The fields that generative AI addresses—knowledge work and creative work—comprise billions of workers. Generative AI can make these workers at least 10% more efficient and/or creative: they become not only faster and more efficient, but more capable than before. Therefore, Generative AI has the potential to generate trillions of dollars of economic value.
Sequoia, Generative AI: A Creative New World
‘Generative AI’ is described as “unsupervised and semi-supervised machine learning algorithms that enable computers to use existing content like text, audio and video files, images, and even code to create new possible content”.
In the last few months we’ve seen a number of advancements in text to image generators, enabling users to create art from text prompts: Open AI’s Dall-e 2, Midjourney and Stability, an open source solution. While the technology showed promise and created a sense of ‘magic’ for users, it was more ‘playful’ rather than ‘useful’. GPT-3, the Large Language Model (LLM) for generative text, has been in development for years, but was not made available for public use.
That changed last week with the launch of ChatGPT, a new generative AI tool released by Open AI, enabling users to ‘converse’ with AI using natural language. Users can write a text prompt or a question and receive email drafts, college level essays and long form blog posts. ChatGTP can also tell jokes, write poetry and even code. With the hype at its peak, ChatGPT is even being talked about as a possible replacement for Google search. Why sift through links if I can get the answer crafted?
While it shows real potential, ChatGPT has a number of limitations. It can struggle with simple math, deliver factually wrong answers very convincingly and in some cases can be exploited for malicious intent, like finding vulnerabilities in smart contracts. ChatGPT’s answers are based on scraped Internet data (up to 2021), which means it’s not yet ‘connected to the live Internet’ and can’t help in certain use cases. It’s also difficult for startups operating in this space to create real defensibility as most are built on the same core APIs. And finally, there are open questions about the use of training data (for text or art) without referencing and credit to the source. OpenAI’s Sam Altman said it’s working to fix these issues.
Investing in generative AI
At Remagine Ventures we believe that in the long run Generative AI tools will exist for every role in the company (coding, marketing, recruiting, sales, finance, etc) and ultimately every vertical, from Saas to gaming. They won’t replace humans, at least not right away. In the same way we use Grammarly to write better emails, we’ll have a set of tools to help us with task automations at work.
We already invested in two startups in the space. HourOne, the text to video company based on real human characters (in 2019), and our latest investment, Munch which uses AI to automatically create highlight clips and video overlays (Oct 2022). In addition, several of our companies are exploring how they may leverage Generative AI to improve their user experience.
We’re excited about the pace of innovation in this space and what it means for the future of content creation in gaming, entertainment, populating the Metaverse with 3D content and more. We plan to continue writing, learning about and investing in this new era of ‘Creative Automation’.
In my previous post, I touched on the potential exponential growth of generative AI on the volume of content. If you believe that the Internet will the flooded with new content in all formats, most of it being synthetic/ machine generated, is there a world in which suddenly verifying the human source of the content can rely on something like the Blockchain? Will students have to submit their essays via blockchain service? It’s too early to tell, but with this new technology come a whole new set of challenges and opportunities.
The post was originally published in the Remagine Ventures Dec newsletter.