As I stepped off the plane from San Francisco, I couldn't help but feel a sense of excitement and anticipation. Having just attended what was claimed to be the first-ever conference on generative artificial intelligence (AI)1, I was left with a powerful impression of what the future may hold. Thanks to our venture capital investments I always felt like I have a kind of crystal ball. This is because innovation that is driven by cutting-edge startups can provide a glimpse into the future.
This time, it felt no different. In this opinion piece, I share my thoughts on generative AI and how I expect it to impact individuals, businesses and society as a whole.
How big will this disruption be?
As I look at the current state of generative AI, particularly the likes of OpenAI’s ChatGPT, I can't help but think of its potential for innovation and disruption on a scale comparable to the introduction of email, the internet, and the smartphone. And it's not just the magnitude of this innovation that's remarkable, but also the speed at which it's being adopted. A UBS study based on Similarweb data ranks ChatGPT as the first product or service to have reached 100 million active users within two months after launch.
Chat-based AI interfaces are already challenging traditional search methods. They may completely replace them. What’s more, personalised chats with AI tools have the potential to at least partially substitute traditional content formats like articles, books and videos. As we enter this new era of generative AI, it seems to me that companies like OpenAI, Google’s LaMDA and others have the potential to become new "AI operating systems" that power a range of new services and that help to create new companies. I see this as comparable to how the smartphone enabled the likes of Uber and WhatsApp.
Will generative AI soon be everywhere?
I expect generative AI to quickly become ubiquitous.
Text generation based on Large Language Models (LLMs) – like OpenAI’s GPT – is becoming available within many types of apps, from internet browsers2 to office tools like Microsoft 3653, to software development tools4.
I anticipate that the integration of existing software interfaces – so-called application programming interface (APIs) – into AI tools, will further enhance their overall capabilities. It doesn’t take much imagination to also visualise speech-to-speech interfaces with chat-based AI tools. And it is entirely possible that these interfaces will start to exhibit empathy and emotions (without being sentient), at least for some applications. In my view, it is therefore only a matter of time before it will become increasingly difficult to distinguish a conversation with an intelligent chatbot from a discussion with a person.
Powerful generative AI tools for images (like OpenAI’s Dall-E – which produced the image, above5 – and Stability.AI’s Stable Diffusion) are spreading fast, with the latter already available in Adobe Photoshop through a plug-in.
Leading firms in the industry have already launched or announced prototypes of AI tools for other types of media, including video and music.
How will this type of AI change how we work?
With the rapid progress in this field, it is my personal view that Generative AI has the potential to fundamentally transform the way many people work. This is an idea that may excite some. Others might wonder what this means for them. Individuals whose jobs involve listening, speaking, writing, typing, or reading (which includes most people working in an office) may experience significant changes.
ChatGPT illustrates how people can interact with an AI tool. The same approach can be applied everywhere else: instead of opening an empty presentation or an empty text document, generative AI can help to fill those documents with content from the start. Beautiful.ai’s DesignerBot is a prototype of this. In the future, users will be able to prompt the AI to change those documents iteratively and in real-time – in the same way as someone prompts ChatGPT to alter a text that it has generated.
I see this rapid fire trial-and-error approach becoming the main way how we will create things in the future. Picture an architect adding the command “add a bathroom”. Something similar could work for music, videos, coding, website and product design – resulting in greater speed, creativity and quality.
Tools like ChatGPT will not only change how we work, but they will also give new opportunities to all of us. They will make it easier to create something. And they will lower the entry barriers to launching and creating a business.
What are the capabilities of generative AI?
While known as generative AI, these tools are capable of much more than generating content from a short prompt. ChatGPT, for example, can generate text in various forms. It can create outline for presentations, it can create books, white papers, essays, speeches and even poems. ChatGPT can also find errors in text, summarise it, complete or rewrite it, or change its style. It can make text easier to understand. It can translate text into multiple languages. It can extract information from text, classify text and analyse its sentiment – all of which can be very useful as input for other applications. It can find errors in computer code and it can complete it. The list goes on.
Other AI tools are or will soon be able to do similar things for images, music and videos. What I find particularly interesting is that the openendedness of generative AI models means that some of their capabilities are only discovered months or years after they were launched.
What about the shortcomings of those tools?
As I reflect on some of the criticisms of today's AI tools, I am reminded of the early days of other breakthrough technologies. From the Apple II to Windows 1.0, from AOL to my first Nokia phone, these products all had their share of shortcomings. But what the history of technology has repeatedly shown us is that exponential progress leads to unforeseen improvements in products and services. I firmly believe that the same will be true for the most pressing issues facing today's Generative AI.
