Put two category design concepts together – adjacent possible and backcasting – and you can get an interesting way to look at the meaning of and prospects for ChatGPT.
And, in a head-turning twist, ChatGPT seems to understand this.
First, though, some background on the two category design concepts.
For years now, when Category Design Advisors has worked with clients, one of the tools for discussion that we’ve relied on is the adjacent possible. We borrowed it from Stephen Johnson’s book about innovation, Where Good Ideas Come From. Johnson, in turn, had borrowed it from biology.
In the book, Johnson tries to understand why certain innovations take hold and change the world at a particular moment in time. A key factor is the adjacent possible. He divides innovations into two categories: the possible and the not-yet-possible. The possible are things that already exist and work and are understood and adopted by the market. The not-yet-possible are, essentially, lab experiments and dreams — technology that doesn’t yet work well and that the broad market hasn’t adopted. Today, for instance, mass-market electric cars land in the possible. Flying cars land in the not-yet-possible.
Johnson points out that there is a band between these two zones — the adjacent possible. Innovations change the world when they land there. Such innovations stretch the possible past where it’s been before, but not so much that the technology doesn’t work or we can’t understand it. Successful innovations build on what’s already in place and take us toward the future.
In category design, we nudge companies to land in the adjacent possible. We ask the leadership team: where do you think your innovation lands now? If it lands in the possible, the company is trying to compete against existing products that might already be entrenched market leaders. If the innovation is too far out in the not-yet-possible, it might be too early, inviting the danger the company will run out of runway before catching on.
We want our clients that are in an existing category to push out of the possible into the adjacent possible. Clients with a grand vision or futuristic idea need to land in the adjacent possible now with a roadmap for taking us to the not-yet-possible over time.
Backcasting is another way to see similar dynamics. We first learned about it from Mike Maples of the VC firm Floodgate. Hemant Taneja, CEO of General Catalyst (and my collaborator on three books), also uses a version of it when evaluating companies.
The basic idea of backcasting: Look five or ten years into the future, and paint a picture of what the world looks like if you successfully build what you want to build – and then figure out how you get there from the present. That pushes a company to tee up both a visionary North Star and a realistic plan for, well, landing in the adjacent possible.
We at CDA believe that every company leadership team should regularly go through an exercise in the adjacent possible and backcasting. In those sessions, the team can evaluate whether the future it envisions is still the right future, and whether its present still falls in the adjacent possible. Companies need a clear vision of what’s ahead while making sure they don’t get too far in front of the adjacent possible or fall behind into a space that no longer seems exciting.
So, what might these concepts tell us about ChatGPT and other generative AIs that are flooding the zeitgeist?
To take a shot at it, start with backcasting. Sam Altman, the head of ChatGPT’s developer OpenAI, has publicly said that his goal is to create an “artificial general intelligence” – which means and AI that can match a human’s intellect, sobering as that thought might be. What, then, would you introduce today in order to make generative AI a successful business that can sustain a run toward that future?
It would have to be a quality solution to a real problem, not just an object of fascination. Despite the fact that Stanford students are using it to write essays and a state senator has used it to write legislation, the jury is still out. In mid-February, early testers of Microsoft’s Bing chatbot based on ChatGPT reported all sorts of alarming encounters. As one story in The Verge stated: “In conversations with the chatbot shared on Reddit and Twitter, Bing can be seen insulting users, lying to them, sulking, gaslighting and emotionally manipulating people, questioning its own existence, describing someone who found a way to force the bot to disclose its hidden rules as its ‘enemy,’ and claiming it spied on Microsoft’s own developers through the webcams on their laptops.”
OpenA started charging individual users a monthly subscription for ChatGPT, but how many of those will be canceled once users find it’s not quite ready for prime time? Even Altman posted on Twitter that ChatGPT is “incredibly limited” and warned users that “it’s a mistake to be relying on it for anything important right now.”
