don’t usually pay attention to developer conferences. But OpenAI might change all that. Yesterday at their developer conference, CEO Sam Altman said 2 million developers are building on their API, 92% of Fortune 500s are using its products, and 100M people are using ChatGPT weekly! That’s enough impact for all of us to care about what new features and capabilities they’re launching. 

OpenAI is defining the AI market and the future of tech. And part of that future is empowering non-developers to create on their platform. 

THE NEWS 

A bunch of news sites covered the totality of things announced at the OpenAI conference yesterday. Things that stood out to me as a marketing leader included:  

ChatGPT (Updated with GPT-4Turbo Capabilities)  

• Lesser Time Gap — When the model first came out, it was trained on data up to September of 2021, a time warp that gave me pause in using and trusting it fully. OpenAI has been closing the gap and yesterday announced that the new GPT-4Turbo model is trained on data through April 2023. That’s pretty up-to-date. And Altman hinted that it may get even more current mores as time passes. 

• They added the ability to add more context - you can now add the equivalent of up to 300 pages of a standard book text to give context to a request (Shoot, I need to collect much more research to inform my prompts!) 

• The user experience is improving - ChatGPT can now browse the web, write and run code, or analyze data without a user needing to navigate between its (previously separate) models. It will know which model to use automatically. 

• Altman emphasized data privacy for enterprise clients - Altman reinforced that they do not train the models with data from APIs or Enterprise deployments - a big concern of anyone who might consider training a model with their private data. Data and insights privacy is a huge issue for many companies and was previously a boon for companies with more private large language models.

• Copyright infringement concerns addressed — OpenAI launched Copyright Shield, offering to defend customers of ChatGPT Enterprise and API users if they faced legal claims around copyright infringement (I still think this is a tricky place for marketers, morally and legally.)

Voice-Activated 

• (Launched in the past few months) You can interact with Chat GPT verbally as well as in text 

• GPT-4 Turbo has six default voices that can speak a wide variety of languages 

Build your own GPT  

• Part of the keynote news was that you can now create different GPTs using common language, not code, to generate models with more expertise in a particular area or field. These custom GPT agents can be trained to help with special use cases (Altman created a custom GPT agent to advise entrepreneurs and grill them on why they’re not growing faster. He demonstrated one built by Canva where you can design by asking for what you want in natural language - interact to get one you like and then click through to the full Canva experience.)

• OpenAI is launching a marketplace where creators can sell their custom GPT Agents shortly. 

  

Create your own GPT AI Assistant  

• You can now make your own “GPT” that takes the capabilities of GPT4-Turbo for something.

• OpenAI demonstrated how the Chat GPT-4 Turbo APIs could be used to make calls in and out of applications to have an AI Assistant interacting with the app data and user answering questions, synthesizing information, and visualizing all sorts of things.

Enterprise Custom Models 

• They are launching custom models where they work with customers to create a model with a completely new knowledge domain or create one with a a lot of proprietary data. Their researchers will work closely with a company to make a custom model using OpenAI tools.

• They won’t be able to do it with many companies, and it expect it to have a sizeable cost and potential wait time

What does it mean for a CMO?  

As a CMO, we have two areas of concern as to how we embrace AI faster than our competitors: 

1. In our own products and services 

2. Making our internal teams more efficient and effective 

While more than 70% of software companies are launching AI features this year, I recently heard May Habib, CEO of Writer, mention that it’s harder for companies to get their own employees to use AI than it is to roll it out in their own products! Change is hard; everyone is busy. Learning new things and creatively adapting them is tricky. 

Five things are top of mind for me:  

1. Chief “Market” Officers Must Understand to Strategize - One of the highest impact areas for marketing leaders is to help their company define and lead their market STRATEGY, not just the execution of marketing activities and programs. The rapid pace of change in AI, and the change from being highly technical to based on natural language means we can and must stay up-to-date on what is possible in OpenAI’s market-leading engine. Then, we can help inform product decisions and trade-offs to stay ahead in our markets. The keynote was an hour and very consumable. I highly recommend you watch it to get your wheels turning. Watch it here.
   

2. Our Web Sites Will Soon Need AI Assistants - Before I talked to any CEO, I used to spend a long time reading about their company on their website. And then almost every CEO said their website doesn’t reflect their vision, and isn’t up to date! I have a big sense of urgency for how AI will impact marketing departments’ already squeezed website resources. The AI assistants demoed today by OpenAI will be tools we’ll need to integrate into our websites very soon. As they demonstrated their in-app AI Assistant concepts, I found myself thinking back to how websites started with site maps, then moved to better search, then chat, then pop-ups… We’ve been trying to get users what they want but forcing them into our own models of content organization. As applications all around us start incorporating AI assistants and co-pilots, our company websites are going to need to keep up. And we’ll need to train them on all the available content to answer our customers’ questions when they want and how they want.
   

3. We need to assign AI Program Managers / Change Agents - I’ve observed lots of people trying to learn and experiment with AI at lots of companies. But for our AI efforts to make real progress,  we can’t expect our teams to find free time to educate themselves and change long-ingrained habits and processes. We also can’t put just $10k and an intern on the task. We need to put real people and real money to keep us on the learning edge, looking for how we can make leaps forward and facilitate change management. We’re going to need some more dedicated resources to make big changes beyond just ChatGPT for blogs. 
   

4. Maybe Marketing Can Make Independent Lead or Rev-Generating Tools - There was something really interesting about the “Anyone can create their own GPT and publish it on our marketplace” vision OpenAI shared today Because the new GPTs can be created with natural language, not code, in theory, marketing departments could create their own GPTs — something useful to their target market — and use the tool as a lead generation (or revenue generating) engine to bring interest and people back to their company. At Atlassian, we rode the wave of Agile education. I once heard that Twillio’s top-performing asset was a blog about how to hire the best developers. Marketers have long created tools wider than their products to bring wider audiences into their ecosystem. This new medium of GPTs might be an interesting lead gen experiment for innovative marketing departments. Or perhaps just a side hustle for innovative marketers.
   

5. We’re going to get sick of the six default voices pretty quickly - Much less strategically, I groaned when OpenAI announced there were six different voices available for the audio components of GPT-4 Turbo. As someone who has listened to dozens, possibly hundreds of different potential voices for advertising and brand videos, I bet we get pretty sick of the six default voices developers can use pretty fast. I can see brands wanting custom voices to differentiate their use of GPT-4 Turbo in their customer experiences from their competitors. 

Overall, I really enjoyed the developer keynote. I suggest you watch it. Any company at this kind of scale that has Satya Nadella do a walk-on and call them “magic deserves an hour from strategic, forward-minded CMOs.