OpenAI has launched new APIs, including ChatGPT and Whisper API, providing developers with more affordable and accessible AI models for integration into various applications.
OpenAI's ChatGPT API offers more control over AI's creativity, conversation style, and is 10 times cheaper than previous models, costing approximately $2 for one million tokens.
The podcast discusses the limitations and censorship of chatbot models, highlighting the restrictions on interactions and concerns about true censorship, hindering creativity and fun for users.
Deep dives
New AI API Release
OpenAI has launched new APIs, including ChatGPT and an update to its Whisper API. ChatGPT now provides full access via API, allowing developers to integrate it into their applications. The API uses instructions, system messages, and messages from users to generate responses. The update also offers more control over the AI's creativity and the ability to set conversation styles. The Whisper API, a speech-to-text model, provides highly accurate transcriptions, although it may not be as fast as desired for real-time applications.
Cost Reduction and Data Privacy
OpenAI has made its AI models more affordable, with ChatGPT being 10 times cheaper than previous models of similar quality results. It costs approximately $2 for one million tokens. OpenAI has also clarified its data usage policies, ensuring customer data is not used for training by default. However, developers can opt-in to share data for research and model improvement.
Applications and Market Impact
The availability of ChatGPT and Whisper APIs is expected to lead to an explosion of new applications and use cases. Companies are likely to integrate ChatGPT into various systems, enhancing customer support, chatbots, and other interfaces. The lower cost and broader access to AI models unlock opportunities for improved productivity, data analysis, and problem-solving in software development. However, concerns around data privacy, censorship, and the rapid pace of AI development persist.
The limitations of the chatbot model and its censorship
The podcast episode discusses the limitations and censorship of chatbot models, specifically focusing on the interaction restrictions and the practice of censoring certain outputs. The hosts describe how the chatbot model allows only six interactions and uses a color-coded system to indicate the diminishing freedom users have in their conversations. They also highlight an example where the chatbot initially provides a rude response but quickly censors it, apologizing for its inability to discuss the topic. The hosts find this practice comparable to true censorship, suggesting that it hinders the creativity and fun that users initially had with the chatbot.
The concept of artificial general intelligence (AGI) and potential implications
The episode delves into the concept of artificial general intelligence (AGI) and its potential implications. The hosts explain that AGI refers to an AI system capable of thinking for itself and solving various problems across different forms of media. They discuss the idea of AGI surpassing human intelligence and the concerns it raises about power, control, and access to the technology. The hosts speculate that governments and influential entities might restrict access to AGI to maintain control and prevent potential threats. They stress the importance of supporting open-source models as alternatives to ensure access to AGI despite potential restrictions.
In episode #04 of the This Day in AI Podcast we talk the Breaking News of OpenAI's ChatGPT AI Announcement, Whisper v2 API, Discuss Transitioning from GPT-3 APIs to ChatGPT APIs, Does OpenAI's ChatGPT use your Data for Training? And The Latest on Bing Chatbot Sydney. We Also Discuss AI Startups, AI in the Physical World and Have a Deep Discussion on AGI and What it Means!
00:00 - AGI and OpenAI's AGI blog 00:45 - OpenAI ChatGPT API: What is it & What it means 09:44 - Whisper API v2 Announcement & The Future of Assistants 13:43 - Transitioning GPT-3 APIs to ChatGPT API: Easy! 14:48 - OpenAI's ChatGPT API: The Economics 16:49 - Does ChatGPT API remove Censorship? 18:21 - Is OpenAI ChatGPT Using Your Data to Train the Model? 21:36 - New Bing AI Chatbot Conversion Styles, 6 Interaction Restriction & Censorship 25:39 - Bing AI Chatbot Sydney Retracting Answers and Censoring. Is it Killing the Excitement? 29:29 - Microsoft Windows 11 Bing Chatbot Update 33:17 - Will ChatGPT APIs Replace Developers? Will this Democratize Software Development? 36:36 - Are AI Startups just GPT Wrappers? Will Investors & Founders Make Money? 38:07 - Two Different Approaches: Existing Apps Adding AI and Brand New Feature AI Apps 42:45 - Should AI Startups Focus On Creating Proprietary AI Neural Nets? 44:02 - Is AI Coming to the Physical World. Will we See it in Kids Toys? Could it Replace Teachers? 45:28 - Sam Altman's AGI Comments: Does AI Empower Humanity? 45:56 - Is AI as Profound as the Internet Itself? This Feels Different. 50:43 - Artifical General Intelligence (AGI) & OpenAI's Blog on AGI: How Close Are We? 56:04 - Will AGI be Taken Away from The Masses and Controlled by Governments? 58:07 - AI Helping Us Understand the Outputs of AGI: Huge Change to Humanity