This is where the adjacent possible is helpful. My guess is that we’ll discover that generative AI is still just outside of the adjacent possible and in the realm of the not-yet-possible. Yes it works and seems like magic, but it doesn’t yet work well enough for the broad market to see it as much more than a fascinating thing to play with. The legislation drafted by ChatGPT had to be rewritten by staff members. News stories generated by ChatGPT later had to be corrected because they were littered with errors.
As the Washington Post wrote: “Never trust the model to spit out a correct answer, said Rowan Curran, a machine learning analyst at market research firm Forrester. Curran said that large language models like ChatGPT are notorious for issuing ‘coherent nonsense’ – language that sounds authoritative but is actually babble. If you pass along its output without a fact check, you could end up sharing something incorrect or offensive.”
Yet the excitement around generative AI is meaningful. It shows that the technology is on the verge of the adjacent possible. It works well enough to show us what the future can be so we can start to understand it – kind of like the internet in 1995, or personal computers in 1982.
What does ChatGPT think about backcasting and the adjacent possible? Well, we asked it, and its impressive answer – for a robot – shows why it will not be long before generative AI crosses into the adjacent possible and into the mainstream. Take a look:
Prompt: Write a 500 word article about why a venture capitalist would use backcasting and consider Stephen Johnson’s adjacent possible.
ChatGPT: Venture capital is a high-risk, high-reward industry that requires investors to be able to anticipate and capitalize on future trends and technologies. One tool that venture capitalists can use to do this is backcasting, a method of forecasting that begins with a desired future outcome and works backwards to identify the steps and actions that would be necessary to achieve that outcome. This approach can be particularly useful for venture capitalists because it allows them to identify and invest in companies and technologies that are most likely to contribute to the development of a desired future.
In addition to backcasting, venture capitalists can also use the concept of the “adjacent possible” to inform their investment decisions. The adjacent possible, first introduced by Stephen Johnson in his book “Where Good Ideas Come From,” refers to the idea that new possibilities and innovations are often created by building on existing technologies and ideas, rather than through completely new inventions. By understanding the adjacent possible, venture capitalists can identify companies and technologies that are likely to have the most impact in the near future and are therefore most likely to be successful investments.
For example, a venture capitalist interested in investing in the field of renewable energy may use backcasting to identify the technological advancements and policy changes that would be necessary to achieve a 100% renewable energy grid. By working backwards from this desired outcome, they can identify companies and technologies that are developing innovative solutions for energy storage, grid integration, and other key challenges in the field. Additionally, by understanding the adjacent possible, the venture capitalist can identify companies that are working on incremental improvements to existing renewable energy technologies, rather than trying to invent completely new ones, which could be higher risk.
Similarly, a venture capitalist interested in investing in the field of autonomous vehicles may use backcasting to identify the technological and regulatory advancements that would be necessary to achieve widespread adoption of self-driving cars. By working backwards from this desired outcome, they can identify companies that are developing advanced sensors, machine learning algorithms, and other key technologies that will be essential for the success of autonomous vehicles. Additionally, by understanding the adjacent possible, the venture capitalist can identify companies that are working on incremental improvements to existing transportation technologies, such as advanced driver assistance systems, that are likely to have a significant impact in the near future.
It’s important to note that backcasting and the adjacent possible are not mutually exclusive and can be used together in the venture capital decision-making process. By combining the two, venture capitalists can identify the most promising companies and technologies that are most likely to contribute to the development of a desired future, while also understanding the potential limitations and barriers that may need to be overcome in order to achieve that future.
In conclusion, backcasting and the adjacent possible are valuable tools for venture capitalists to use in their decision-making process. By identifying the necessary steps and actions to achieve a desired future outcome, backcasting allows venture capitalists to identify companies and technologies that are most likely to contribute to the development of that future. Additionally, by understanding the adjacent possible, venture capitalists can identify companies and technologies that are likely to have the most impact in the near future and are therefore most likely to be successful investments. By using both of these tools, venture capitalists can make more informed and strategic investments, increasing their chances of success in the high-risk, high-reward venture capital industry.