At an MIT event, OpenAI CEO Sam Altman took part remotely and was questioned about Artificial Intelligence by Lex Fridman, a computer scientist and podcaster. Altman confirmed that OpenAI are not currently developing a fifth version of their Generative Pre-trained Transformer model (GPT), and instead, are focusing on enhancing the capabilities of their already-released fourth version, GPT-4.
Altman was asked about an open letter suggesting AI developers to pause training on models bigger than GPT-4 model for six months, and their stance on the matter. He backed up the notion of ensuring AI models are acting in accordance with human values and ethics, however, he also added that it was lacking technical precision when it came to order to put a stop on the development.
Before GPT-4, OpenAI released GPT-3 with a whopping 175 billion parameters, making it one of the largest existing language models. OpenAI has no revealed the exact number of parameters GPT-4 owns, but it’s estimated to be 1 trillion. In OpenAI’s blog post they stated that GPT-4 is now more creative and collaborative than ever before, and with its greater general knowledge and problem-solving abilities, it can solve more intricate issues with more accuracy.
To prove its prowess, GPT-4 was given a simulated law bar exam, in which it scored in the top 10 percent, unlike GPT-3.5 which achieved bottom 10 percent. OpenAI, which is one of the foremost Artificial Intelligence research laboratories, has used GPT models for various applications such as language translation, chatbots, and content creation. Altman’s comments imply OpenAI is well aware of the doubts about their GPT models and are looking into ways to tackle the problem.
OpenAI is a United States-based research laboratory founded in 2015. It initially started as a nonprofit organization but reorganized as a nonprofit with a for-profit branch in 2018. It was founded by Elon Musk, Sam Altman, and other prominent personalities, investors, and scientists to promote responsible utilization of wide-reaching Artificial Intelligence (AI). OpenAI has received strong public and private support including early investments from many technology entrepreneurs familiar with the power of AI, such as Peter Thiel, Reid Hoffman and Jessica Livingston.Projects such as their GPT-3 and the launch of their GPT-4 have pioneered applications based on the development of both language models and deep learning.
Lex Fridman is an American computer scientist and podcaster. He holds a degree in robotics from MIT and is currently a Postdoctoral Associate at the MIT Media Lab. His research has focused on autonomous driving and self-driving cars, as well as other AI-related topics like Robotics, Machine Learning, Artificial General Intelligence, Autonomous Racing and Human-Robot Interaction. He has founded robotics companies and is the host of the Artificial Intelligence podcast. Structured as long conversations with industry-leading experts, the podcast is intended to educate and inform on the implications of able technological advances and their potential to humans.