Of course, there are valid concerns to be addressed, such as the potential for so-called hallucinations. These are machine answers that appear to be true but are not. There is also the issue of the black-box nature of these systems: it is often difficult or impossible to make transparent how an AI tool came to the results it provided.
There are some missing capabilities, such the inability of some AI chatbots to reliably do calculations. And there is risk of harmful, biased or inappropriate behaviour which must be scrutinised: who determines the parameters of the AI? And to what end? Moreover, the limitations of the data used to train these models, the energy consumption required to power neural networks, and the regulatory, legal and ethical questions surrounding this new technology6 cannot be ignored. Yet, from my vantage point, many industry participants are already taking steps to mitigate these concerns. I expect the growing attention and investment that the generative AI industry is attracting to further accelerate progress in this field.
Nevertheless, generative AI tools do have some notable shortcomings that I expect to persist for a long time. These tools lack an understanding of content and the ability to reason or to exhibit intuition like humans. Despite their power, they are still trained for relatively narrow tasks. I am convinced that progress will be made, and machines will eventually achieve general artificial intelligence, but this point in time is likely still several decades away.
What will be the impact on the job market?
As someone with a background in economics, I am acutely attuned to the potential impact of emerging technologies on the job market. The rise of generative AI is no exception.
On one hand, there is no denying that this technology has the potential to significantly increase productivity and reduce costs in industries where automation is feasible. This could result in job losses or, at the very least, fewer new jobs created in those sectors. On the other hand, it's worth considering that lower prices resulting from the use of generative AI could lead to increased access to related products and services for consumers from all walks of life and across various industries (e.g. for education and healthcare), which could drive up production volumes and ultimately lead to more job opportunities.
Generative AI may lead to deflation in industries that can be further automated. However, as most central banks are unlikely to allow deflation to occur for the overall economy, this could ultimately result in a price increase for those goods and services that do not benefit from AI-based automation. This, in turn, could prompt a re-evaluation of roles that require uniquely human skills, such as empathy, intuition, and critical thinking, or that are simply too important to be automated.
My optimistic inclination leads me to believe that these dynamics will help to dampen job losses in industries that permit automation and will create new jobs in others. However, the future is unpredictable, and the short and long-term effects of emerging technologies on the job market remain ambiguous.
Is this the peak of a hype cycle?
The current buzz surrounding ChatGPT and its explosive adoption appears to be following the familiar pattern of the so-called hype cycle that I have observed with other disruptive technologies. I do not believe that we have reached the peak of the initial hype yet, as I anticipate the rapid pace of developments in generative AI to continue. And it is only a matter of weeks since ChatGPT was launched to the public. When this hype cycle inevitably subsides at some point, I expect that some doubts will arise regarding the potential of this technology. Nevertheless, I have learned from past disruptions that it's crucial to look through the ups and downs of the hype cycle and to focus on the mid to long-term impact that a new technology such as generative AI will have.
What will be the impact on society?
I am cautiously optimistic for the short and mid-term adoption of generative AI, but I cannot ignore the significant unknowns we face as a society regarding its longer-term implications. My personal crystal ball leaves me with many important questions, such as: will the widespread use of generative AI ultimately foster greater creativity and intelligence, or will it stifle them? Will it enable better knowledge sharing or fuel rampant misinformation?
Most significantly, will it mitigate social inequality, or will it exacerbate the concentration of power and wealth among a select few? In my view, the answers to these questions are not predetermined; rather, they will be shaped by the choices we make as a society about how we use these powerful new tools.
A cautiously optimistic conclusion
I believe that the rapid development of generative AI is poised to revolutionise our lives, and in particular our professional lives, in ways we cannot yet fully imagine. This will create new opportunities for many people around the world. The economist in me recognises the potential for both disruption and opportunity in the job market. And the technologist in me is fully cognizant of the risks and the potential of this new technology. Overall, I am cautiously optimistic that existing concerns can be effectively addressed, and that generative AI will have a significant and positive transformative impact on society as long as it is approached responsibly. This is the responsibility of all of us as society. And it is our responsibility as investors in companies that create and apply this new technology.
All content in the document is based on the original ideas of the author. The text has been reviewed by the author. The editing of the document was supported by ChatGPT.
1 Jasper AI’s GenAI conference, it was sold-out with 1,200 participants
2 On February 7, 2023, Microsoft announced the integration of OpenAI’s technology into its Edge browser for content summaries and chat functionalities
3 On February 10, 2023, The Verge reported that Microsoft is preparing to integrated OpenAI’s capabilities into Word, PowerPoint and Outlook
4 e.g. Github Copilot or Replit Ghostwriter
5 Artwork by OpenAI’s Dall-E produced in response to the instruction: "Draw me a minimalist but sophisticated San Francisco Skyline”
6 For instance, concerning the protection of intellectual property rights related to the data utilized in training generative AI models