Tencent Technology News on April 21, during the second live speech of the "2023TED" conference held in Vancouver, Canada, Greg Brockman, co-founder of artificial intelligence start-up OpenAI, showed many new features of the popular chatbot ChatGPT and introduced the progress made in supporting the technology large language model GPT-4. Brockman believes that artificial general intelligence (AGI) has ushered in a historic moment.
Brockman projected his laptop onto the big screen in the TED Theater, showcasing an exciting, yet-to-be-released array of ChatGPT plugins.
For example, ChatGPT can help users make dinner recipes, generate images of prepared dishes, tweet about the dish, and build corresponding shopping lists on Instacart. In doing these things, the user does not need to leave the chatbot interface.
Brockman then showed off other new features of ChatGPT, such as the ability to fact-check his work. It can interpret spreadsheets full of data even when given relatively vague instructions. Brockman says the idea was to teach machines to "align" with the user's intent.
Just as you might teach a child to complete a task without giving precise instructions, AI gradually learns to apply its knowledge to new situations through a feedback loop with the user. As it takes increasingly difficult tasks, he hopes this will help shape a deep, trustworthy model of cooperation between humans and AI. Humans will be the managers and supervisors of this work, while machines will be responsible for performing specific tasks.
Brockman acknowledges that many people are nervous about the potential of AI general, but he still believes that the technology will create a better world for everyone. He said the key to achieving this goal will be broad participation and input from people figuring out how to put safety barriers for AI.
After the talk, TED head Chris Anderson took the stage with Brockman to dive into ChatGPT's timeline and the concerns of many in the tech industry and beyond, what are the risks of bringing such a powerful tool to the world. Was the launch of GPT-4 a responsible decision, or a reckless one?
Brockman's position on this is that the best way to do this is to release the machine to the public before it becomes super powerful, see how it actually works, and then "let reality slap it in the face." Since then, he said, our collective responsibility has been to provide feedback to AI as it progressively improves.
Brockman concluded by saying that AI is poised to change every aspect of how we use computers, and it's time for us all to familiarize ourselves with the technology. (Golden Deer)