“Extremely impressive.” “Incredibly rich.” “Super exciting.”
That’s how tech leaders are describing ChatGPT, the new conversational chatbot model released last week by OpenAI.
The bot builds on existing GPT natural language technology developed by OpenAI, the San Francisco-based organization formed by tech leaders Sam Altman and Elon Musk, and backed by Microsoft, which helps power the back-end cloud computing for OpenAI products.
But the latest update sparked huge reaction over the past week — and more than 1 million users — given how ChatGPT is able to quickly answer complicated questions and instantly produce content.
For example, we asked it to write a short news story about the Seattle startup ecosystem, and the result was relatively on point. It can easily do homework or create weight loss plans or even draft lines of code. The tech is making college professors rethink how they create exam questions.
“There’s a certain feeling that happens when a new technology adjusts your thinking about computing,” said Box CEO Aaron Levie. “Google did it. Firefox did it. AWS did it. iPhone did it. OpenAI is doing it with ChatGPT.”
But there are already concerns being raised about the technology, part of a newer group of language models called “generative AI,” such as the ability to discern when something is created by a machine or not, or misuses and unintended consequences.
And for now, the bot is not always accurate or truthful. Seattle tech vet Jensen Harris, co-founder of Textio, shared a response from ChatGPT describing “The Ohio and Indiana War,” which, of course, did not happen.
We reached out to Seattle-area tech investors, entrepreneurs, and researchers to get their quick takes on ChatGPT and its potential. Read on for their answers.
Greg Gottesman, managing director at Pioneer Square Labs:
- “We are seeing the first inning of technology that will turn information workers — everyone from journalists to marketers to programmers — from the generators of content and code to editors of it. I think over time it will change the type of work we do and the way we work.”
- “Startups, of course, have a major role to play. I think the big companies may control much of the underlying technology, but there will be plenty of room for startups creating the applications that stem from this technology and building on top of it and enhancing it in ways that enable purpose-built, vertical solutions.”
Yejin Choi, University of Washington computer science professor
- “I think ChatGPT is both surprising and unsurprising. What it is capable of doing, when it is successful to carry on, is super exciting and impressive. It does reach a new milestone that the community at large didn’t necessarily consider as feasible within this year.
- “It is also unsurprising in that it does make mistakes easily when users try to fool the system deliberately. This weakness to the adversarial queries might turn out to be an extremely challenging problem to solve in the coming years, as such queries challenging the system with less frequent, novel situations. The current deep learning training paradigm relies a great deal on frequent patterns or patterns of previously-seen situations, and lacks common sense capabilities to reason fluidly about trivial yet novel/unfamiliar situations.”
Geoff Harris, managing partner at Flying Fish Partners
- “This has the potential to be game-changing technology. What is less clear right now is the strategy behind it. This could simply be a case of OpenAI demonstrating what is possible on top of their core platform. They could extend this to be an app model for others — one where there are vertical implementations akin to Alexa Skills, but one where the discoverability challenges are solved because the UX is visual not simply audio.”
- “An alternative path is that this is OpenAI’s ‘Office’ to their ‘Windows’ — the killer first-party app implementation of GPT. In this path, many incumbents, most notably Google, should be nervous.”
Richard Yonck, futurist and emerging tech consultant
- “There are certainly a range of business use cases, but I’m concerned about its intentional misuses (e.g., multimodal deepfakes and other forms of automated cybercrime) and unintended consequences (information and its expression converging on the mean as well as the atrophying of writing skills).”
David Shim, founder and CEO of Read AI:
- “It won’t be a surprise to see in the next 24 months, multiple billion-dollar companies built on top of OpenAI’s foundational models. The startups that will be the most successful won’t be the best at prompt engineering, which is the focus today, rather success will be found in what novel data and use case do they incorporate into OpenAI’s models. This novel data and application will be the moat that establishes the next set of AI unicorns.”