Tech insiders are excited about ChatGPT, a chatbot that can answer questions and write essays, because it represents a significant advance in natural language processing. Many believe that ChatGPT could be a major step forward in the development of AI, with potential implications for how we interact with machines.
OpenAI’s ChatGPT Chatbot Impresses Tech Execs with Its Ability to Generate Detailed Responses and Code
Tobias Zwingmann, the managing partner of RAPYD.AI, has been using ChatGPT to generate lecture notes for his online AI courses. Zwingmann recently asked the chatbot to explain the mechanisms and workings of a machine learning technology called DBSCAN, and was impressed with the detailed response he received. He said it would have taken him a whole day to write the notes himself, but with ChatGPT it only took 30 minutes.
ChatGPT was developed by San Francisco-based OpenAI, a research company led by Sam Altman and backed by Microsoft, LinkedIn co-founder Reid Hoffman, and Khosla Ventures. The chatbot has quickly become a viral sensation, with people asking it questions like “Are NFTs dead” and requesting jokes about the tax risks of international remote work. ChatGPT’s responses, which are often (but not always) coherent, are being shared on Twitter.
In a year that has been tough for the tech sector, with layoffs, stock price declines, and crypto disasters making headlines, ChatGPT has been a reminder that innovation is still happening.
Tech executives and venture capitalists have praised ChatGPT on Twitter, with some comparing it to Apple’s debut of the iPhone in 2007. Five days after OpenAI released the chatbot, it had reached one million users.
In 2016, tech companies like Facebook, Google, and Microsoft were predicting that chatbots would be the next big thing in human-computer interaction, with the potential to order Uber rides, buy plane tickets, and answer questions in a life-like manner. However, progress has been slow, and most chatbots are still relatively primitive.
But with early adopters of ChatGPT showing that it can carry a conversation through multiple queries and generate software code, the world of natural language processing appears to be entering a new phase. This is part of a broader trend of tech investors pouring billions of dollars into startups specializing in generative AI, which refers to computers automatically creating text, videos, photos, and other media using machine learning technologies.
Brendan Burke, an analyst at tech industry data firm PitchBook, said that many early-stage investors have shifted their attention from cryptocurrencies and web3 to generative AI technologies. He added that many web3 developers have also moved from NFTs and crypto projects to open-source generative AI initiatives.
What Is ChatGPT?
ChatGPT is a variant of OpenAI’s popular GPT-3.5 language-generation software that has been designed to carry conversations with people. Some of its features include answering follow-up questions, challenging incorrect premises, rejecting inappropriate queries, and admitting its mistakes.
ChatGPT was trained on a large amount of text data, and learned to recognize patterns that enable it to produce its own text mimicking various writing styles, according to Bern Elliot, a vice president at Gartner. OpenAI does not reveal what data was used for training ChatGPT, but the company generally crawls the web, uses archived books, and Wikipedia.
Elliot said that ChatGPT is currently more of a way for OpenAI to gain publicity and show what is possible with large language models, rather than being a useful piece of software for businesses. ChatGPT is free, but OpenAI sells access to its underlying language and related AI models for businesses to use.
“ChatGPT, as currently conceived, is a parlor trick,” Elliot said. “It’s something that isn’t actually itself going to solve what people need, unless what they need is sort of a distraction.”
Ironclad, a digital contract management startup, is exploring the use of ChatGPT, a language model developed by OpenAI, to summarize changes to legal documents. This feature would be helpful for the startup’s legal clients who frequently alter documents and need to notify colleagues of the changes. Cai GoGwilt, the company’s CTO, said that ChatGPT offers more creative responses compared to other language models developed by major tech companies. Ironclad uses both GPT and RoBERTa, a language tool from Meta, in its digital document software.
LexisNexis, a legal research and data company, is also testing ChatGPT. Min Chen, a vice president at the company, said that they already use OpenAI’s GPT-3 software through Microsoft’s Azure cloud. Chen said that GPT-3 is more suitable for LexisNexis because it is an enterprise product and can be customized. However, the team has been experimenting with ChatGPT and she said that it sometimes generates “sensible answers” that are “very impressive.” Despite this, Chen acknowledged that ChatGPT has its flaws.
Min Chen, a vice president at LexisNexis, said that ChatGPT is not reliable enough as a decision-making tool for serious legal research. She cited instances where the software produces verbose answers that appear to make sense but are not factually correct. Chen also mentioned the issue of bias in AI-powered software, citing an example where ChatGPT produced song lyrics that implied women who wear lab coats are just there to clean the floor, whereas men in lab coats have the knowledge and skills required.
In a blog post, OpenAI acknowledged the limitations of ChatGPT and said that the company is working to mitigate bias problems. However, the software “will sometimes respond to harmful instructions or exhibit biased behavior,” according to the post. Sam Altman, CEO of OpenAI, also tweeted that ChatGPT is “incredibly limited” and should not be relied upon for anything important at this time.