ChatGPT to Emerge as a "Super App" - New Features and the Future of Assistants
ChatGPT to Emerge as a "Super App" - New Features and the Future of Assistants
This really is the inside story on where ChatGPT is headed.
“This technology looks different. Different from anything we anticipated.” - OpenAI’s Greg Brockman.
TLDR;
Greg Brockman demonstrated how ChatGPT Plugins will usher in a new app store adding far more value beyond its current capabilities. See my summary of the 9 top new features here.
His video demonstrates several new features that go well beyond anything you will find in the iOS store.
Brockman also explains several important principles about ChatGPT’s purpose, training, and expected performance.
ChatGPT is on course to become the virtual assistant we wanted all along. Siri, Alexa, Google Assistant, and Bixby all came up short, but ChatGPT is borrowing the vision of earlier innovators and using a new technology to overcome key barriers. And it has a revenue model!
Greg Brockman, the president of OpenAI, delivered a TED Talk this week that offers more clarity on ChatGPT’s future than anything we’ve seen before. It is on a course to become a “super app” or “everything app” and might be the virtual assistant we’ve wanted all along.
Yesterday, I broke down the nine most interesting features Brockman demonstrated. Today, we go deeper into the product philosophy and the emergence of ChatGPT as an assistant.
Part 1: ChatGPT Product Strategy
In addition to the new features, Brockman revealed several key concepts guiding the ChatGPT product strategy during the TED Talk. Four of these include:
How conversational interactions backed by an intelligent “assistant” make it easier for users to get what they want.
How a natural language UI will complement existing UIs and services.
How ChatGPT is trained to understand requests and what applications will fulfill the user intent.
How ChatGPT is trained to anticipate unarticulated user needs based on the conversation context.
We have all become accustomed to ChatGPT as a search, summarization, question-answering, and creative tool. However, the next development phase will greatly extend the capabilities for task execution. This was ushered in by the announcement of ChatGPT Plugins in March.
You will notice in the video that Brockman refers to these as applications and not plugins. That seems more accurate regarding their functionality, while the plugins term fits the description of digital services that augment the core ChatGPT application. Regardless, you may recall that Sam Altman said on November 30th that he expected ChatGPT to become a “helpful assistant.”
Brockman shared more on this idea of an assistant that selects the right tools for you in his Ted Talk.
You can see that ChatGPT is selecting all of these different tools without me having to tell it explicitly, which ones to use in any situation. And this, I think, shows a new way of thinking about user interface. We are so used to thinking of, well, we have these apps. We click between them. We copy / paste between them, and usually it’s a great experience within an app as long as you kind of know the menus and know all the options.
And by having this unified language interface on top of tools, the AI is able to sort of take away all of those details from you. So, you don’t have to be the one who spells out every little piece of what is supposed to happen…
And the thing that’s really interesting is that the traditional UI is still very valuable, right? If you look at this (i.e. an Instacart app shopping screen), you can still modify the actual quantities. And that’s something that I think shows that they’re not going away—traditiona UIs. It’s just that we have a new, augmented way to build them…
Now, the important thing about how we do this. It’s not not just about building these sorts of tools. It’s about teaching the AI how to use them.
Part 2: The Virtual Assistant We Were Promised
The four concepts listed above are not new to anyone that closely followed the rise of voice assistants. The original vision for Siri, outlined by Adam Cheyer and Dag Kittlaus, followed similar ideas even though Apple decided to curtail these ambitions after acquiring the company and launching the service in 2011.
You could also take the bullet points above and replace ChatGPT with Viv. You then have the essence of Cheyer’s and Kittlaus’ Siri follow-up company Viv Labs which was acquired by Samsung in 2016 to power its Bixby voice assistant.
Around the same time, Jeff Bezos attempted to implement a version of these principles with Amazon Alexa. In early November 2022, Amazon essentially admitted that Alexa would not be “everywhere” after all. On the final day of that month, ChatGPT launched. It now seems clear that “ChatGPT everywhere” is a real possibility. It has far more momentum behind the broad vision summarized above than any of its predecessors.
One area where ChatGPT and the large language model (LLM) approach are clearly different from these earlier visions is in its generative capabilities. While it is clearly better than Alexa or other predecessors at a variety of assistant tasks that rely on information retrieval or task execution, it can also create entirely new content. That is novel and may be critical to its success.
Earlier assistants made it easier for users to perform tasks than established methods. They also centralized some of the executive functions by bringing it all into one conversational interface which provided more flexibility.
However, these assistants were fundamentally a replacement technology. You could still perform all of these functions in other places. There was a promise of more personalization that might augment the user experience through the assistant, but the outcomes were not novel. The value was based on a better user experience.
ChatGPT adds novel generative functionality from writing email responses and essays to summarization and image generation. Yes, you can now perform those tasks in other applications—most of them based on OpenAI technologies—but few consumers have habits or loyalty formed around them.
ChatGPT was the first experience with these features for many people. That means they are forming new habits around new activities as opposed to displacing existing habits. Thus the low-friction rise to 100 million monthly active users.
You can also use ChatGPT to answer questions for you about specific legal documents, software code, and data. Earlier assistants were not expected to be that versatile. The idea of what an “everything app” or “everything assistant” can be has clearly expanded.
Part 3: Will ChatGPT Make It?
The history of voice assistants is a gradual regression in scope to execute select user tasks. Voice assistants today are very useful for some tasks and in some situations. They are also daily use applications for many consumers.
However, voice assistants didn’t do much to break down the siloed nature of our digital services and eventually became viewed as UI affordances for certain devices—iPhone for Siri, Android phones for Google Assistant, Samsung phones for Bixby, smart speakers and TVs for Alexa.
ChatGPT has the potential to maintain the perception of full device independence and aggregate access to a wide range of daily-use digital services. Everyone is enthusiastic about this vision today, but they were also excited about Siri, Alexa, Google Assistant, and Viv at one time. There is still a lot of work ahead of ChatGPT. Take not that each its predecessors fell short on realizing the vision.
One thing ChatGPT has going for it is the rapid accumulation of a lot of users. Another is revenue. ChatGPT crossed the chasm into some very large existing markets and helped establish new segments with potentially attractive economics.
ChatGPT offers so much value that OpenAI had a solid business model from day one. Some applications will pay to integrate and use ChatGPT’s foundation model. Some consumers are paying a subscription fee. It also appears that some plugin makers will have a clear path to monetization which we can assume will include flow-back fees to OpenAI. These factors were almost entirely absent from the early launch of previous virtual assistants.
LLM-based solutions are not free from problems. Truth, privacy, trust and safety, content rights, and other challenges lie ahead. However, there is now little doubt about market interest in the functionality. And if it can do some of the things Brockman just demonstrated, I expect ChatGPT will quickly become my most-used application.
I recommend you watch the first 16 minutes of the video if you want to see where this is headed. Great stuff! 🤯
Check out This Interview from 2018
If you would like to learn about the history of virtual assistants going back five decades, I strongly recommend you listen to my 2018 podcast interview with Adam Cheyer, one of the three founders of Siri, and co-founder of Viv Labs. It will provide some valuable context about how ChatGPT is fulfilling a long-held virtual assistant vision and building on earlier advances.
